Preprint
Review

This version is not peer-reviewed.

Bridging AI-Assisted Cancer Screening and Public Health Surveillance: Evidence and Challenges in the Philippines (2020–2026)

Submitted:

21 October 2026

Posted:

22 October 2026

You are already at the latest version

Abstract
Artificial intelligence (AI) is reshaping cancer screening globally, yet evidence supporting its clinical application and integration into public health systems remains limited in countries such as the Philippines. This narrative review synthesised documented Philippine applications of AI-assisted cancer screening from January 2020 to September 2026, examining their implications for cancer surveillance, health-system readiness, and policy. A structured review of scientific publications, government and health technology assessment documents, and grey literature revealed a limited and heterogeneous evidence base. Only two peer-reviewed studies reported separately identifiable Philippine patient data, both retrospective, addressing mammography classification and computer-aided colonoscopy. Multinational evaluations and reported commercial deployments suggested opportunities for AI-assisted detection but provided insufficient Philippine-specific evidence of clinical effectiveness and patient outcomes. No prospective validation studies in Filipino screening populations or local evaluations of AI-assisted cervical or liver cancer screening were identified. Beyond clinical validation, major barriers included predominantly opportunistic screening, incomplete population-based cancer registry implementation, fragmented digital infrastructure, workforce constraints, regulatory uncertainty, and insufficient economic evaluation. Recent national screening financing policies provide opportunities for integrating innovative technologies into established healthcare pathways. These findings underscore that the transition toward AI-assisted cancer screening requires not only technological advancement but also locally validated evidence, strengthened population-based surveillance, equitable access, and coordinated public health governance.
Keywords: 
;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  

1. Introduction

Neoplasms were the second leading registered cause of death in the Philippines in 2024, accounting for 77,504 of 701,884 registered deaths, after ischaemic heart disease and ahead of cerebrovascular disease [1]. That figure is a count of registered deaths coded to neoplasms of any type. Meanwhile, the Global Cancer Observatory's modelled estimate for the same year, which projects national mortality trends and applies subnational registry incidence rates to the 2024 population, gives 86,338 cancer deaths and 149,852 new cases, with breast, colorectal, lung, prostate and liver cancers making up just over half of new diagnoses [2,3]. The two sources measure different aspects of cancer mortality and should not be treated as equivalent: the Philippine Statistics Authority reports officially registered deaths, whereas GLOBOCAN estimates the total national burden using statistical modelling. Each figure is therefore used here only for the outcome it measures. Although the earlier GLOBOCAN 2022 estimates were higher, successive editions incorporate different data and modelling inputs; changes between editions cannot therefore be interpreted as a temporal trend.
A 2023 analysis summarised by the Philippine Institute for Development Studies estimated an annual cancer-related economic burden of PHP 35.3 billion, including PHP 29.2 billion in direct health-care costs and PHP 1.1 billion in wages lost because of illness and caregiving [4]. These estimates should be interpreted in the context of the underlying analysis and its assumptions. Limited screening uptake may contribute to delayed diagnosis, although causal attribution is not established by the available evidence. In a 2020 community-based survey of 338 women in rural Masbate, 13.9 percent reported ever having undergone cervical cancer screening [5]. The study cited earlier registry evidence suggesting that many cervical cancers were diagnosed at an advanced stage, but stage at diagnosis was not measured in the survey itself. Comparable evidence was not identified for other cancer sites; the cervical-cancer findings should therefore not be generalised to all cancers in the Philippines.
Recent policy changes may improve access to selected cancer-screening tests. The National Integrated Cancer Control Act (Republic Act No. 11215, 2019) provides for a Philippine Cancer Center and a population-based cancer registry and directs PhilHealth to expand benefit coverage for screening and early detection [6]. PhilHealth Circular No. 2025-0014 established outpatient benefit packages for mammography, breast ultrasound, low-dose chest computed tomography, alpha-fetoprotein testing, liver ultrasound, and screening colonoscopy [7,8]. Under the circular, eligibility assessment and test requests are linked to accredited primary-care providers, which receive results and arrange referral when indicated. In March 2025, the Health Technology Assessment Council issued positive recommendations concerning government financing of high-risk HPV DNA testing for women aged 30–65 years every five years and visual inspection with acetic acid where HPV DNA testing is not operational; the recorded Secretary of Health decisions were acknowledged on 25 June 2025 [9,10]. This review did not identify a separate Department of Health issuance establishing a complete national cervical-screening and clinical-management algorithm. The HTA documents are therefore described here as financing recommendations rather than as a comprehensive national algorithm.
Financing screening tests does not, by itself, establish an effective screening programme. Screening also requires sufficient personnel to interpret images and laboratory samples and to ensure the diagnostic follow-up of positive results. Oncology specialists are unevenly distributed, with 0.32 medical oncologists, 0.15 surgical oncologists, and 0.09 radiation oncologists per 100,000 population; five of the country’s 17 regions have no radiation oncologist [11,12]. These figures describe oncology treatment capacity rather than the workforce responsible for interpreting screening tests. This review identified no current national estimates of the numbers or geographic distribution of radiologists, breast imagers, pathologists, cytotechnologists, or endoscopists involved in screening, which constitutes an important evidence gap. The 2025 review of readiness for artificial intelligence in Philippine health care identified fragmented electronic health-record adoption, limited interoperability, and unreliable digital connectivity, particularly outside major urban centres [13]. Software-assisted image interpretation has been proposed as one means of extending limited reader capacity. In the Swedish MASAI trial, AI-supported mammography reduced screen-reading workload by 44%, increased cancer detection without increasing false-positive results, and subsequently produced a lower interval-cancer rate than standard double reading [14,15,16]. Whether these findings are transferable to the Philippines, where screening remains largely opportunistic and health-system capacity varies geographically, requires prospective local evaluation.
This review has three objectives: first, to identify applications of interpretive software in cancer screening in the Philippines documented between 2020 and 2026 and to assess the quality of the sources in which they were reported; second, to examine the evidentiary, infrastructural, and governance gaps that may impede their integration into routine practice; and third, to propose feasible responses within the capacity of the current health system.

2. Materials and Methods

2.1. Design and Reporting, Sources and Search Strings

This narrative review used a structured, documented search and applied SANRA principles to justify its scope, describe the search, grade the evidence, and link findings to conclusions [17]. As SANRA is a quality-assessment tool rather than a systematic-review reporting standard, systematic coverage is not claimed.
Table 1 reports the searched sources, principal terms, final search dates, and numbers of full-text records assessed. Web-based sources and the reference lists of relevant reviews and included studies were also hand-searched.

2.2. Eligibility

An item was included in the core evidence set if it reported a model, product, deployment, evaluation or policy proposal in which software interprets a screening test (mammography, breast ultrasound, chest radiography, low-dose CT, colonoscopy, cervical cytology, colposcopic or visual-inspection images, HPV testing, liver ultrasound or alpha-fetoprotein) for asymptomatic cancer detection, and the work was conducted in the Philippines, used Philippine patient data, or included Philippine sites or readers with results reported. Items were placed in an adjacent set, reported separately, if they concerned diagnostic rather than screening use, a non-cancer indication for the same class of software, or a Philippine model developed for a disease category that includes but is not limited to cancer. Philippine screening policy and workforce documents were included as context. International trials were used only as comparators for the strength of evidence.

2.3. Source Grading and Credibility Assessment

Core evidence was graded by source type: A, peer-reviewed prospective or randomised studies involving Filipino patients; B, peer-reviewed retrospective studies involving Filipino patients; C, peer-reviewed multinational studies that included Philippine sites or readers but reported pooled results; D, conference abstracts; E, government, health technology assessment, insurer, or programme documents; and F, news reports or vendor releases. For each item, the study design, sample, reference standard, principal threats to validity, and generalisability to Philippine screening populations were recorded (Table 3). Numerical performance claims from category F sources were not considered evidence of accuracy.

3. The Cancer Screening Landscape in the Philippines

Table 2 summarises modelled 2024 incidence and mortality for cancer sites relevant to screening. In these estimates, breast cancer accounted for 18.8 percent of new diagnoses, colorectal cancer ranked second by incidence, lung cancer accounted for the largest number of cancer deaths, and cervical cancer accounted for 7,829 new cases and 3,389 deaths [2].
Cancer screening in the Philippines remains predominantly opportunistic. A 2025 ASEAN scoping review grouped the Philippines with Myanmar, Laos, Cambodia, Vietnam, Brunei, Indonesia, and Timor-Leste as countries relying primarily on opportunistic screening, whereas Singapore, Malaysia, and Thailand place greater emphasis on organised programmes [18]. Available reports indicate limited cervical screening uptake. In the Cordillera Administrative Region, 10,238 women—3.01 percent of the target population of 340,614—were screened in 2025 [19]. A 2025 media report accompanying the approval of HPV DNA testing also stated that fewer than 1 percent of Filipino women underwent regular cervical screening, although the report did not identify the underlying national data source [20].
Three developments since 2024 provide potential entry points for interpretive software. First, PhilHealth Circular No. 2025-0014 established outpatient benefits of PHP 2,610 for mammography, PHP 1,350 for breast ultrasound, PHP 7,220 for low-dose chest computed tomography, PHP 1,230 for alpha-fetoprotein testing, PHP 960 for liver ultrasound, and PHP 23,640 for screening colonoscopy [7]. The circular requires results to be returned to the referring primary-care provider, recorded in the patient’s health record, and followed by referral when indicated. A subsequent PhilHealth standard limits annual low-dose CT screening to asymptomatic persons aged 50–80 years who meet specified smoking or family-history criteria. These requirements could support, but do not themselves establish, a screening register. Second, media reports on the Centralized Laboratory Model for HPV DNA Screening in Metro Manila described self-collection and centralised laboratory testing, with 4,969 women screened, a 7 percent positivity rate, and 190 receiving preventive treatment or referral [22,23]. Because no primary programme report was identified, these figures should be treated as programme-reported, and the 190 should not be interpreted as histologically confirmed precancer cases. Third, Republic Act No. 11215 mandates a Philippine Cancer Center and a population-based cancer registry [6]; if implemented with appropriate data linkage, these could support outcome evaluation of screening software.

4. Documented Applications, 2020–2026

Table 3 presents the core evidence on software used to interpret cancer-screening tests with a Philippine component. Table 4 summarises related diagnostic, non-cancer, or category-level work relevant to the evaluation and governance of comparable technologies. The core evidence was limited and heterogeneous; two peer-reviewed studies reported separately identifiable Philippine patient data, and both were retrospective.

4.1. Breast Cancer

Two studies were relevant to mammography. Tiangco and colleagues reported a pilot study conducted at The Medical City in 2020 [24]. A deep-learning model was developed using 900 images from 215 mammographic examinations obtained from women who subsequently underwent biopsy and was evaluated against surgical pathology. The report presented sensitivity of 95.7 percent, specificity of 87.0 percent, accuracy of 90.7 percent, and an AUROC of 0.76. Because the cohort was restricted to women who underwent biopsy, these estimates may not represent performance in an unselected screening population. The report did not clearly state whether images from each woman were confined to a single dataset partition; consequently, patient-level separation of training and test data could not be verified. The authors also proposed a potential role for the model in BI-RADS 0 examinations; however, binary model classification would not replace the additional imaging assessment ordinarily required for BI-RADS 0. This was the only Philippine-developed mammography model identified through the review search.
The second study was a 2026 multinational multireader evaluation by Ong and colleagues [25]. Nine breast radiologists from several Asian institutions, including radiologists affiliated with Philippine hospitals, interpreted 302 mammograms with and without commercial decision-support software. The case set included 89 biopsy-proven cancers. Mean reader AUC increased from 0.799 to 0.851, specificity increased from 77.0 percent to 88.4 percent, sensitivity was unchanged, and reading time decreased. Because the cases were pooled and were not identified as Philippine, the study evaluates the response of Philippine readers to software support rather than the software’s performance in Filipino women. It was the first indexed study identified in this review to include Philippine breast radiologists using such software under study conditions.
The two studies provide a substantially different level of evidence from the Swedish MASAI trial, which randomised 105,934 women to AI-supported screening or standard double reading. Reported findings included a 44 percent reduction in screen-reading workload [14], higher cancer detection without a higher false-positive rate [15], and non-inferiority for interval-cancer occurrence at follow-up [16]. The reported relative reduction in interval-cancer rate should be interpreted with its confidence interval and trial context. This review did not identify comparable prospective evidence in Filipino screening populations or a Philippine patient-validation study of a vendor product reported as active in the country.

4.2. Lung Cancer

Lung cancer had the most varied body of Philippine material identified in this review, although much of it was not peer reviewed. In 2024, AstraZeneca and AC Health announced the use of Qure.ai chest-radiograph software at Healthway Cancer Care Hospital in Taguig and Healthway QualiMed Hospital in Santa Rosa to assist pulmonary-nodule identification [26,27]. The initiative was described as part of a commitment to screen one million Filipinos by 2026. A separate collaboration reported extending software-assisted chest-radiograph assessment to the Aventus Medical and Intellicare networks [28]. Reference [26] stated that the software could provide a preliminary assessment in approximately 20 seconds; however, associated performance statements were vendor attributed and should not be interpreted as Philippine validation evidence. This review did not identify published Philippine data on screening yield, positive predictive value, or subsequent CT uptake. Global programme figures in reference [38] were likewise treated as vendor-reported implementation data rather than independent evidence of effectiveness.
Two additional reports provided limited implementation information. A 2025 conference abstract described lung-nodule triage using chest radiographs obtained through tuberculosis-screening programmes in Ethiopia, the Philippines, and Vietnam [29]. Because results were pooled and Philippine-specific denominators and pathology outcomes were not reported, the abstract does not provide country-specific diagnostic validation; it suggests only the feasibility of using tuberculosis-screening infrastructure for lung-nodule triage. Separately, a media report attributed to the Philippine Cancer Society stated that approximately 25,000 individuals had undergone software-assisted chest radiography, approximately 1,600 people with detected nodules had been referred for low-dose CT, and the software-assisted interpretation cost approximately PHP 50 per examination [30]. These figures were not supported by a peer-reviewed programme evaluation, and the report did not provide stratification by risk group, CT findings, pathology outcomes, or follow-up. The reported pathway nevertheless illustrates a possible triage approach before low-dose CT. PhilHealth reimburses low-dose CT under Circular No. 2025-0014 [7], while Circular No. 2025-0027 describes annual screening eligibility for selected asymptomatic persons aged 50–80 years [21].

4.3. Colorectal Cancer

The only indexed Philippine patient study of software-assisted colorectal cancer screening identified in this review concerned colonoscopy. Tan and colleagues retrospectively compared AI-assisted with cap-assisted colonoscopy at Chinese General Hospital and Medical Center in Manila among adults aged 40 years or older undergoing screening or surveillance colonoscopy between January and May 2022 [31]. Propensity-score matching produced 49 patient pairs. Adenoma detection was 51.0 percent with AI assistance and 46.9 percent with cap assistance, with no statistically significant difference between groups (p = 0.686); polyp detection also did not differ significantly. The accompanying editorial noted that transparent caps are inexpensive and accessible and discussed their potential relevance in resource-limited settings, given the absence of a significant advantage for AI in this small study [32]. The study was retrospective and single-centre, and the absence of a standard-colonoscopy arm means that it cannot determine whether either adjunct improves detection relative to usual care. It nevertheless provides initial evidence of local evaluation in a screening or surveillance population. PhilHealth now reimburses screening colonoscopy at PHP 23,640 [7], which may facilitate broader access to the procedure but does not specifically fund software evaluation.

4.4. Cervical Cancer

No Philippine study or pilot evaluation of software-assisted cervical screening was identified through the review search. A June 2025 press report stated that AI-enhanced cervical-screening technologies were presented at a summit in Manila, but it did not identify the technologies or document Philippine deployment or evaluation [33]. The HTA Council recommendation supports visual inspection with acetic acid as a screening method where HPV DNA testing is not operational, rather than specifically as triage after a positive HPV test [10]. Visual interpretation is operator dependent and requires training and quality assurance. In Zambia, an automated visual-evaluation system developed using smartphone images from 8,204 women was assessed as a primary screening and HPV-triage tool [39]. A subsequent prospective study in five African countries reported sensitivity of 60.1 percent and specificity of 81.9 percent for CIN2+, compared with 36.6 percent and 94.2 percent, respectively, for naked-eye visual inspection [40]. Differences in participant characteristics, disease prevalence, programme design, and reference-standard procedures limit direct transfer of these findings to Philippine screening populations; prospective local evaluation would therefore be required.

4.5. Liver and Other Cancers

For liver disease, a 2021 conference paper from the University of the Philippines Los Baños and The Medical City described a convolutional neural network trained on 1,895 ultrasound images from Filipino cohorts to classify several liver-image categories [35]. The model was not developed or evaluated specifically for hepatocellular-carcinoma screening and is therefore treated as adjacent evidence. PhilHealth reimburses alpha-fetoprotein testing and liver ultrasound for eligible individuals at increased risk of liver cancer [7], but this review did not identify a Philippine application of interpretive software to either test. A commentary suggested that recorded Philippine skin-cancer counts may underestimate the burden because national registration is incomplete [37]; no local study of software-assisted skin-cancer screening was identified. UTAK AI is a diagnostic brain-tumour imaging project led by UP Manila with Philippine and Taiwanese collaborators [36]. The programme report described high accuracy but did not provide numerical results or sufficient methodological detail for independent appraisal; the project was therefore classified as diagnostic and adjacent to the screening evidence.

4.6. What the Tuberculosis Programme Shows

The most detailed peer-reviewed Philippine evidence identified on programme-scale chest-radiograph software concerned tuberculosis. Marquez and colleagues retrospectively analysed 5,740 individuals who underwent chest radiography and molecular testing during case-finding activities in four Philippine regions between May 2021 and March 2024 [34]. At an abnormality threshold of 0.5, the study reported pseudo-sensitivity of 95.6 percent, pseudo-specificity of 28.1 percent, positive predictive value of 18.4 percent, and an AUROC of 0.82. The authors estimated that increasing the threshold to 0.68 would reduce molecular testing by 42 percent while missing 10 percent of detected cases. Because confirmatory testing was incomplete among software-negative individuals, the estimates were appropriately described as “pseudo” measures and may be affected by verification bias. The study illustrates the importance of local threshold calibration, subgroup assessment, and transparent reporting. These principles may inform lung-nodule triage evaluation, although tuberculosis detection and cancer screening are distinct tasks and require separate validation. The study also indicates that programme-scale evaluation of deployed imaging software is feasible in Philippine settings; this review did not identify a comparable peer-reviewed evaluation in a cancer-screening pathway.

4.7. The Regional Picture

Tun and colleagues searched for ASEAN studies published from January 2019 through May 2024 and included 14 studies across six cancer-screening areas [18]. The included studies were conducted in Singapore, Malaysia, Vietnam, Thailand, and Indonesia; none used Philippine data. The Philippine evidence summarised in Table 3 was not included, apparently because relevant items fell outside the review period or did not meet its eligibility criteria. The scoping review therefore indicates an absence of included Philippine evidence for the period and criteria searched, rather than proving that no Philippine activity occurred. The colonoscopy and multireader mammography studies published in 2026 provide more recent indexed evidence.

5. Gaps

The gaps fall into six groups, summarised with candidate responses in Table 5.

5.1. Evidence

This review did not identify a prospective diagnostic-accuracy or outcome study of cancer-screening software in Filipino patients. Two peer-reviewed studies with separately identifiable Philippine patient data were retrospective [24,31]. A multinational multireader study included Philippine-affiliated radiologists but did not report a Philippine patient case set [25], while a conference abstract on lung-nodule triage pooled Philippine data with data from Ethiopia and Vietnam [29]. Reported commercial deployments and the Philippine Cancer Society programme did not provide peer-reviewed Philippine diagnostic-performance or outcome data [26,27,28,30]. The geographic composition of the training data for individual commercial products could not be fully verified from the cited public sources; nevertheless, the national readiness review cautioned that under-representation in training data may reduce model performance in some populations [13]. Philippine-specific clinical factors, including previous pulmonary tuberculosis, may also affect transportability, although direct evidence for cancer-screening performance remains limited [30]. Prospective local evaluation is therefore warranted before effectiveness and value for reimbursement are inferred.

5.2. Programme Structure

Interpretive software is more readily evaluated and governed when embedded within an organised screening programme comprising a defined target population, systematic invitation or recruitment, result tracking, referral and recall, outcome linkage, and quality assurance. Cancer screening in the Philippines remains predominantly opportunistic, and no nationally organised programme incorporating all of these components was identified [18]. PhilHealth Circular No. 2025-0014 links covered screening tests to primary-care providers and requires results to be recorded and patients with positive findings to be referred for further evaluation [7]. However, the circular does not establish a population-based invitation system or national screening register, and no published evaluation of compliance with its referral and follow-up requirements was identified. The clinical value of software-assisted triage depends on timely access to confirmatory testing and continuity of care. The tuberculosis programme provides a Philippine example of software integrated into case finding with confirmatory molecular testing [34], but its applicability to cancer screening requires separate evaluation. No published assessment of a comparable end-to-end cancer-screening pathway was identified in this review.

5.3. Data and Registry

Section 28 of Republic Act No. 11215 requires the establishment of a national, population-based cancer registry and monitoring system [6]. However, a fully implemented registry with national population coverage was not identified. GLOBOCAN continues to estimate Philippine cancer incidence by applying rates from subnational registries to the national population [2], while available specialty and hospital-based registries do not provide comprehensive national coverage; Philippine dermatologists have similarly described recorded skin-cancer counts as probable underestimates [37]. Comprehensive linkage between screening records and cancer outcomes is needed to measure interval-cancer rates reliably at the population level and to assess false-negative findings, stage at diagnosis, and downstream outcomes. The national readiness review also identified fragmented health-information systems, inconsistent data standards, and limited interoperability [13]. Although hospital-based cancer datasets are available, no shared, nationally representative repository of locally labelled cancer-screening images was identified for model development or independent validation.

5.4. Workforce

Published workforce evidence indicates shortages and marked geographic concentration among oncology specialists, including radiation oncologists and training programmes [11,12]. These data describe treatment capacity rather than the workforce responsible for interpreting screening examinations. No current national estimates of the numbers or geographic distribution of radiologists, breast imagers, pathologists, cytotechnologists, or endoscopists involved in screening were identified in this review, nor were published studies of Filipino clinicians’ readiness for or training in interpretive software. The multinational mammography study was the first indexed study identified in this review to document Philippine-affiliated radiologists using such software under study conditions [25]. The national readiness review identified limited multidisciplinary expertise spanning clinical practice, data science, and implementation as an important challenge to responsible adoption [13]. The benefits of software-assisted triage will depend on sufficient capacity for confirmatory testing and diagnostic follow-up. The Philippine Cancer Society’s media-reported programme referred approximately 1,600 of 25,000 individuals for low-dose CT, but no data on CT completion, findings, pathology outcomes, or follow-up were reported [30].

5.5. Infrastructure and Equity

The Philippine deployments and patient studies identified in this review were concentrated mainly in urban tertiary facilities and managed-care networks, including sites in Taguig, Santa Rosa, Pasig, and Manila. Uneven digital connectivity and health-information infrastructure may constrain some deployment models, particularly those requiring continuous cloud access [13]. Concentration in relatively well-resourced settings could also limit generalisability to rural or underserved populations. The tuberculosis programme shows that software-assisted chest radiography can be deployed across several Philippine regions [34], while the three-country feasibility abstract suggests that tuberculosis-screening infrastructure may support lung-nodule triage [29]. However, the latter did not provide Philippine-specific implementation or outcome data. Whether these approaches improve equitable access to cancer screening in underserved Philippine populations remains uncertain.

5.6. Governance and Economics

Medical-device software in the Philippines is governed by the general medical-device authorisation framework under Administrative Order No. 2018-0002 [41]. In May 2025, the Food and Drug Administration issued draft guidance addressing the qualification, risk classification, and registration of medical-device software [42]; a separate 2026 draft addressed post-authorisation variations for medical and health-related devices [43]. This review did not identify final software-specific guidance explicitly addressing adaptive updates, model drift, or requirements for local revalidation. Philippine AI policy and proposed legislation were also evolving during the review period [44,45]. The national readiness review recommended risk-proportionate regulation, supervised testing environments, and stronger institutional capacity for health-care AI [13]. No Philippine economic evaluation of interpretive cancer-screening software was identified. A media report cited an approximate software-assisted radiograph interpretation cost of PHP 50, while PhilHealth reimburses low-dose CT at PHP 7,220 [7,30]; these figures concern different stages of care and cannot establish cost-effectiveness. Economic evaluation should include programme costs, confirmatory testing, false-positive and false-negative findings, cancers detected, downstream care, and health outcomes.

6. Proposed Responses

Table 5 summarises potential responses to the identified gaps. Four are discussed in greater detail because existing policy, financing, or programme structures may provide opportunities for their evaluation or implementation.

6.1. Use the Two Pathways That Already Have Financing and a National Test Recommendation

The lung pathway now includes reimbursement for low-dose CT, whereas the cervical pathway has a positive national financing recommendation for HPV DNA testing. These developments may provide practical settings in which to evaluate software-assisted triage. For lung cancer, a pragmatic study could assess radiograph-based triage in appropriately accredited facilities and report the flag rate, uptake of confirmatory CT, cancers detected per CT examination, and cost per cancer detected. Decision thresholds should be calibrated using local data, as illustrated by the Philippine tuberculosis programme, rather than adopted without local evaluation [34]. If sufficiently complete data become available, the Philippine Cancer Society programme could provide a potential starting cohort. For cervical cancer, automated visual evaluation could be assessed as a triage approach within established HPV-screening services, using histopathology as the reference standard and adapting relevant prospective study designs [39,40] to the Philippine context. Such studies could provide locally relevant performance and implementation evidence.

6.2. Borrow the Tuberculosis Programme's Evaluation Practice, and Correct Its Verification Gap

The tuberculosis study offers a potentially useful evaluation framework: prespecified thresholds, comparison against a reference standard, stratification by population and setting, explicit consideration of tests avoided relative to cases missed, and transparent reporting [34]. It also illustrates the risk of partial verification. Because confirmatory testing was not uniformly applied to software-negative individuals, the authors appropriately described sensitivity and specificity estimates as “pseudo” measures. Evaluations of cancer-screening software should therefore minimise differential verification. Where applying the reference standard to all participants is infeasible, investigators could apply it to a representative sample of software-negative examinations and use appropriate weighted analyses. A radiologist over-read may be clinically informative but would not necessarily constitute an equivalent reference standard for the target condition.

6.3. Make Reimbursement Conditional on Evidence and Reporting

Republic Act No. 11215 directs PhilHealth to expand benefits for cancer screening and early detection [6]. PhilHealth could consider evidence-development conditions for software used within reimbursed screening pathways, including locally relevant performance evaluation and standardised reporting of flag rates, confirmatory testing, and outcomes. Such requirements could help convert selected deployments into structured sources of evidence, although their feasibility, administrative burden, and cost would require assessment. The HTA Council’s HPV DNA recommendation refers to minimum performance criteria and international target product profiles [9]; a comparable evidence framework might inform future assessment of screening software after relevant regulatory guidance is finalised.

6.4. Finish the Registry and Build a Shared Image Library

Interval-cancer occurrence is an important outcome reported by the MASAI trial [16], but reliable population-level estimation in the Philippines would require sufficiently complete linkage between screening records and cancer outcomes. Continued implementation of the population-based cancer-registry mandate under Republic Act No. 11215 could therefore strengthen the evaluation of screening technologies [6]. In parallel, a governed and de-identified collection of locally acquired mammograms, chest radiographs, colonoscopy images or video, cervical images, and cytology slides linked to appropriate reference standards could support independent validation and, where justified, model adaptation. Any such resource would require robust governance, privacy protection, data-quality standards, and equitable institutional access consistent with applicable Philippine law [47].
The wider literature reinforces the importance of evaluating AI-assisted cancer screening within the broader health system. A recent review of mammography access in Latin America identified persistent socioeconomic, geographic, and structural barriers, suggesting that technological availability alone may be insufficient to ensure equitable screening participation [48]. Similarly, a global scoping review of 84 studies on AI applications in breast cancer care found that most studies originated from developed countries and that patient-centred clinical outcomes were frequently unreported [49]. A separate methodological review of nine AI-based cancer diagnosis studies identified potential biases and deficiencies in methodological reporting and data transparency [50]. Together, these findings support the need for Philippine evaluations to extend beyond algorithmic accuracy and consider equitable access, diagnostic follow-up, transparent reporting, and clinically meaningful outcomes.

7. Limitations of the Review

This narrative review should be interpreted in light of several limitations. Although the search was structured and documented (Table 1), it was not designed to establish systematic-review-level completeness. Relevant studies, conference abstracts, institutional reports, and other grey literature may therefore have been missed, particularly in a rapidly evolving and heterogeneous field. The identification of additional eligible publications during subsequent checking further illustrates this possibility.
The available evidence base also imposes important constraints. Information on real-world deployments often came from press releases, media reports, or institutional communications rather than peer-reviewed studies. These sources may be incomplete or selectively reported and were therefore used primarily to document reported activities, not to establish diagnostic accuracy, clinical effectiveness, or cost-effectiveness. Evidence grading by source type was used to make these distinctions explicit; however, the grading scheme has not been externally validated.
Finally, the evidence base continues to change. Studies, deployments, regulatory documents, and performance evaluations released after the final search dates are not represented. The findings should therefore be understood as a time-bounded synthesis of the sources identified through September 2026.

8. Conclusions

From January 2020 through September 2026, the identified Philippine evidence on interpretive software in cancer-screening pathways was limited and heterogeneous. It included a retrospective single-centre mammography-classifier study, a retrospective matched comparison of computer-aided and cap-assisted colonoscopy, a multinational mammography reader study that included Philippine-affiliated radiologists, a three-country feasibility abstract that included Philippine sites, reported commercial chest-radiograph deployments without published Philippine performance results, and a cancer-society programme described in a press report. This review did not identify a prospective diagnostic-accuracy or outcome study in Filipino screening participants, nor a Philippine evaluation of software-assisted cervical- or liver-cancer screening. The principal evidence and implementation gaps concern local validation, organised programme infrastructure, registry completeness, screening-workforce data, connectivity and interoperability, software-specific regulatory guidance, and local economic evaluation. Recent financing and health-technology-assessment decisions for lung and cervical screening may create opportunities for prospective evaluation. Experience from the Philippine tuberculosis programme also suggests that programme-scale evaluation of imaging software is feasible, while highlighting the importance of representative verification and local threshold calibration. Priority actions include prospective, transparently reported evaluations embedded in complete screening pathways and proportionate evidence requirements for publicly financed use.

Author Contributions

Conceptualization, B.J.J.S.; methodology, B.J.J.S.; investigation, B.J.J.S.; formal analysis, B.J.J.S.; resources, B.J.J.S.; writing—original draft preparation, B.J.J.S.; writing—review and editing, B.J.J.S.; visualization, B.J.J.S. The author has read and agreed to the published version of the manuscript.:

Funding

This research received no external funding.

Conflicts of Interest

The author declares no conflict of interest.

References

  1. Philippine Statistics Authority. Registered Deaths in the Philippines, 2024. Quezon City, Philippines, 2026. Available online: https://psa.gov.ph/statistics/vital-statistics/node/1684082390 (accessed on 1 September 2026).
  2. Ervik, M.; Laversanne, M.; Lam, F.; Colombet, M.; Ferlay, J.; Filho, A.M.; Bray, F. Global Cancer Observatory: Cancer Today (2024 estimates, version 1.0). Philippines fact sheet; International Agency for Research on Cancer: Lyon, France, 2026; Available online: https://gco.iarc.who.int/media/globocan/factsheets/populations/608-philippines-fact-sheet.pdf (accessed on 4 September 2026).
  3. Sung, H.; Filho, A.M.; Laversanne, M.; Ferlay, J.; Siegel, R.L.; Soerjomataram, I.; Jemal, A.; Bray, F. Global cancer statistics 2024: GLOBOCAN estimates of incidence and mortality worldwide for 34 cancers in 186 countries. CA Cancer J. Clin. 2026, 76(4), e70090. [Google Scholar] [CrossRef]
  4. Philippine Institute for Development Studies. Filipinos lose Php 35 billion annually due to cancer: Philippines the worst in ASEAN in cancer premature deaths (summary of Ulep, V.G., The Economic Burden of Cancer in the Philippines, presented at the Philippine Cancer Summit, 24 February 2023). Available online: https://www.pids.gov.ph/details/news/in-the-news/filipinos-lose-php-35-billion-annually-due-to-cancer-philippines-the-worst-in-asean-in-cancer-prema-2 (accessed on 29 August 2026).
  5. Imoto, A.; Honda, S.; Llamas-Clark, E.F. Human papillomavirus and cervical cancer knowledge, perceptions, and screening behavior: A cross-sectional community-based survey in rural Philippines. Asian Pac. J. Cancer Prev. 2020, 21(11), 3145–3151. [Google Scholar] [CrossRef]
  6. Republic of the Philippines. Republic Act No. 11215: An Act Institutionalizing a National Integrated Cancer Control Program and Appropriating Funds Therefor. 14 February 2019. Available online: https://lawphil.net/statutes/repacts/ra2019/ra_11215_2019.html (accessed on 8 September 2026).
  7. Philippine Health Insurance Corporation. PhilHealth Circular No. 2025-0014: PhilHealth Benefit Packages for Selected Outpatient Cancer Screening Tests. Pasig City, Philippines, 2025. Available online: https://www.philhealth.gov.ph/circulars/2025/PC2025-0014.pdf (accessed on 1 September 2026).
  8. Philippine News Agency. PhilHealth to cover outpatient cancer screening tests starting Aug 14. 8 August 2025. Available online: https://www.pna.gov.ph/articles/1256145 (accessed on 1 September 2026).
  9. Health Technology Assessment Council; Department of Science and Technology. High-Risk Human Papilloma Virus (HrHPV) Testing (DNA Test) for early detection of cervical cancer: HTA Council Final Recommendation (13 March 2025) and Secretary of Health decision (acknowledged 25 June 2025). Available online: https://hta.dost.gov.ph/published-assessment/high-risk-human-papilloma-virus-hrhpv-testing-dna-test-for-early-detection-of-cervical-cancer/ (accessed on 9 September 2026).
  10. Health Technology Assessment Council; Department of Science and Technology. Visualization in Acetic Acid (VIA) for Screening of Cervical Cancer: HTA Council Final Recommendation, 13 March 2025; Secretary of Health decision acknowledged 25 June 2025. HTA Council final VIA assessment. Available online. (accessed on 11 October 2026).
  11. Eala, M.A.B.; Maslog, E.A.S.; Dee, E.C.; Ting, F.I.L.; Toral, J.A.B.; Dofitas, R.B.; Co, H.C.S.; Cañal, J.P.A. Geographic distribution of cancer care providers in the Philippines. JCO Glob. Oncol. 2022, 8, e2200138. [Google Scholar] [CrossRef]
  12. Jacomina, L.E.; Agas, R.A.F.; Benedicto, M.T.J.U.; Vega, G.P.; Paulino, A.C.; Mejia, M.B.A. Radiation oncology training in the Philippines: Bridging gaps for improved cancer care in low- and middle-income countries. JCO Glob. Oncol. 2024, 10, e2300462. [Google Scholar] [CrossRef]
  13. Sarmiento, R.F.R.; Overgaard, S.M.; Gai, C.; Overgaard, J.D.; Ohde, J.W. Guiding responsible AI in healthcare in the Philippines. npj Digit. Med. 2025, 8, 338. [Google Scholar] [CrossRef]
  14. Lång, K.; Josefsson, V.; Larsson, A.M.; et al. Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (MASAI): A clinical safety analysis of a randomised, controlled, non-inferiority, single-blinded, screening accuracy study. Lancet Oncol. 2023, 24, 936–944. [Google Scholar] [CrossRef]
  15. Hernström, V.; Josefsson, V.; Sartor, H.; Schmidt, D.; Larsson, A.M.; Hofvind, S.; Andersson, I.; Rosso, A.; Hagberg, O.; Lång, K. Screening performance and characteristics of breast cancer detected in the Mammography Screening with Artificial Intelligence trial (MASAI): A randomised, controlled, parallel-group, non-inferiority, single-blinded, screening accuracy study. Lancet Digit. Health 2025, 7, e175–e183. [Google Scholar] [CrossRef]
  16. Gommers, J.; Hernström, V.; Josefsson, V.; Sartor, H.; Schmidt, D.; Hjelmgren, A.; Larsson, A.M.; Hofvind, S.; Andersson, I.; Rosso, A.; Hagberg, O.; Lång, K. Interval cancer, sensitivity, and specificity comparing AI-supported mammography screening with standard double reading without AI in the MASAI study: A randomised, controlled, non-inferiority, single-blinded, population-based, screening-accuracy trial. Lancet 2026, 407, 505–514. [Google Scholar] [CrossRef]
  17. Baethge, C.; Goldbeck-Wood, S.; Mertens, S. SANRA—a scale for the quality assessment of narrative review articles. Res. Integr. Peer Rev. 2019, 4, 5. [Google Scholar] [CrossRef]
  18. Tun, H.M.; Rahman, H.A.; Naing, L.; Malik, O.A. Artificial intelligence utilization in cancer screening program across ASEAN: A scoping review. BMC Cancer 2025, 25, 703. [Google Scholar] [CrossRef]
  19. Philippine Information Agency. DOH-CAR expands cervical cancer screening with additional test kits. May 2026. Available online: https://pia.gov.ph/news/doh-car-expands-cervical-cancer-screening-with-additional-test-kits/ (accessed on 1 September 2026).
  20. Standard, Manila. DOH approves HPV DNA testing for cervical cancer. September 2025. Available online: https://manilastandard.net/lifestyle/314638267/doh-approves-hpv-dna-testing-for-cervical-cancer.html (accessed on 8 September 2026).
  21. Philippine Health Insurance Corporation. PhilHealth Circular No. 2025-0027: Quality Standards for Lung Cancer Screening Using Low-Dose Computed Tomography. Pasig City, Philippines, 2025. Available online: https://www.philhealth.gov.ph/circulars/2025/PC2025-0027.pdf (accessed on 1 September 2026).
  22. Limos, A. Self-sampling tests pushed for HPV, cervical cancer prevention in the Philippines. The Manila Times. 20 August 2025. Available online: https://www.manilatimes.net/2025/08/20/news/national/self-sampling-tests-pushed-for-hpv-cervical-cancer-prevention-in-the-philippines/2170505 (accessed on 8 September 2026).
  23. Daily Tribune. Drive to eliminate cervical cancer gains push. 19 August 2025. Available online: https://tribune.net.ph/2025/08/19/drive-to-eliminate-cervical-cancer-gains-push (accessed on 8 September 2026).
  24. Tiangco, B.J.; Koa-Sales, R.; Saludares, M.I.; Domingo, M. A pilot study on artificial intelligence-powered medical image analysis for breast cancer detection. Med. City J. 2020, 3(1). Available online: https://www.themedicalcity.com/ctri/files/journals/Tiangco%20et%20al%202020-%20A%20Pilot%20Study%20on%20Artificial%20Intelligence-Powered%20Medical%20Image%20Analysis%20for%20Breast%20Cancer%20Detection.pdf (accessed on 9 September 2026).
  25. Ong, W.L.L.; Tay, W.M.I.; Buenaflor, M.T.; Elmaadawy, M.; Dharmarajan, J.P.; Lastrilla, N.J.A.; Kushvaha, S.; Alabado, R.L.; Lai, C.; Sharma, R.K.; Chotai, N. Bolstering the performance of breast radiologists with AI-CAD in mammography: A multireader study. Acad. Radiol. 2026, 33(8), 3287–3297. [Google Scholar] [CrossRef]
  26. Cabalza, D. AI-powered technology for early lung cancer detection now in PH. Philippine Daily Inquirer. 5 August 2024. Available online: https://newsinfo.inquirer.net/1969383/ai-powered-technology-for-early-lung-cancer-detection-now-in-ph (accessed on 28 August 2026).
  27. Manila Bulletin. Health center integrates AI into early detection of lung cancer in the country. 30 July 2024. Available online: https://mb.com.ph/2024/7/30/health-center-integrates-ai-into-early-detection-of-lung-cancer-in-the-country (accessed on 8 September 2026).
  28. Qure.ai. AstraZeneca, Qure.ai, Aventus and Intellicare join forces to advance lung nodule risk identification in the Philippines. Vendor news release, undated on page. Available online: https://www.qure.ai/us/news-press-coverages/astrazeneca-qure-ai-aventus-and-intellicare-join-forces-to-advance-lung-nodule-risk-identification-in-the-philippines (accessed on 8 September 2026).
  29. Balanag, V.M.; Elepano, I.V.; Codlin, A.J.; Dao, T.P.; Binegdie, A.B. P1.17.31 Feasibility of integrating AI-assisted lung cancer screening in high tuberculosis burden settings. J. Thorac. Oncol. 2025, 20((10) Suppl. 1, S225. [Google Scholar] [CrossRef]
  30. Sampang, D. PCS pushes for AI-assisted X-rays for early cancer diagnosis. Philippine Daily Inquirer. 13 August 2025. Available online: https://newsinfo.inquirer.net/2095308/pcs-pushes-for-ai-assisted-x-rays-for-early-cancer-diagnosis (accessed on 9 September 2026).
  31. Tan, J.R.L.; Enriquez, K.B.T.; Aballe, K.V.T.; Go, M.A.G.; Lim, M.L.O.; Co, J.T. Comparison of colon adenoma detection rate using cap-assisted and artificial intelligence-assisted colonoscopy at a tertiary hospital in the Philippines: A propensity score-matched analysis. Clin. Endosc. 2026, 59(1), 106–114. [Google Scholar] [CrossRef]
  32. Kim, T.W.; Na, S.-Y. Is cap still useful for colon adenoma detection rate improvement in the artificial intelligence era? Clin. Endosc. 2026, 59(1), 73–75. [Google Scholar] [CrossRef]
  33. Philippine Daily Inquirer (Global Nation). Philippines urged to adopt global standards towards cervical cancer elimination by 2030. June 2025. Available online: https://globalnation.inquirer.net/279082/philippines-urged-to-adopt-global-standards-towards-cervical-cancer-elimination-by-2030 (accessed on 1 September 2026).
  34. Marquez, N.; Carpio, E.J.; Santiago, M.R.; Calderon, J.; Orillaza-Chi, R.; Salanap, S.S.; Stevens, L. Performance of chest X-ray with computer-aided detection powered by deep learning-based artificial intelligence for tuberculosis presumptive identification during case finding in the Philippines. BMC Glob. Public Health 2025, 3, 74. [Google Scholar] [CrossRef]
  35. Villanueva, D.A.; Baltazar, L.R.; Roxas-Villanueva, R.M.; Tiangco, B.; Viray, E.D.; Albia, J. Artificial intelligence for liver disease detection using ultrasound images. Proc. Samahang Pisika Ng Pilipinas 2021, 39, SPP-2021-2G-06. Available online: https://proceedings.spp-online.org/article/view/SPP-2021-2G-06 (accessed on 8 September 2026).
  36. Department of Science and Technology–Philippine Council for Health Research and Development. DOST-PCHRD presents AI-driven innovations at AmCham Philippines. DOST-PCHRD presents AI-driven innovations at AmCham Philippines. Available online. 18 September 2024. (accessed on 11 October 2026).
  37. Tan, N.M.G.; Arevalo, M.V.P.N.; Eala, M.A.B.; Siripunvarapon, A.H. Skin cancer in the Philippines: The Filipino narrative. JAAD Int. 2022, 8, 163–164. [Google Scholar] [CrossRef]
  38. eHealth Network. AstraZeneca and Qure.ai achieve 5 million AI-powered lung cancer screenings, flag 50,000 high-risk cases across 20 countries (report of a vendor release). 3 May 2025. Available online: https://ehealth.eletsonline.com/2025/05/astrazeneca-and-qure-ai-achieve-5-million-ai-powered-lung-cancer-screenings-flag-50000-high-risk-cases-across-20-countries/ (accessed on 8 September 2026).
  39. Hu, L.; Mwanahamuntu, M.H.; Sahasrabuddhe, V.V.; Barrett, C.; Horning, M.P.; Shah, I.; Laverriere, Z.; Banik, D.; Ji, Y.; Shibemba, A.L.; et al. Internal validation of Automated Visual Evaluation (AVE) on smartphone images for cervical cancer screening in a prospective study in Zambia. Cancer Med. 2024, 13(11), e7355. [Google Scholar] [CrossRef]
  40. African AVE Collaborative; Barrett, C.; Joseph, J.T.; Horning, M.P.; et al. Performance of an artificial intelligence-based tool for cervical precancer screening in five countries in Africa: A prospective, observational, diagnostic accuracy study. Lancet Glob. Health 2026, 14(1), e49–e60. [Google Scholar] [CrossRef]
  41. Food and Drug Administration; Republic of the Philippines. Administrative Order No. 2018-0002: Guidelines Governing the Issuance of an Authorization for a Medical Device based on the ASEAN Harmonized Technical Requirements. Department of Health: Manila, Philippines, 2018. [Google Scholar]
  42. Food and Drug Administration; Republic of the Philippines. Draft for Comments: Guidelines on the Regulation of Medical Device Software (MDSW) by the Food and Drug Administration. May 2025. Available online: https://www.fda.gov.ph/draft-for-comments-guidelines-on-the-regulation-of-medical-device-software-mdsw-by-the-food-and-drug-administration-fda/ (accessed on 9 September 2026).
  43. Food and Drug Administration; Republic of the Philippines. Draft for Comments: Guidelines on the Application for Variation of the Issued Product Authorizations for Medical and Health-Related Devices. 2026. Available online: https://www.fda.gov.ph/draft-for-comments-guidelines-on-the-application-for-variation-of-the-issued-product-authorizations-for-medical-and-health-related-devices/ (accessed on 9 September 2026).
  44. Regulations. ai. Philippines—AI Governance Framework (secondary summary of Philippine AI policy). Available online: https://regulations.ai/regulations/philippines-summary (accessed on 1 September 2026).
  45. House of Representatives; Republic of the Philippines. House Bill No. 6028, 20th Congress (proposed legislation on the regulation and risk classification of artificial intelligence systems). Available online: https://docs.congress.hrep.online/legisdocs/basic_20/HB06028.pdf (accessed on 8 September 2026).
  46. World Health Organization. Ethics and Governance of Artificial Intelligence for Health: WHO Guidance; World Health Organization: Geneva, Switzerland, 2021; Available online: https://iris.who.int/bitstream/handle/10665/341996/9789240029200-eng.pdf (accessed on 8 September 2026).
  47. Republic of the Philippines. Republic Act No. 10173: Data Privacy Act of 2012. 15 August 2012. Available online: https://www.privacy.gov.ph/data-privacy-act/ (accessed on 9 September 2026).
  48. Méndez-Domínguez, N.; Arce Medina, M.J.; Balam Gomez, M.; Morales Rojas, M.E.; Novelo Moreno, E. Regional Inequities in Mammography Access and Utilization in Latin America: Ethnic, Rural, and Structural Barriers Identified Through a Narrative Review. Epidemiologia 2026, 7(1), 25. [Google Scholar] [CrossRef]
  49. Chia, J.L.L.; He, G.S.; Ngiam, K.Y.; Hartman, M.; Ng, Q.X.; Goh, S.S.N. Harnessing Artificial Intelligence to Enhance Global Breast Cancer Care: A Scoping Review of Applications, Outcomes, and Challenges. Cancers 2025, 17(2), 197. [Google Scholar] [CrossRef]
  50. Smiley, A.; Reategui-Rivera, C.M.; Villarreal-Zegarra, D.; Escobar-Agreda, S.; Finkelstein, J. Exploring Artificial Intelligence Biases in Predictive Models for Cancer Diagnosis. Cancers 2025, 17(3), 407. [Google Scholar] [CrossRef]
Table 1. Search sources, strings, dates and yield.
Table 1. Search sources, strings, dates and yield.
Source Search string or terms Last searched Retrieved in full
PubMed / PubMed Central (Philippines OR Filipino) AND (cancer OR neoplasm) AND (screening OR "early detection") AND ("artificial intelligence" OR "deep learning" OR "machine learning" OR "computer-aided" OR "convolutional neural network"); site-specific variants for mammography, "chest X-ray", "low-dose CT", colonoscopy, colposcopy, cytology, HPV, ultrasound 1 Sept 2026 14
Google Scholar Same terms; additionally "Philippine" AND ("adenoma detection" OR "lung nodule" OR "cervical" OR "mammogram") AND ("AI" OR "deep learning") 2020–2026 9 Sept 2026 9
Publisher platforms (Springer Nature, Elsevier/ScienceDirect, Wiley, ASCO, Lancet, Clinical Endoscopy) Title/abstract search for Philippines AND cancer screening AND AI terms; citing-article lists of included studies 9 Sept 2026 8
Local journals and proceedings The Medical City Journal; Philippine Journal of Science; Samahang Pisika ng Pilipinas proceedings; Philippine College of Radiology and Philippine Society of Medical Oncology publication pages 1 Sept 2026 3
Government, HTA and insurer hta.dost.gov.ph (HPV, VIA); philhealth.gov.ph circulars 2025; doh.gov.ph issuances; fda.gov.ph draft circulars (medical device software); psa.gov.ph causes of death; congress.hrep.online (artificial intelligence bills) 30 Aug 2026 9
News archives and vendor sites Philippine Daily Inquirer, Manila Bulletin, Manila Times, Manila Standard, BusinessWorld, BusinessMirror, Daily Tribune, PNA, PIA, Sunstar; qure.ai, lunit.io: "AI" AND (cancer OR lung OR breast OR cervical) AND Philippines, 2020–2026 5 Sept 2026 13
Comparator trials (international) MASAI (Lancet Oncology 2023; Lancet Digital Health 2025; Lancet 2026); automated visual evaluation (Cancer Medicine 2024; Lancet Global Health 2026) 8 Sept 2026 5
Full-text retrieval counts include documents read for context and later excluded. HTA, health technology assessment; PNA, Philippine News Agency; PIA, Philippine Information Agency.
Table 2. Estimated incidence and mortality for the leading screening-relevant cancers in the Philippines, 2024 [2].
Table 2. Estimated incidence and mortality for the leading screening-relevant cancers in the Philippines, 2024 [2].
Cancer site New cases Share of all cases (%) Deaths Share of all deaths (%)
Breast 28,124 18.8 11,160 12.9
Colorectum 17,462 11.7 9,302 10.8
Lung 15,312 10.2 11,715 13.6
Prostate 9,195 6.1 4,170 4.8
Liver 8,638 5.8 7,969 9.2
Cervix uteri 7,829 5.2 3,389 3.9
Lip and oral cavity 1,633 1.1 1,004 1.2
All cancers 149,852 100 86,338 100
Source: Global Cancer Observatory, Cancer Today, 2024 estimates (version 1.0). Incidence is modelled from subnational registry rates applied to the 2024 population; mortality is projected from national vital registration. These are estimates, not counts, and differ from the 77,504 registered neoplasm deaths reported by the Philippine Statistics Authority for 2024 [1].
Table 3. Core evidence on interpretive software used or evaluated in cancer-screening pathways with a Philippine component, January 2020–September 2026.
Table 3. Core evidence on interpretive software used or evaluated in cancer-screening pathways with a Philippine component, January 2020–September 2026.
Cancer site Application and setting Design and result Grade Main threats to validity Ref.
Breast Deep-learning classifier for benign versus malignant lesions on screening mammograms, The Medical City, Pasig Retrospective; 900 images from 215 screening mammograms of women who subsequently underwent biopsy; the abstract reports sensitivity of 95.7%, specificity of 87.0%, accuracy of 90.7%, and an AUROC of 0.76 against surgical pathology; radiologist readings in the same series, with sensitivity of 86% and specificity of 46%, were computed on 137 women after excluding BI-RADS 0 and 6 B Cancer-enriched biopsied cohort; patient-level separation of training and test images not stated, so image-level leakage cannot be excluded; radiologist and model comparisons used different denominators; institutional journal [24]
Breast Multireader study of commercial mammography support software with nine Asian breast radiologists, including radiologists affiliated with Philippine hospitals Retrospective; 302 mammograms with 89 biopsy-proven cancers; mean reader AUC rose from 0.799 to 0.851 (p = 0.0151), specificity from 77.0% to 88.4%, sensitivity unchanged; reading time reduced C Case set not reported as Philippine; cancer-enriched; reader rather than patient outcome; not local patient validation [25]
Lung Chest radiograph software (qXR, Qure.ai) for nodule triage to low-dose CT, Healthway Cancer Care Hospital and Healthway Qualimed Hospital, AstraZeneca–AC Health partnership; separate deployment across Aventus Medical and Intellicare networks Deployments announced July–August 2024; stated aim of one million Filipinos screened by 2026; no Philippine yield, positive predictive value or CT uptake published F Vendor and partner statements; performance figures quoted are global vendor claims [26,27,28]
Lung Feasibility of lung nodule detection on chest radiographs taken for tuberculosis screening, using the same software, in Ethiopia, the Philippines and Vietnam Conference abstract reporting pooled feasibility results from Ethiopia, the Philippines, and Vietnam; country-specific Philippine denominators and pathology outcomes were not reported D Results pooled across countries; no Philippine denominator; no pathology outcome; abstract only [29]
Lung Philippine Cancer Society programme and proposal to DOH and PhilHealth to cover software-assisted radiograph reading as a triage step before low-dose CT Press statement, August 2025: about 25,000 individuals screened with software-assisted chest radiography, 1,600 with nodules forwarded to low-dose CT; software cost quoted at about PHP 50 per read F Unpublished programme figures reported verbally; no denominator by risk group, no CT results, no pathology [30]
Colorectum Computer-aided polyp detection versus cap-assisted colonoscopy, tertiary hospital, Manila (Chinese General Hospital and Medical Center) Retrospective propensity-matched cohort, January–May 2022; 49 matched pairs from screening and surveillance colonoscopies in adults aged 40 and over; adenoma detection rate 51% with software versus 47% with cap (p = 0.69); polyp detection rate also not different B Small matched sample; retrospective; no standard-colonoscopy arm, so the study compares two adjuncts rather than software against usual care; single centre [31]
Grades: A, peer-reviewed prospective or randomised study in Filipino patients; B, peer-reviewed retrospective study in Filipino patients; C, peer-reviewed multinational study including Philippine sites or readers with pooled results; D, conference abstract; E, government or programme document; F, news or vendor release. AUROC/AUC, area under the receiver operating characteristic curve; BI-RADS, Breast Imaging Reporting and Data System; DOH, Department of Health.
Table 4. Adjacent work: diagnostic, non-cancer or category-level applications relevant to evaluation and governance.
Table 4. Adjacent work: diagnostic, non-cancer or category-level applications relevant to evaluation and governance.
Area Work and setting Grade Relevance Ref.
Tuberculosis case finding (same software class) qXR version 3 in active and intensified case finding in four regions, May 2021–March 2024, USAID TBIHSS project; 5,740 individuals with radiograph and molecular test; pseudo-sensitivity 95.6%, pseudo-specificity 28.1%, AUROC 0.82; raising the threshold to 0.68 saved 42% of confirmatory tests at 10% missed cases Peer-reviewed retrospective programme evaluation Provides a Philippine programme-scale example of threshold calibration and the effects of partial verification [34]
Liver disease (category, not cancer screening) Convolutional neural network trained on 1,895 liver ultrasound images from Filipino cohorts to classify liver-disease categories; University of the Philippines Los Baños and The Medical City, 2021 proceedings D Shows local capacity for ultrasound model development; not a hepatocellular carcinoma screening model [35]
Brain tumour (diagnosis) UTAK AI: AI-assisted brain-tumour detection and analysis led by UP Manila in collaboration with Philippine General Hospital, Taipei Veterans General Hospital, and Taiwan AI Labs; supported through the MECO–TECO HAT initiative F (programme report) Publicly supported cancer-imaging project relevant to cross-institutional data collaboration; diagnostic rather than screening work [36]
Skin cancer (burden) Commentary: 2,102 basal cell and 614 squamous cell carcinomas recorded by the Philippine Dermatological Society 2011–2021, described as underestimates for want of a national registry Peer-reviewed commentary No Philippine dermatology screening software identified; illustrates registry gap [37]
Cervical cancer policy and technology context AI-enhanced cervical-cancer screening technologies presented at the June 2025 Cervical Cancer Elimination Summit in Manila F Technologies were not specified, and no Philippine deployment or evaluation was documented [33]
DOST-PCHRD, Department of Science and Technology–Philippine Council for Health Research and Development; TBIHSS, TB Innovations and Health Systems Strengthening.
Table 5. Gaps identified and possible responses.
Table 5. Gaps identified and possible responses.
Domain Gap Possible response
Evidence No prospective Philippine validation; two retrospective patient studies; deployment and programme reports lack published diagnostic-performance and outcome data Commission multicentre prospective evaluations in Philippine screening populations; require prespecified reporting of performance, follow-up, and patient outcomes; consider evidence and reporting requirements for reimbursement or public-sector endorsement
Programme Predominantly opportunistic screening; Circular No. 2025-0014 requires result recording and referral but does not establish systematic invitation, a national screening register, or evaluated follow-up Develop a minimal screening register linked to claims and primary-care records; audit referral and follow-up completion; pilot end-to-end evaluation in financed lung and cervical screening pathways
Data National population-based cancer registry not fully implemented; population-level interval-cancer rates cannot be measured reliably; no shared, nationally representative repository of locally labelled screening images identified Implement the registry mandate under Republic Act No. 11215 with a defined timetable; link screening records to cancer outcomes; develop governed, de-identified image repositories consistent with data-protection and interoperability requirements
Workforce No current national estimates of the screening-reader workforce or published evidence on clinician readiness; downstream diagnostic capacity may be insufficient for increased referrals Map the numbers and geographic distribution of screening-related specialists; align software deployment with diagnostic capacity; assess training needs and incorporate supervised evaluation into relevant professional training programmes
Infrastructure and equity Cancer-screening deployments concentrated in urban facilities and managed-care networks; uneven connectivity and health-information infrastructure may constrain implementation in underserved areas Evaluate connectivity-appropriate deployment models in regional and underserved settings; test integration within existing care pathways; monitor access, referrals, and outcomes by geography and socioeconomic group
Governance and economics Medical-device-software guidance remains in draft; no final guidance specific to adaptive model updates, drift, or local revalidation was identified; no Philippine economic evaluation of interpretive cancer-screening software was identified Finalise software-specific guidance with proportionate requirements for post-market monitoring, updates, and local revalidation; commission economic evaluations of priority pathways using programme and PhilHealth costs; strengthen regulatory and health-technology-assessment expertise
DOH, Department of Health; DOST-PCHRD, Department of Science and Technology–Philippine Council for Health Research and Development; FDA, Food and Drug Administration; HTAC, Health Technology Assessment Council; MDSW, medical device software; RA, Republic Act; YAKAP, Yaman ng Kalusugan Program.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.