Abstract: Background: Cang-ai volatile oil (CAVO) is a traditional Chinese medicine with properties that soothe the liver and alleviate depression. CAVO is widely utilized in the field of antidepressant research and has surfaced as a possible treatment for depression. Depression is a common affective disorder and effective treatment methods are still limited. CAVO is effective in treating depression; however, the exact mechanism is still unclear. This study aimed to explore the likely mechanism by which CAVO reduces symptoms of depression in rats exposed to chronic unpredictable mild stress (CUMS). Methods: We established a CUMS model in Sprague–Dawley rats and administered CAVO via nebulization to evaluate its therapeutic effect. Behavioral and histology tests were conducted to evaluate brain tissue damage. We utilized metabolomics combined with proteomics to analyze the effects of CAVO. We then assessed molecular validation to further clarify the molecular mechanism of its activity. Results: In CUMS model rats, inhaling aerosolized CAVO significantly reduced brain pathology and depression-like behaviors. CAVO significantly changed serum levels of inflammatory cytokines and neurotrophic factors. Biomarkers linked to CAVO's antidepressant effects were found via metabolomics. Functional analyses highlighted key molecular players such as TrkB, and CREB, and a close association with the antidepressant action of CAVO was confirmed. Conclusion: This study reveals that CAVO reduces depression-like behaviors in CUMS rats by regulating the NT/Trk signaling pathway. These results demonstrate CAVO's therapeutic potential and lay the groundwork for future studies and the creation of depressive treatments.

Abstract: Vitamin B12 is an essential water-soluble vitamin required for critical biological processes such as DNA synthesis, erythrocyte maturation, and maintenance of nervous system integrity. Deficiency of vitamin B12 can lead to serious clinical outcomes, including megaloblastic anemia, and potentially irreversible neurological damage. Conversely, hypercobalaminemia may be associated with severe disorders, including solid neoplasms, hematological malignancies and, in some cases, may result from inappropriate supplementation or immunoglobulin–B12 macro-complexes. Although current guidelines recommend total serum vitamin B12 and holotranscobalamin (holoTC) as first-line biomarkers, total serum vitamin B12 remains the most widely used test in routine clinical practice. However, since holoTC represents the biologically active fraction of vitamin B12 available for receptor-mediated cellular uptake, it appears to provide a more reliable assessment of cobalamin status, particularly in specific clinical contexts. Compared with total vitamin B12 measurement, holoTC is assessed using a more limited number of analytical methods and the majority of available kit are aligned with the WHO reference standard, thereby improving inter-assay harmonization. This review explores literature data about the role of vitamin B12 and holoTC, discussing analytical challenges and clinical interpretation, highlights the potential advantages of holoTC over total serum B12.

Abstract: Background: At present, the gold standard for gastric cancer (GC) confirmation relies mostly on histopathology, an invasive procedure. Noninvasive detection methods using serum for large-scale screening maybe useful for the early diagnosis of GC. Helicobacter pylori (HP) infection and chronic atrophic gastritis are major GC risk factors. We recently developed a noninvasive test called DSC test—, based on patient’s age, sex, their serum PGI and PGII, anti-HP immunoglobulin (IgG), and gastrin G17 levels –predicting GC risk as low (score 0, S0) or high (score 2, S2). The comparative investigation at serum protein level of the two different patient groups detected by our DCS test (S0 and S2) may undoubtedly help to identify gastric disease-dependent proteins, resulting from bacterial infection or gastric mucosa inflammation, as well as get better insight the molecular scenario associated to pre-cancerous conditions. Methods: Mass spectrometry-based protein analysis of tryptically digested proteins was performed, followed by univariate statistical analysis for the different DSC groups from two cohort of patients (exploratory and validation). Significantly differentially abundant proteins differing more than 1.5-fold between groups were selected and validated, and their putative role(s) in gastritis and GC discussed. Results: We present data of comparative protein analysis of sera from patients at high risk to develop gastric cancer or advanced atrophy, depending on their DSC score. In particular, we used untargeted liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based proteomics as semiquantitative method to profile proteins specifically associated with score 2 in sera of 80 patients. In both the exploratory and the validation cohorts, four proteins (beta-2-microglobulin, EGF-containing fibulin-like extracellular matrix protein 1, complement factor D, and Cystatin-C) were more abundant, while two (sex hormone-binding globulin and pregnancy zone protein) were less abundant in sera of S2 individuals (|fold change|≥0.6, p < 0.05, t-test). The higher presence of beta-2-microglobulin (B2M) and the lower content of pregnancy zone protein (PZP) in S2 sera were validated by immunoblotting. Conclusions: This study identified a proteomic signature differentially associated to sera of patients with a different risk to develop GC/advanced atrophy according to our DSC test. The protein marker panels presented in this work will contribute to improve GC diagnostics, once they have been transferred from a research result to a practical tool.

Abstract: Adolescents growing up in multilingual regions experience diverse educational contexts that may shape their daily routines and psychosocial environments, but their independent relevance for mental health remains unclear. South Tyrol, with its parallel German-, Italian-, and Ladin-language school systems, provides a unique setting to examine these associations. This study assessed whether school language and home–school language mismatch are associated with mental health, psychosomatic symptoms, and health-related behaviors among adolescents. We analyzed data from a population-based survey of 2005 adolescents aged 11–19 years who provided self-reported information on mental health, psychosomatic complaints, school stress, social support, digital behaviors, lifestyle, and sleep. Group comparisons by school language were conducted using general linear models and χ² tests with effect sizes. Multivariable regression analyses examined the independent association of home–school language mismatch with mental health outcomes, adjusting for sociodemographic and educational factors and further incorporating sleep-related behaviors. Mental health outcomes, psychosomatic symptoms, and most health-related behaviors showed little variation by school language, with generally small effect sizes. Home–school language mismatch was associated with slightly higher depressive symptom scores in unadjusted analyses but was not independently associated with mental health outcomes after adjustment. In contrast, weekly sleep problems emerged as the strongest correlate of depressive symptoms, accounting for a substantial proportion of explained variance. These findings indicate that adolescent mental health in this multilingual context is shaped less by the language of schooling itself than by broader behavioral and developmental factors, highlighting sleep-related behaviors as a central and modifiable target for prevention.

Abstract: Physics-informed neural networks (PINNs) have emerged as powerful tools for solving partial differential equations, but their training remains challenging due to ill-conditioned loss landscapes. While adaptive methods like Adam dominate deep learning, they exhibit instability on stiff PDEs, and second-order methods are com- putationally prohibitive. We present EPANG-Gen (Enhanced Physics-Aware Natural Gradient with Generalization), a novel optimizer that combines memory-efficient eigen- decomposition with lightweight Bayesian uncertainty quantification. EPANG-Gen in- troduces three key innovations: (1) a randomized eigenspace estimator that approx- imates Hessian curvature with O(dk) memory (k ≪ d), (2) Bayesian R-LayerNorm for per-activation uncertainty estimation, and (3) adaptive rank selection (PASA) that dynamically adjusts to problem difficulty. We evaluate EPANG-Gen on four bench- mark PDEs—Poisson 1D, Burgers’ equation, Darcy flow, and Helmholtz 2D—and on the challenging Taylor-Green vortex at Re = 100, 000, a canonical 3D turbulence problem. Results show that EPANG-Gen matches Adam’s performance on the toughest turbulent regime while eliminating the 25% catastrophic failure rate of ADOPT across 72 runs. Ablation studies confirm that eigen-preconditioning improves performance by 11–35%. The built-in uncertainty estimates provide actionable confidence metrics at negligible cost. EPANG-Gen represents the first optimizer specifically designed for geo- metric and physical AI that combines theoretical convergence guarantees with practical robustness for safety-critical applications.

Abstract: Non-dilute emulsions are emulsions where the concentration of the droplets is high enough for the neighboring droplets to interact with each other hydrodynamically but is still smaller than the packed bed concentration where the droplets are packed and deformed against each other. Thus, they cover a broad range of droplet concentration. Many emulsions encountered in industrial applications fall under this category. Non-dilute emulsions exhibit rich rheological behavior from a simple Newtonian fluid to a highly non-Newtonian fluid reflecting shear-thinning, shear-thickening, yield stress, viscoelasticity, etc. In this article, the rheology of non-dilute emulsions is re-viewed comprehensively. Emulsions of hard-sphere type droplets and deformable droplets, with and without surfactants, are covered. The mathematical models de-scribing the rheological behavior of non-dilute emulsions are discussed. The influences of electric charge and interfacial rheology on the rheological behavior of emulsions are covered in detail. The flocculation of droplets caused by different mechanisms such as depletion and bridging induced by additives and their effect on emulsion rheology are investigated thoroughly. Finally, the dynamic rheology of non-dilute emulsions is dis-cussed covering both pure oil-water interfaces and additive-laden interfaces. The mathematical models describing the dynamic rheological behavior of non-dilute emulsions are described. Based on the existing theoretical and empirical models, it is possible to a priori predict the rheology of non-dilute emulsions. However, serious gaps in the existing knowledge on non-dilute emulsion rheology remain. This review identifies the gaps in existing knowledge and points out future directions in research related to non-dilute emulsion rheology.

Abstract: Background: Mechanical thrombectomy is established for acute ischaemic stroke due to large vessel occlusion, particularly in patients with moderate to severe neurological deficits. However, the benefit–risk profile in patients presenting with mild neurological deficits despite confirmed large vessel occlusion remains uncertain.Objective: To systematically synthesise evidence on the effectiveness and safety of mechanical thrombectomy compared with best medical therapy in adults with mild acute ischaemic stroke and large vessel occlusion. Methods: Randomised controlled trials and comparative observational studies will be included. Primary outcomes will include functional outcomes at 90 days, with secondary outcomes addressing safety, reperfusion, and early neurological deterioration. Meta-analysis will be performed using random-effects models where appropriate. Risk of bias and certainty of evidence will be assessed using Cochrane and GRADE methodologies. Registration: The systematic review protocol was prospectively submitted to PROSPERO with registration granted on the 6th of March, 2026 [CRD420261323013].

Abstract: The aerodynamic performance of an aircraft wing is influenced by the angle of attack (AoA), which directly affects lift, drag, and overall efficiency. This study presents a theoretical analysis of the effect of AoA on a symmetric thin aerofoil using aerodynamic models including Thin Aerofoil Theory and Lifting-Line Theory are discussed. Results are discussed under three regimes; linear AoA range where lift increases proportionally while maintaining steady improvement in lift to drag ratio, pre-stall regime where partial flow separation takes place by reducing lift growth, near-stall where flow separation causes deterioration in aerodynamic efficiency. The study highlights the importance of wing geometry including aspect ratio, camber, and sweep angle. Understanding complex interactions between AoA, lift generation, drag forces, and wing geometry is crucial for optimizing aircraft design and improving aerodynamic performance.

Abstract:

This paper proposes a gravitational theoretical framework based on the dynamics of discrete spacetime units. The core idea is that there exists a conserved ”spacetime raw material”, and quan- tum virtual processes of matter continuously produce new spacetime units by consuming this raw material, forming local density gradients—which manifest as spacetime curvature. This mechanism naturally eliminates action-at-a-distance, is compatible with general relativity under covariance con- straints, and provides a unified explanation for dark matter, dark energy, black hole singularities, and other long-standing puzzles.First, we clarify the meta-principle of ”global common covariance”, and on this basis, give the ultimate explanation of symmetry breaking: symmetry is not ”broken”, but a local cost paid for global covariance. Then we systematically elaborate twelve core arguments of the framework, and starting from the only fundamental equation (the second-order discrete wave equation of complex fields), we rigorously and step-by-step derive the Newtonian gravity limit, mass-energy equation E = mc2 , the principle of constant speed of light, Maxwell’s equations, Newton’s three laws, Schr¨odinger equation, Dirac equation, the origin of spin-1/2, and the geometric formula of the fine-structure constant. All physical laws are derived results rather than external inputs. Finally, we present quantitative predictions that can be tested by future experiments.

Abstract: Along with some known and less known results, we discuss new insights relating combinatorics of words and the ordering of the rationals from a dynamical systems point of view, somehow continuing along the path started in [BI]. We obtain in particular a set of results that structure and enrich the correspondence between the Stern-Brocot (SB) ordering of rational numbers and the corresponding ordering of Farey-Christoffel (FC) words, a class of words that, since their appearance in literature at the end of the 18th century, have revealed numerous relationships with other fields of mathematics. Among the results obtained here is the construction of substitution rules that act on the FC words in a parallel way to the maps on the positive reals that generate the permuted SB tree both vertically and horizontally. A complete correspondence is obtained between the vertical and horizontal motions on the SB tree and the geodesic motions along scattering geodesics and the horocyclic motion along Ford circles in the upper half-plane, respectively.

Abstract: Urinary Tract Infections (UTIs) represent one of the most prevalent bacterial infections globally, posing significant health burdens, especially in low- and middle-income countries (LMICs), due to delayed diagnoses, limited access to laboratory services, and rising antimicrobial resistance. This study presents a machine learning (ML)-based diagnostic support framework for early UTI detection, leveraging structured clinical data and explainable artificial intelligence (XAI) techniques to enhance interpretability and trust among healthcare providers. A patient dataset containing 4,865 records was used in the study to train and test Extreme Gradient Boosting (XGBoost), Decision Tree (DT), and Random Forest (RF) classifiers, while class imbalance was addressed using the Synthetic Minority Over-sampling Technique (SMOTE). The performance of the models was evaluated through accuracy, precision, recall, F1-score, log loss, and AUC-ROC, and random forest showed the best results (accuracy: 86.43%, F1-score: 86.71%, AUC-ROC: 0.8695). To ensure that such models can be adopted by stakeholders in the health sector, Local Interpretable Model-agnostic Explanations (LIME) was integrated, which identified painful urination, urinary frequency, and suprapubic pain as primary predictors in the model. This study shows that interpretable ML models can be helpful in resource-limited regions in predicting UTIs, thereby rendering a solution to improve the management of infections in these regions.

Abstract: The effect of high-gravity fields, generated by rapid rotation, on CO₂ adsorption in activated carbon beds is examined. Adsorption-desorption kinetics is monitored before, during, and after short rotation periods at up to 5,000rpm. Rotation induced a reproducible transient bump in headspace pressure, quantitatively attributed to a centrifugal free energy shift (~12.2 J/mol) that overfilled weak adsorption sites beyond their static equilibrium. The bump mechanism is described by fold catastrophe theory, with a critical angular velocity (ωc=3,500rpm) triggering a sudden transition to a high-occupancy branch. Post-rotation, constant-rate zero-order desorption from shallow sites overlapped with a slower pseudo-first-order adsorption process as deep, previously inaccessible pores became available, increasing CO₂ capacity by 18.4%. Kinetic modelling produced an apparent diffusivity of 1.2x10-5m2/s and a structural accessibility time constant of ~25h. Thermodynamic analysis showed that rotation improved the overall free energy of adsorption and altered entropy in a manner consistent with the observed adsorption-desorption sequence. These results demonstrate that rotational fields can enhance CO₂ uptake, modify kinetic pathways, and trigger threshold phenomena in porous adsorbents.

Abstract: We study the canonical string-based Assembly Index (ASI), defined as the minimum number of binary concatenations needed to construct a target word under full reuse. NP-completeness of ASI-DEC over general finite alphabets and an equivalence between ASI plans and straight-line programs (SLPs) under the same size convention has been established. We emphasize that all transfers between decision variants are effected by explicit polynomial-time mappings and (where needed) an explicit reparameterization of the threshold by an absolute constant or a simple affine function. The remaining technical obstacle for the binary alphabet is that a naive encoding reduction may allow an optimizer to exploit “cross-boundary” substrings created by overlaps of codewords. We give a fully self-contained binary-alphabet proof: we construct an explicit self-synchronizing (comma-free) codebook of 17 fixed-length binary codewords and prove a boundary-normalization lemma showing that optimal plans can be assumed aligned to codeword boundaries. This yields a polynomial reduction from fixed-alphabet ASI-DEC to binary ASI-DEC, proving NP-completeness over {0, 1}. Using the recalled ASI–SLP equivalence (with a short proof for completeness), we obtain NP-completeness of binary SLP-DEC. We additionally provide an explicit, fully formal translation between our binary-rule counting convention and the standard SGP size measure (sum of right-hand side lengths), showing that the NP-completeness classification transfers to common one-string SGP/SLP decision variants over {0, 1}.

Abstract: Peripheral artery disease (PAD) is a pervasive atherosclerotic condition affecting well over 100 million adults worldwide and associated with major functional limitations, reduced quality of life, and elevated risks of myocardial infarction, stroke, limb events, and mortality. Exercise therapy—preferably supervised or delivered through structured, monitored home based programs—is a first line, guideline endorsed therapy that improves walking performance and patient reported outcomes and contributes to comprehensive secondary prevention. This review synthesizes mechanistic underpinnings (endothelial, angiogenic, metabolic, autonomic) and appraises the comparative effectiveness, safety, and implementation models of supervised exercise therapy (SET), structured home based and hybrid programs, and alternative modalities in PAD. Finally, we summarize policy aspects and persistent gaps to guide clinical practice and future research.

Abstract: Background: Tuberculosis (TB), caused by Mycobacterium tuberculosis, remains a major global health challenge and a leading infectious cause of death, with 10.6 million cases and 450,000 rifampicin-resistant cases reported in 2021. The rise of multi-drug-resistant (MDR) and extensively drug-resistant TB (XDR-TB), driven by mutations in genes such as rpoB, katG, inhA, gyrA, and rrs, threatens the practical control of TB. In the Eastern Cape, South Africa, limited data exist on the patterns of resistance-conferring mutations. This study investigated the molecular profiles of genetic mutations associated with first-line and second-line anti-tuberculosis drug resistance, including fluoroquinolones and injectable agents, among Mycobacterium tuberculosis isolates to inform region-specific diagnostics and treatment strategies. Methods: A retrospective cross-sectional laboratory-based design was used to analyze 112 phenotypically confirmed drug-resistant isolates. Molecular DST for first- and second-line anti-tuberculosis drugs was performed at the National Health Laboratory Service (NHLS) TB reference laboratory. Drug-resistance profiles were classified according to World Health Organization (WHO) definitions. Results: rpoB (D435V 40.2%; S450L 36.6%) and katG (S315T 80.4%) mutations predominated, forming the MDR backbone, while 15% harbored inhA promoter mutations linked to low-level cross-resistance. Nearly 48.2% showed dual resistance to fluoroquinolones and second-line injectables. A significant association between rpoB S450L and dual second-line resistance (p=0.0019) suggests genomic progression toward XDR-TB. The predominance of stable high-fitness resistance mutations and the substantial burden of dual second-line resistance suggest sustained community transmission of established multidrug-resistant strains. These findings underscore the importance of integrating molecular surveillance with community-engaged prevention strategies and strengthened clinical governance to interrupt transmission and limit progression toward advanced resistance in high-burden rural settings. They further reinforce the value of genotype-based diagnostics and expanded genomic surveillance within routine TB programs. Incorporating predictive analytics into programmatic practice will enhance early detection, optimize treatment selection, and support sustained progress toward TB control and eventual elimination.

Abstract: Hypertension is an important target for primordial prevention of complex, noncommunicable diseases and its prevalence remains high across populations. Urban population in India is at a high risk of hypertension but the genetic basis of hypertension in this population remains poorly understood. We conducted a pooled whole-blood genome-wide association study of 28 pools representing 1,402 participants of the Diabetes In Sindhi Families In Nagpur (DISFIN) study which enrolled families of probands with type 2 diabetes (T2D). Genotyping was done using Illumina’s Global Screening Array. From a total of 608,550 single nucleotide variants, 191 were found to be significantly associated with hypertension even after adjusting for metabolic comorbidities, batch effects, pooling error, kinship status and pooling variation. These variants mapped to 180 well-characterized genes that comprised 55 (31%) genes encoding long noncoding RNA (lncRNA). Many of the genes significantly associated with hypertension (including 35% of the lncRNAs) have also been reported by other studies. However, we identified novel genes (SBF2, ARHGAP12, EPAS1, CLEC16A and LRPPRC) to be associated with hypertension. The most significantly associated lncRNA gene was FLYWCH-AS1. Bioinformatic analyses indicated that these novel genes are likely to have functional importance in hypertension. Our study thus points to the potential candidate genes associated with hypertension in endogamous Sindhi families with T2D patients. The replicable and functional role of these candidate genes should be investigated in future studies.

Abstract: Urban shrinkage, driven by demographic and socioeconomic change, has become a pressing issue across Europe, particularly in small peripheral towns and semi-urban settlements that have historically relied on a single industry or company. This study investigates the demographic and socioeconomic factors contributing to the decline in Latvian mono-towns, thereby filling a void in empirical research on urban development in post-socialist contexts. Principal component analysis (PCA) was applied to a set of key demographic and socioeconomic indicators derived from census and administrative data to identify the principal dimensions driving urban shrinkage. The analysis reveals three principal components explaining 87% of the variance: socioeconomic vitality (57.1%), population change and peripherality (17.2%), and aging society dynamics (12.6%). The results contribute to a nuanced understanding of how mono-functional urban contexts shape the intensity and character of shrinkage. These results establish a basis for specific policy measures designed to promote resilience in small-settlement settings and contribute to the understanding of spatial planning and regional development approaches in the post-socialist urban transition context. The research underscores the need for context-specific approaches to address the multifaceted challenges of urban shrinkage.

Abstract: This study addresses the need for intelligent condition monitoring in high-complexity medical imaging systems by proposing a smart sensing architecture for the Revolution EVO Computed Tomography (CT) scanner. Ensuring operational reliability and minimizing unexpected downtime remain critical challenges in advanced CT platforms, motivating the integration of distributed sensing and data-driven analytics. The proposed framework combines Smart Sensor Networks with Machine Learning (ML)-based analysis to enable continuous acquisition and synchronization of heterogeneous operational data from key subsystems, including the X-ray tube assembly, detector array, rotational gantry mechanism, and data acquisition and processing unit. Multivariate feature extraction and sensor-level data fusion are employed to support anomaly detection and predictive assessment of system behavior. The methodology is informed by technical documentation and system specifications provided by GE HealthCare, together with established approaches in intelligent sensing and predictive analytics. The results demonstrate that structured integration of multi-sensor data and ML-based inference can enhance diagnostic sensitivity and enable early identification of abnormal operational patterns. It is concluded that a sensor-centric monitoring architecture provides a feasible pathway toward improved reliability, reduced unplanned interruptions, and more efficient lifecycle management of CT imaging systems.

Abstract: We deploy a spiking neural network (SNN)-equivalent intrusion detection system (IDS) on the STM32N6570-DK, a commodity ARM Cortex-M55 MCU with the Neural-ART NPU. Exploiting the approximate equivalence between single-timestep (T=1) SNN inference and INT8 quantized ANN inference, we compile a lightweight MLP classifier to the NPU without neuromorphic hardware. Evaluated on NSL-KDD (5-class) and UNSW-NB15 (10-class) with 10 random seeds, the ReLU model achieves 78.86±1.32% and 64.75±0.61% overall accuracy, respectively. INT8 accuracy stays within 1 percentage point of FP32 across all 24 tested calibration configurations, and layer-wise analysis shows 99.0% final prediction agreement between FP32 and INT8 models. On the NPU, the INT8 model infers in 0.46 ms on NSL-KDD and 0.29 ms on UNSW-NB15 (100% NPU execution), occupying 120.6–137.7 KB Flash and 0.5–1.25 KB RAM. A comparison with QCFS activation reveals that the Floor operator falls back to CPU on this NPU, adding 17.6% latency. Tree-based baselines (Random Forest, XGBoost) confirm that the MLP offers the best accuracy on NSL-KDD while being the only model eligible for NPU acceleration. To our knowledge, this is the first IDS deployment on an ARM Cortex-M NPU and the first empirical validation of T=1 SNN–ANN equivalence on commercial NPU silicon.

Abstract: Background: Advances in machine learning (ML) based survival modeling enable the analysis of high-dimensional biomedical data. However, many approaches rely on the proportional hazards (PH) assumption, which is frequently violated in oncology and can limit the interpretability of hazard ratio–based results. Using Estrogen Receptor (ER) status in the METABRIC breast cancer cohort as a case study, we propose a framework that integrates machine learning survival models with Restricted Mean Survival Time (RMST) to provide a more robust and clinically interpretable approach for survival analysis under non-proportional hazards. Methods: Overall survival was analyzed in 1104 patients. PH violations were confirmed using Schoenfeld residuals and Kaplan–Meier inspection. We compared four models: stratified Cox Elastic Net (Cox E-Net), Random Survival Forest (RSF), Gradient Boosting Survival Analysis (GBSA), and DeepHit. Performance was assessed using Harrell’s C-index, time-dependent IPCW C-index, and Integrated Brier Score (IBS). RMST at 180 months was utilized to quantify absolute survival differences between ER subgroups. To improve the stability of the estimates, 200 bootstrap resamples were performed, and 95% confidence intervals were derived from the bootstrap distribution. Results: ER status demonstrated significant PH violation (p < 0.005) with crossing survival curves. Discrimination (C-index 0.664–0.725) and calibration (IBS 0.149–0.169) were comparable across models, with RSF achieving the highest overall performance. Despite similar accuracy, survival curve structures differed substantially. Cox E-Net and RSF reproduced the observed crossing pattern, whereas GBSA generated smoother trajectories and DeepHit showed marked compression of subgroup separation. In the independent test cohort, the empirical RMST difference at 180 months was 16.6 months (ER-positive: 130.4; ER-negative: 113.8). Model-based RMST differences ranged from 1 month (DeepHit) to 27 months (Cox E-Net), with RSF and GBSA (12.8 and 13.8 months) most closely approximating the empirical benchmark. Conclusions: We propose a novel, model-agnostic ML + RMST framework that addresses non-proportional hazards while providing quantifiable, time-specific clinical benefit. Moreover, models with similar discrimination and calibration produced markedly different survival curve behavior and absolute RMST estimates, demonstrating that accuracy metrics alone are insufficient for clinical interpretation. By linking predictive modeling with absolute survival quantification, this framework advances survival evaluation beyond relative risk ranking toward clinically meaningful decision support.