Real-Time Inferential Analytics Based on Online Databases of Trends : A Breakthrough within the Discipline of Digital Epidemiology in Dentistry and Dental Anatomy

CORRESPONDING AUTHOR Dr Ahmed Al-Imam: Department of Anatomy and Cellular Biology, College of Medicine, University of Baghdad, Bab Al-Moadham, 10053, Baghdad, Iraq. Email: tesla1452@gmail.com | a.m.al-imam@herts.ac.uk Phone number: +964 (0) 771 433 8199 ResearchGate Account: https://www.researchgate.net/profile/Ahmed_Al-Imam Scopus Account: https://www.scopus.com/authid/detail.uri?authorId=57191594132 Scopus Author ID: 57191594132


INTRODUCTION
Digital epidemiology is an emerging discipline of public health and epidemiological sciences, and it has been evolving rapidly over the past few decades.[1,2] It can be implemented based on data from online resources of the surface web including trends databases, online drug fora and blogs, and social communication media.[3,4] Google Trends database perfectly fits this purpose as it contains an automated and up-to-date collection of data based on queries of users of the web from all over the world, including millions if not billions of users.[2,5] Epidemiologist never attempted to carry out real-time or predictive analytics within the context of digital epidemiology and in connection with the discipline of dentistry or medicine.[6][7][8] In this study, we will explore this concept via the integrative use of Python programme language, statistical packages, and spreadsheet templates in an aim to demonstrate a prototype for real-time analytic of data retrieved from Google Trends.
The primary objective of this study is to demonstrate a prototype of real-time analysis and to infer data on the digital epidemiology in connection with topics that are related to the maxillary sinus and the Schneiderian membrane, posterior maxillary teeth, and related oral pathologies including periapical abscess formation, periodontal pathologies, and complicated dental implants.Real-time analytics are of considerable impact when implemented in biological and life sciences as they will tremendously reduce the required resources for research.On the other hand, Predictive analytics based on artificial neural networks and machine learning algorithms, can be the next step to be deployed in continuation of the real-time systems to prognosticate changes in the temporal trends and the digital epidemiology of phenomena of interest in medicine, dentistry, as well as other subdisciplines of biological and life sciences.
The paranasal sinuses are of prime importance for the region of the head.[9,10] Leonardo da Vinci (1452-1519) made the very first illustration of the maxillary sinus as well as other paranasal sinuses (Figure 1).[11,12] Following a maxillary molar tooth extraction, the treatment modalities routinely involve dental prostheses.However, the central fossa of the candidate implant site may require bone grafting techniques to carry out a satisfactory surgery outcome at the prospective implantation site.[13] Oberli and workmates analysed a series of one hundred thirteen periapical radiographs of maxillary premolars and molars with periapical radiolucency indicating chronic apical periodontitis.The cohort was evaluated for the occurrence of maxillary sinus perforations and postoperative complications.Perforation of the Schneiderian membrane occurred in 9.6% of the cases, while membrane exposure without rupture existed in 12%.The distance between the apex of the periapical lesion and the sinus floor did not serve as a predictor of a potential sinus membrane rupture.[14] In 2013, Dagassan-Berndt and fellow workers measured the thickness of the Schneiderian membrane via dental Cone-Beam Computed Tomography (CBCT).It was significantly higher in the dentate group compared to the edentulous group in connection with the position of the first and second molar.Further, in the dentate group, clinical signs of periodontal destruction were not associated with Schneiderian membrane thickness.[15] 2. MATERIALS AND METHODS

ETHICAL APPROVAL
This study has been ethically permitted by the Institute Review Board (IRB) of the College of Medicine at the University of Baghdad and in compliance with the authority of the IRB meeting number seven that took place on the 20 th of December in 2016.

REVIEW OF LITERATURE
An analysis of the existing body of literature was conducted systematically from the 1 st to the 15 th of August 2018 via medical and paramedical databases including NCBI-PubMed, the Cochrane Library, and Elsevier.The unpublished grey literature was also consulted for data of interest.The concept of real-time and predictive analytics was never explored medical and paramedical literature (Table 1A).Further, keywords of different themes were utilised in the process of examining the databases of published research in connection with the maxillary sinus anatomy and related pathologies (Table 1B).Themes included five different topics including "Premolars and Molars", "Maxillary Sinus", "Pathologies", "Surgical Procedures", and "Radiology".We applied different combinations of keywords and themes, via the implementation of Boolean Operators (AND, OR, NOT).[16] Bibliographic materials of interest were assessed and appraised for validity and rigorousness via critical appraisal tools.[17,18] Duplicate publications were eliminated, and studies that successfully passed the critical appraisal were deemed as satisfactory reference materials.Those studies were conducted on humans as well as non-human species, and written exclusively in the English language.Priority was given to recently-published literature within the past 5-10 years.

EXPLORATION OF GOOGLE TRENDS
Data were extracted from Google Trends database for the past five years from the 18 th of August 2013 to the 18 th of August 2018.[19] We used five keywords to retrieve raw numerical week-by-week particulars on the temporal trends, geographic mapping, and related queries by web users.Keywords included "Schneiderian membrane", "Maxillary Sinus", "Sinus lift", "Endodontics", and "Periodontal disease".Our study is a hybrid of a cross-sectional analysis via an internet snapshot, as well as real-time analytics of the trends.Hence, the level-of-evidence for this study cannot be categorised in correspondence with the Oxford Centre for Evidence-Based Medicine (CEMB).[20] Real-time analysis was attempted via the integration of Python programming language and Microsoft Excel 2016.This concept was never tried as confirmed the complete absence of published literature relevant to the topic of real-time analysis (Table 1B).Statistical analyses and hypotheses testing, descriptive and inferential, were carried out via Microsoft Excel 2016 and the Statistical Package for Social Sciences (SPSS v.24).The implemented statistical tests included the Analysis of Variance and Covariance (ANOVA), Student's t-test, and Linear Regression.An alpha value (α) of 0.05 and a confidence interval of 95% (95% CI) are considered as the cut-off margin for statistical inference.

A PROTOTYPE OF REAL-TIME ANALYSIS OF GOOGLE TRENDS
To achieve real-time analytics based on data already available on Google Trends (Figure 2), we wrote a script (programming code) via Python high-level programming language version 3.6.6,32-bit using Linux Deepin 15.6 (64-bit) and Windows 10 Pro (64-bit) operating systems.[21] We applied Thonny version 2.1.21interpreter, a Python Integrated Development Environment (IDE).The interpreter is a computer program that directly executes the programming script.[22,23] Two libraries (modules) were imported, Pytrends and OpenPyXl.Those modules are a collection of precompiled routines that a program can use.[24,25] 3. RESULTS

DATABASES OF LITERATURE
The systematic inspection of databases of interest of the published literature yielded a total of 491813 hits distributed as 488884 (PubMed-NCBI), 1611 (the Cochrane Library), and 1318 (Elsevier).The most successful keywords to retrieve data addressing the research questions included two combinations of keywords seen in bold fonts (Table 1) generating 441 and 30 hits respectively.

GOOGLE TRENDS
Exploration of Google Trends database gave data on related queries from users of the surface web.Those queries were not limited to "Maxillary sinus cyst", "Maxillary sinus infection", "Maxillary sinus pain", "Maxillary sinusitis", "Maxillary sinus retention cyst", "Sinus lift surgery", "Sinus graft", "Sinus augmentation", "Dental implants", "Gum disease", "Periodontitis", "Gingivitis", "Gum disease treatment", and "Periodontal treatment" (Table 2).Google Trends also led to accurate data about the geographic mapping (geo-mapping) of the web users queries towards topics of interest in connection with the maxillary sinus, the posterior maxillary teeth, and related oral pathologies.Geo-mapping was limited to forty-seven countries including Japan, Taiwan, Chile, Germany, Ecuador, United Kingdom, Bulgaria, Ireland, Peru, Italy, Spain, Austria, Venezuela, Brazil, Mexico, Colombia, Greece, United States, Australia, Norway, New Zealand, South Korea, Sweden, Switzerland, Portugal, Ukraine, Singapore, Russia, France, Canada, Romania, Belgium, the Netherlands, Philippines, Argentina, South Africa, Egypt, Malaysia, United Arab Emirates, Kingdom of Saudi Arabia, Poland, India, Pakistan, Thailand, Indonesia, Iran, and Turkey.
Countries from the Middle East accounted for 10.64% while countries that represented statistical outliers has contributed to 6.38% of the global map (Figure 3).Those outliers were related to the keyword "Sinus lift" and included Austria, Romania, and Turkey.Concerning geo-mapping, the "Schneiderian Membrane" generated no hits at all while other keywords averaged 2.81 +/-0.63 (Maxillary sinus), 1.47 +/-0.39 (Sinus lift), 27.43 +/-2.76 (Endodontics), and 68.30 +/-3.14 (Periodontal Disease) (Table 3).Based on Student's t-test statistics, there was a statistically significant difference between all keywords with an exception for "Maxillary sinus" versus "Sinus lift" (p-value=0.091)(Table 4).Therefore, surface web users were most interested in periodontal diseases and endodontics.

REAL-TIME ANALYSIS OF GOOGLE TRENDS
The programming script (Python's) enables an automatic retrieval of data from Google Trends based on keywords of interest (up to five).The retrieval process is in real-time, and at a regular interval (weekly, biweekly, monthly, etc.) that can be customised according to the research requirement.The data were self-regulated to be transferred to an Excel spreadsheet template (Microsoft Excel 2016, 64-bits) that had built-in formulas for statistical analysis and multiple hypothesis testing for statistical inference and in realtime.A Complimentary statistical analysis was carried out via SPSS.Eventually, we succeeded in generating a real-time stream of inferential analytics.

THE CONCEPT OF REAL-TIME ANALYSIS
The concept of real-time analysis was never explored before not only within the field of dental anatomy and dental pathology but also in connection with the entire discipline of Medicine and Dentistry.[26] We used keywords that are specific to the proposed research questions on the maxillary sinus, maxillary teeth and related oral pathologies (Table 1B).The total number of hits, representing published papers was 346146, most of which (342841) were indexed via PubMed-NCBI, and much less (3305) were found on the Cochrane Library, while none existed on Elsevier database.Nevertheless, there was no single study attempting real-time analytic.Hence, there is a full deficit within the existing body of literature about the objectives of this study.Our study is the first of its kind according to which an automated real-time stream of data accompanied by a statistical analysis was applicable based on data extrapolated from Google Trends.

LIMITATIONS OF THE REAL-TIME ANALYSIS
The digital epidemiological analysis can be applied via online databases of trends.Geographic mapping of the top contributing countries originated mainly from the developed world, as well as few countries from the Middle East, some Latin countries, and others from Eastern Europe.This study may have some limitations due to the sole reliance on Google Trends as a representative of the databases of trends existing on the surface web.Besides, the retrospective analytic part of the study was specific for a restricted period (2013-2017).Data collected from Google Trends might be occasionally faulty or misleading as some web users might be deploying the use of an incognito mode of web browsing or dedicated incognito web browsers including Tor Browser, virtual private networks and internet protocol masking.Subsequent studies should incorporate more than one trends database for cross-validation.
Additionally, ventures into the deep web and the darknet should be attempted to recover any relevant data including those on the geographic mapping and temporal trends whenever feasible.

LITERATURE REVIEW OF RELEVANCE TO MAXILLARY SINUS AND POSTERIOR MAXILLARY TEETH
Lozano-Carrascal (2014) confirmed, by means of a cross-sectional analytic study, that CBCT scanning has been shown to be a useful tool for evaluating maxillary sinus variations of anatomical parameters including the residual ridge height and width, ridge bone density, maxillary sinus angle, maxillary sinus lateral wall thickness, Schneiderian membrane thickness, maxillary sinus septa, and the posterior superior alveolar artery.CBCT represents the gold standard tool for evaluating the maxillary sinus area.[27] Concerning the maxillary septa, Bornstein and co-workers (2016) studied the of frequency, morphology, and locations of septa using CBCT.Septa are common anatomical structures that often exist in the region of the first or second molar on the floor of the maxillary sinus.Proper pre-operative assessment of the septa, via three-dimensional radiographic examination, is mandatory to avoid potential complications during sinus floor elevation procedures.[28] In 2015, Goller-Bulut and teammates studied the relationship between periapical and marginal bone loss and the mucosal thickness of the maxillary sinus in connection with maxillary premolars and molars.
Retrospective analysis showed that mucosal thickening (MT) of the maxillary sinus was common among patients with the periodontal bone loss (PBL), and it was significantly associated with apical lesions and PBL.[29] Bayrak and colleagues (2018) conducted another retrospective evaluation and found no statistically significant relationship between nasal septum deviation (NSD) and Schneiderian membrane thickness (SMT).The deviation of the nasal septum was found in 50.6% of patients without gender bias.However, the average SMT was found to be higher in males for all the examined tooth areas (p-value≤0.05).The average thickness of the membrane was highest in the first molar region and least in the third molar region.[30] Earlier in 2017, Khorramdel and fellows published similar data confirming that periapical lesions and periodontal infections in the posterior maxilla were associated with Schneiderian membrane thickening.Besides, there was a significant relationship between the location of the posterior maxillary teeth and the thickness of the Schneiderian membrane.[31] In 2014, a study by Acharya and co-workers revealed that the incidence of advanced periodontal disease was common among Hong Kong Chinese and Asian Indian subjects who sought tooth replacement.Ethnicity, sex, and sinus membrane thickening affected the available bone height in the region below the maxillary sinus.[32] Lu and colleagues (2012) validated, via CBCT imaging, that the prevalence and extent severity of the maxillary sinus mucosal thickening were positively associated with the severity of apical periodontitis.CBCT imaging is invaluable for the assessment of the maxillary sinuses and related teeth primarily molars and premolars.Patients in their 7 th decade of life had the highest prevalence of mucosal thickening.[33] During the same year, Bornstein and co-authors explored the characteristics and dimensions of the Schneiderian membrane and apical bone in maxillary molars in patients referred for apical surgery.The thickness of the apical bone and the Schneiderian membrane were generally higher in patients with periapical pathoses.[34] For single-tooth implant replacement in the posterior maxilla, it is often necessary to do an augmentation of the alveolar process to overcome the post-extraction bone resorption.In 2011, Kahnberg presented a technique for a local sinus lift with autogenous bone in a one-stage procedure.The researchers validated that local sinus lift with simultaneous bone augmentation and single-tooth replacement in the posterior maxilla is a predictable method although a certain bone volume reduction around the implant was evident at the 2-year follow-up.[35] On the other hand, Taschieri and co-researchers (2014) validated via their experimental trial, that the use of platelet concentrates can be effective in reducing the pain, swelling, and surgery-related side effects.[36] In 2017, Eggmann and colleagues concluded, based on a systematic review of studies implementing CBCT scans, that periapical lesions in the posterior maxilla are likely to be associated with Schneiderian membrane thickening.On the other hand, the current evidence regarding the relation between periodontal diseases and the appearance of the Schneiderian membrane in CBCT is inconclusive.[37] Monje and co-workers (2016), based on their systematic review and meta-regression analytic of thirtyone studies, reported that the overall mean ± SE maxillary Schneiderian membrane thickness (SMT) was 1.17 ± 0.1 mm.The mean SMT for the three-dimensional radiography (3DR) group was 1.33 mm, and 0.48 mm for the histology group.Three-dimensional technologies overestimate the SMT by approximately 2.5 folds in comparison with histologic analysis, yet the difference between the two measurement methods was not statistically significant.Regression analyses, though statistically insignificant, demonstrated that patient with periodontitis and smoking had thicker SMT.The thicker SMT might be more vulnerable to perforation (p-value=0.14).On average, SMT is 1 mm in patients seeking sinus augmentation.[38] Several other pathologies can affect the maxillary sinus and the subjacent alveolar ridge as documented by Evrard, Jham, and Kessler.[39][40][41] Evans reported a tumour attached to the cemento-enamel junction of an impacted third molar at the region of the maxillary tuberosity.The lesion represented an unusual presentation for an odontogenic myxoma in which an aggressive squamous epithelial component existed, altogether with a mucus-secreting component.[34] Jham published a report of a case of an extensive central ossifying fibroma (COF) affecting the maxilla of a 62-year-old patient.The lesion had undefined limits on the right side of the upper alveolar ridge and extending towards the palate.The considerable growth potential of COF manifested in this case by the invasion of the maxillary sinus, nasal cavity and the orbit.[35] Kessler reported a case of respiratory epithelial adenomatoid hamartoma of the maxillary sinus.
Those hamartomas are rare lesions occurring primarily in the nasal cavity, nasopharynx, and the paranasal sinuses.[36] Table 1A.Investigation of Databases of Literature: The Topic of "Real-time" Analysis.

Figure 1 .
Figure 1.Sagittal Anatomy of the Maxillary Sinus: A Duplicate of Leonardo da Vinci's Original Sketch.† Concept Art by Dawoude Kaouche (co-author).

Figure 2 .Figure 3 .
Figure 2. Flowchart of the Implemented Methods of Research.

Table 1B .
Investigation of Databases of Literature: The Maxillary Sinus.

Table 2 .
Google Trends: Top Related Queries on the Maxillary Sinus.