PLOS digital health最新文献

{"title":"Multicenter comparative analysis of local and aggregated data training strategies in COVID-19 outcome prediction with Machine learning.","authors":"Carine Savalli, Roberta Moreira Wichmann, Fabiano Barcellos Filho, Fernando Timoteo Fernandes, Alexandre Dias Porto Chiavegatto Filho","doi":"10.1371/journal.pdig.0000699","DOIUrl":"10.1371/journal.pdig.0000699","url":null,"abstract":"<p><p>Machine learning (ML) is a promising tool in assisting clinical decision-making for improving diagnosis and prognosis, especially in developing regions. It is often used with large samples, aggregating data from different regions and hospitals. However, it is unclear how this affects predictions in local centers. This study aims to compare data aggregation strategies of several hospitals in Brazil with a local training strategy in each hospital to predict two COVID-19 outcomes: Intensive Care Unit admission (ICU) and mechanical ventilation use (MV). The study included 6,046 patients from 14 hospitals, with local sample sizes ranging from 47 to 1500 patients. Machine learning models were trained using extreme gradient boosting, lightGBM, and catboost for structured data. Seven data aggregation strategies based on hospital geographic regions were compared with local training, and the best strategy was determined by analyzing the area under the ROC curve (AUROC). SHAP (Shapley Additive exPlanations) values were used to assess the contribution of variables to predictions. Additionally, a metafeatures analysis examined how hospital characteristics influence the selection of the best strategy. The study found that the local training strategy was the most effective approach, in the case of ICU outcomes, for 11 of the 14 hospitals (79%), and, in the case of MV, for 10 hospitals (71%). Metafeatures analysis suggested that hospitals with smaller sample sizes generally performed better using an aggregated data strategy compared to local training. Our study brings to light an important concern about the impact of grouping data from different hospitals in predictive machine learning models. These findings contribute to the ongoing debate about the trade-off between increasing sample size and bringing together heterogeneous scenarios.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"3 12","pages":"e0000699"},"PeriodicalIF":0.0,"publicationDate":"2024-12-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11670925/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142901028","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A cluster randomized trial assessing the effect of a digital health algorithm on quality of care in Tanzania (DYNAMIC study).","authors":"Rainer Tan, Godfrey Kavishe, Alexandra V Kulinkina, Sabine Renggli, Lameck B Luwanda, Chacha Mangu, Geofrey Ashery, Margaret Jorram, Ibrahim Evans Mtebene, Peter Agrea, Humphrey Mhagama, Kristina Keitel, Marie-Annick Le Pogam, Nyanda Ntinginya, Honorati Masanja, Valérie D'Acremont","doi":"10.1371/journal.pdig.0000694","DOIUrl":"10.1371/journal.pdig.0000694","url":null,"abstract":"<p><p>Digital clinical decision support tools have contributed to improved quality of care at primary care level health facilities. However, data from real-world randomized trials are lacking. We conducted a cluster randomized, open-label trial in Tanzania evaluating the use of a digital clinical decision support algorithm (CDSA), enhanced by point-of-care tests, training and mentorship, compared with usual care, among sick children 2 to 59 months old presenting to primary care facilities for an acute illness in Tanzania (ClinicalTrials.gov NCT05144763). The primary outcome was the mean proportion of 14 major Integrated Management of Childhood Illness (IMCI) symptoms and signs assessed by clinicians. Secondary outcomes included antibiotic prescription, counseling provided, and the appropriateness of antimalarial and antibiotic prescriptions. A total of 450 consultations were observed in 9 intervention and 9 control health facilities. The mean proportion of major symptoms and signs assessed in intervention health facilities was 46.4% (range 7.7% to 91.7%) compared to 26.3% (range 0% to 66.7%) in control health facilities, an adjusted difference of 15.1% (95% confidence interval [CI] 4.8% to 25.4%). Only weight, height, and pallor were assessed statistically more often when using the digital CDSA compared to controls. Observed antibiotic prescription was 37.3% in intervention facilities, and 76.4% in control facilities (adjusted risk ratio 0.5; 95% CI 0.4 to 0.7; p<0.001). Appropriate antibiotic prescription was 81.9% in intervention facilities and 51.4% in control facilities (adjusted risk ratio 1.5; 95% CI 1.2 to 1.8; p = 0.003). The implementation of a digital CDSA improved the mean proportion of IMCI symptoms and signs assessed in consultations with sick children, however most symptoms and signs were assessed infrequently. Nonetheless, antibiotics were prescribed less often, and more appropriately. Innovative approaches to overcome barriers related to clinicians' motivation and work environment are needed.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"3 12","pages":"e0000694"},"PeriodicalIF":0.0,"publicationDate":"2024-12-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11666054/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142883922","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Measurement of breast artery calcification using an artificial intelligence detection model and its association with major adverse cardiovascular events.","authors":"Suzanne J Rose, Josette Hartnett, Zachary J Estep, Daniyal Ameen, Shweta Karki, Edward Schuster, Rebecca B Newman, David H Hsi","doi":"10.1371/journal.pdig.0000698","DOIUrl":"10.1371/journal.pdig.0000698","url":null,"abstract":"<p><p>Breast artery calcification (BAC) obtained from standard mammographic images is currently under evaluation to stratify risk of major adverse cardiovascular events in women. Measuring BAC using artificial intelligence (AI) technology, we aimed to determine the relationship between BAC and coronary artery calcification (CAC) severity with Major Adverse Cardiac Events (MACE). This retrospective study included women who underwent chest computed tomography (CT) within one year of mammography. T-test assessed the associations between MACE and variables of interest (BAC versus MACE, CAC versus MACE). Risk differences were calculated to capture the difference in observed risk and reference groups. Chi-square tests and/or Fisher's exact tests were performed to assess age and ASCVD risk with MACE and to assess BAC and CAC association with atherosclerotic cardiovascular disease (ASCVD) risk as a secondary outcome. A logistic regression model was conducted to measure the odds ratio between explanatory variables (BAC and CAC) and the outcome variables (MACE). Out of the 99 patients included in the analysis, 49 patients (49.49%) were BAC positive, with 37 patients (37.37%) CAC positive, and 26 patients (26.26%) had MACE. One unit increase in BAC score resulted in a 6% increased odds of having a moderate to high ASCVD risk >7.5% (p = 0.01) and 2% increased odds of having MACE (p = 0.005). The odds of having a moderate-high ASCVD risk score in BAC positive patients was higher (OR = 4.27, 95% CI 1.58-11.56) than CAC positive (OR = 4.05, 95% CI 1.36-12.06) patients. In this study population, the presence of BAC is associated with MACE and useful in corroborating ASCVD risk. Our results provide evidence to support the potential utilization of AI generated BAC measurements from standard of care mammograms in addition to the widely adopted ASCVD and CAC scores, to identify and risk-stratify women who are at increased risk of CVD and may benefit from targeted prevention measures.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"3 12","pages":"e0000698"},"PeriodicalIF":0.0,"publicationDate":"2024-12-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11665981/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142883930","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Does providing atrial fibrillation patients, after pulmonary vein isolation, with a 1-lead ECG device relieve the emergency department?-A historically controlled prospective trial.","authors":"Jasper L Selder, Mark J Mulder, Willem R van de Vijver, Philip M Croon, Leontine E Wentrup, Stéphanie L van der Pas, Jos W R Twisk, Igor I Tulevski, Albert C Van Rossum, Cornelis P Allaart","doi":"10.1371/journal.pdig.0000688","DOIUrl":"10.1371/journal.pdig.0000688","url":null,"abstract":"<p><p>Atrial fibrillation (AF) is a prevalent and clinically significant cardiac arrhythmia, with a growing incidence. The primary objectives in AF management are symptom relief, stroke risk reduction, and prevention of tachycardia-induced cardiomyopathy. Two key strategies for rhythm control include antiarrhythmic drug therapy and pulmonary vein isolation (PVI), with PVI being recommended for selected patients. Even though PVI is effective, post procedural health care utilization is high, contributing to emergency department (ED) overcrowding, which is a global healthcare concern. The use of remote rhythm diagnostics, such as a 1-lead ECG device (KM), may mitigate this issue by reducing ED visits and facilitating more plannable AF care.</p><p><strong>Objective: </strong>This study aimed to assess whether providing AF patients with a 1-lead ECG device for 1 year after PVI would reduce ED utilization compared to standard care. Additionally, the study assessed whether this intervention would render AF care more plannable by reducing the incidence of unplanned cardioversions.</p><p><strong>Methods: </strong>A historically controlled, prospective clinical trial was conducted. The intervention group were patients undergoing PVI for AF between September 2018 and August 2020, all patients in this group received a 1-lead ECG device for 1 year for remote rhythm monitoring. The historical control group were patients undergoing PVI between January 2016 and December 2017; these patients did not receive a 1-lead ECG device. Data on ED visits, planned and unplanned cardioversions, and outpatient contacts in the year after the PVI were collected for both groups.</p><p><strong>Results: </strong>The study included 204 patients, 123 in the 1-lead ECG group and 81 in the standard care group. There was no statistically significant difference in the number of all-cause ED visits (63 vs 68 per 100 pts, respectively, p = 0.72), ED visits for possible rhythm disorders, or ED visits for definite rhythm disorders between the two groups. However, the 1-lead ECG group demonstrated a higher proportion of planned cardioversions compared to unplanned ones (odds ratio 4.9 [1.57-15.85], p = 0.007).</p><p><strong>Conclusion: </strong>Providing patients with AF following PVI with a 1-lead ECG device did not result in a statistically significant reduction in ED visits during the first year. However, it did improve the management of recurrent AF episodes by substituting unplanned cardioversions with scheduled ones. Clinical Trials Registration Number NCT06283654.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"3 12","pages":"e0000688"},"PeriodicalIF":0.0,"publicationDate":"2024-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11661629/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142869673","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A voice-based algorithm can predict type 2 diabetes status in USA adults: Findings from the Colive Voice study.","authors":"Abir Elbéji, Mégane Pizzimenti, Gloria Aguayo, Aurélie Fischer, Hanin Ayadi, Franck Mauvais-Jarvis, Jean-Pierre Riveline, Vladimir Despotovic, Guy Fagherazzi","doi":"10.1371/journal.pdig.0000679","DOIUrl":"10.1371/journal.pdig.0000679","url":null,"abstract":"<p><p>The pressing need to reduce undiagnosed type 2 diabetes (T2D) globally calls for innovative screening approaches. This study investigates the potential of using a voice-based algorithm to predict T2D status in adults, as the first step towards developing a non-invasive and scalable screening method. We analyzed pre-specified text recordings from 607 US participants from the Colive Voice study registered on ClinicalTrials.gov (NCT04848623). Using hybrid BYOL-S/CvT embeddings, we constructed gender-specific algorithms to predict T2D status, evaluated through cross-validation based on accuracy, specificity, sensitivity, and Area Under the Curve (AUC). The best models were stratified by key factors such as age, BMI, and hypertension, and compared to the American Diabetes Association (ADA) score for T2D risk assessment using Bland-Altman analysis. The voice-based algorithms demonstrated good predictive capacity (AUC = 75% for males, 71% for females), correctly predicting 71% of male and 66% of female T2D cases. Performance improved in females aged 60 years or older (AUC = 74%) and individuals with hypertension (AUC = 75%), with an overall agreement above 93% with the ADA risk score. Our findings suggest that voice-based algorithms could serve as a more accessible, cost-effective, and noninvasive screening tool for T2D. While these results are promising, further validation is needed, particularly for early-stage T2D cases and more diverse populations.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"3 12","pages":"e0000679"},"PeriodicalIF":0.0,"publicationDate":"2024-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11658629/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142866564","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"EXAM: Ex-vivo allograft monitoring dashboard for the analysis of hypothermic machine perfusion data in deceased-donor kidney transplantation.","authors":"Simon Schwab, Hélène Steck, Isabelle Binet, Andreas Elmer, Wolfgang Ender, Nicola Franscini, Fadi Haidar, Christian Kuhn, Daniel Sidler, Federico Storni, Nathalie Krügel, Franz Immer","doi":"10.1371/journal.pdig.0000691","DOIUrl":"10.1371/journal.pdig.0000691","url":null,"abstract":"<p><p>Deceased-donor kidney allografts are exposed to ischemic injury during ex vivo transport due to the lack of blood oxygen supply. Hypothermic machine perfusion (HMP) effectively reduces the risk of delayed graft function in kidney transplant recipients compared to standard cold storage. However, no free software implementation is available to analyze HMP data for state-of-the-art visualization and quality control. We developed the tool EXAM (ex-vivo allograft monitoring) as an interactive analytics dashboard. We wrote functions in the R programming language to read, process, and analyze HMP data from the LifePort kidney transporter (Organ Recovery Systems, USA). Time series for pressure, flow rate, organ resistance, and temperature are visualized, and relevant statistical indicators have been developed. We explain how data were processed, and indicators were calculated, and we present summary statistics for N = 255 kidney allografts receiving machine perfusion in Switzerland between 2020 and 2023. Median (interdecile range, IDR) of the main indicators were as follows: perfusion duration 5.18 hours (2.29-11.2), flow rate 110 ml/min (52.9-167), ice temperature 1.97°C (1.53-3.07), and perfusate temperature 6.68°C (5.58-8.36). We implemented the dashboard to identify issues, such as atypical perfusion parameters, high ice, or high perfusate temperature to inform transplant centers for quality assurance. In conclusion, EXAM is a free tool that statisticians and data scientists can quickly deploy to enable quality control at transplant organizations that use LifePort kidney transporters. An online viewer is available at https://data.swisstransplant.org/exam/.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"3 12","pages":"e0000691"},"PeriodicalIF":0.0,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11654979/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142857230","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The feasibility of a visuo-cognitive training intervention using a mobile application and exercise with stroboscopic glasses in Parkinson's: Findings from a pilot randomised controlled trial.","authors":"Julia Das, Gill Barry, Richard Walker, Rodrigo Vitorio, Yunus Celik, Claire McDonald, Bryony Storey, Paul Oman, Rosie Morris, Samuel Stuart","doi":"10.1371/journal.pdig.0000696","DOIUrl":"10.1371/journal.pdig.0000696","url":null,"abstract":"<p><strong>Background: </strong>There is currently no pharmacological treatment for visuo-cognitive impairments in Parkinson's disease. Alternative strategies are needed to address these non-motor symptoms given their impact on quality of life. Novel technologies have potential to deliver multimodal rehabilitation of visuo-cognitive dysfunction, but more research is required to determine their feasibility in Parkinson's.</p><p><strong>Objective: </strong>To determine the feasibility and preliminary efficacy of a home-based, technological visuo-cognitive training (TVT) intervention using a mobile application and exercise with stroboscopic glasses compared to non-technological care in people with Parkinson's.</p><p><strong>Methods: </strong>This 18-month, parallel, two-arm pilot trial took place between July 2021-December 2022. Participants were community-dwelling individuals with a diagnosis of Parkinson's, aged over 50 years. Participants were randomly allocated to one of two active four-week interventions, TVT (n = 20) or standard care (SC) (n = 20). A physiotherapist delivered 8 home visits over 4 weeks, lasting 45-60 mins. Participants were evaluated at baseline and then on completion of the intervention. Primary outcomes were feasibility of the study design and intervention (recruitment/retention, adherence, assessment time scale, equipment and safety). Exploratory outcomes included assessments of cognitive, visual, clinical and motor function. (Blinding of participants was not possible due to the nature of the intervention).</p><p><strong>Results: </strong>The recruitment rate was 60% (40/67), and the retention rate was 98% (39/40). Adherence to both arms of the intervention was high, with participants attending 98% of visits in the TVT group and 96% of visits in the SC group. 35% (9/20) of participants in the TVT group experienced mild symptoms associated with use of the stroboscopic glasses which included dizziness, queasiness and unsteadiness. There were minimal between group differences, with both interventions having positive effects on a variety of clinical, cognitive, and physical performance outcomes.</p><p><strong>Conclusions: </strong>Our findings suggest that home-based TVT with a physiotherapist is feasible in people with Parkinson's and could provide an alternative approach to addressing cognitive and motor dysfunction in this population. We make recommendations for future trials and invite ensuing studies to improve upon the design and utilise stroboscopic visual training and digital tools to investigate this emerging area of multimodal rehabilitation. This trial was prospectively registered at ISRCTN (registration number: ISRCTN46164906; https://doi.org/10.1186/ISRCTN46164906).</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"3 12","pages":"e0000696"},"PeriodicalIF":0.0,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11654989/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142855866","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Differential behaviour of a risk score for emergency hospital admission by demographics in Scotland-A retrospective study.","authors":"Ioanna Thoma, Simon Rogers, Jillian Ireland, Rachel Porteous, Katie Borland, Catalina A Vallejos, Louis J M Aslett, James Liley","doi":"10.1371/journal.pdig.0000675","DOIUrl":"10.1371/journal.pdig.0000675","url":null,"abstract":"<p><p>The Scottish Patients at Risk of Re-Admission and Admission (SPARRA) score predicts individual risk of emergency hospital admission for approximately 80% of the Scottish population. It was developed using routinely collected electronic health records, and is used by primary care practitioners to inform anticipatory care, particularly for individuals with high healthcare needs. We comprehensively assess the SPARRA score across population subgroups defined by age, sex, ethnicity, socioeconomic deprivation, and geographic location. For these subgroups, we consider differences in overall performance, score distribution, and false positive and negative rates, using causal methods to identify effects mediated through age, sex, and deprivation. We show that the score is well-calibrated across subgroups, but that rates of false positives and negatives vary widely, mediated by various causes including variability in demographic characteristics, admission reasons, and potentially differential data availability. Our work assists practitioners in the application and interpretation of the SPARRA score in population subgroups.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"3 12","pages":"e0000675"},"PeriodicalIF":0.0,"publicationDate":"2024-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11651550/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142848384","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Neonatal apnea and hypopnea prediction in infants with Robin sequence with neural additive models for time series.","authors":"Julius Vetter, Kathleen Lim, Tjeerd M H Dijkstra, Peter A Dargaville, Oliver Kohlbacher, Jakob H Macke, Christian F Poets","doi":"10.1371/journal.pdig.0000678","DOIUrl":"10.1371/journal.pdig.0000678","url":null,"abstract":"<p><p>Neonatal apneas and hypopneas present a serious risk for healthy infant development. Treating these adverse events requires frequent manual stimulation by skilled personnel, which can lead to alarm fatigue. This study aims to develop and validate an interpretable model that can predict apneas and hypopneas. Automatically predicting these adverse events before they occur would enable the use of methods for automatic intervention. We propose a neural additive model to predict individual occurrences of neonatal apnea and hypopnea and apply it to a physiological dataset from infants with Robin sequence at risk of upper airway obstruction. The dataset will be made publicly available together with this study. Our proposed model allows the prediction of individual apneas and hypopneas, achieving an average AuROC of 0.80 when discriminating segments of polysomnography recordings starting 15 seconds before the onset of apneas and hypopneas from control segments. Its additive nature makes the model inherently interpretable, which allowed insights into how important a given signal modality is for prediction and which patterns in the signal are discriminative. For our problem of predicting apneas and hypopneas in infants with Robin sequence, prior irregularities in breathing-related modalities as well as decreases in SpO2 levels were especially discriminative. Our prediction model presents a step towards an automatic prediction of neonatal apneas and hypopneas in infants at risk for upper airway obstruction. Together with the publicly released dataset, it has the potential to facilitate the development and application of methods for automatic intervention in clinical practice.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"3 12","pages":"e0000678"},"PeriodicalIF":0.0,"publicationDate":"2024-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11642933/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142822814","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Classification of periodontitis stage and grade using natural language processing techniques.","authors":"Nazila Ameli, Tahereh Firoozi, Monica Gibson, Hollis Lai","doi":"10.1371/journal.pdig.0000692","DOIUrl":"10.1371/journal.pdig.0000692","url":null,"abstract":"<p><p>Periodontitis is a complex and microbiome-related inflammatory condition impacting dental supporting tissues. Emphasizing the potential of Clinical Decision Support Systems (CDSS), this study aims to facilitate early diagnosis of periodontitis by extracting patients' information collected as dental charts and notes. We developed a CDSS to predict the stage and grade of periodontitis using natural language processing (NLP) techniques including bidirectional encoder representation for transformers (BERT). We compared the performance of BERT with that of a baseline feature-engineered model. A secondary data analysis was conducted using 309 anonymized patient periodontal charts and corresponding clinician's notes obtained from the university periodontal clinic. After data preprocessing, we added a classification layer on top of the pre-trained BERT model to classify the clinical notes into their corresponding stage and grades. Then, we fine-tuned the pre-trained BERT model on 70% of our data. The performance of the model was evaluated on 32 unseen new patients' clinical notes. The results were compared with the output of a baseline feature-engineered algorithm coupled with MLP techniques to classify the stage and grade of periodontitis. Our proposed BERT model predicted the patients' stage and grade with 77% and 75% accuracy, respectively. MLP model showed that the accuracy of correct classification of stage and grade of the periodontitis on a set of 32 new unseen data was 59.4% and 62.5%, respectively. The BERT model could predict the periodontitis stage and grade on the same new dataset with higher accuracy (66% and 72%, respectively). The utilization of BERT in this context represents a groundbreaking application in dentistry, particularly in CDSS. Our BERT model outperformed baseline models, even with reduced information, promising efficient review of patient notes. This integration of advanced NLP techniques with CDSS frameworks holds potential for timely interventions, preventing complications and reducing healthcare costs.</p>","PeriodicalId":74465,"journal":{"name":"PLOS digital health","volume":"3 12","pages":"e0000692"},"PeriodicalIF":0.0,"publicationDate":"2024-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11642968/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142822807","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}