EP145/#106 Application of machine learning in endometrial cancer: a systematic review. (4th December 2022)

• Record Type: Journal Article
• Authors: Piedimonte, Sabrina; Rosa, Gabriela; Gerstl, Brigitte; Coronel, Ana; Sopocado, Mars; Vicus, Danielle; Llenno, Salvador
• Is Part Of: International journal of gynecological cancer. Volume 32(2022)Supplement 3
• Page Start: A106
• Page End: A106
• Publication Date: 2022-12-04
• Subjects: Generative organs, Female -- Cancer -- Periodicals; 616.99465
• DOI: 10.1136/ijgc-2022-igcs.236
• Languages: English
• ISSNs: 1048-891X

Abstract

Abstract : Objectives: To review the literature on the application of machine learning(ML) in endometrial cancer(EC) and report the most commonly used algorithms and their performance Methods: This is a systematic review of the literature from January 1985-March 2021 on the use of ML in EC. Four independent reviewers screened the articles initially by title then full text. Quality was assessed using the MINORS criteria. P-values were derived using the Pearson's Chi-squared(x2) test. Results: Among 4, 295 articles, 30 studies were included. The mean age of EC patients was 61.3 years(SD:6.6). The most frequent applications of ML were in: patient datasets(n=10), preoperative diagnostics(n=9), genomics(n=7), and serum biomarkers(n=4). The most commonly used ML models were Neural Networks(n=10) and Support Vector Machines(n=6). Over the past two decades, the number of publications on ML in EC increased from 1(2010) to 29(2021). Only 8/30 studies compared ML techniques to traditional statistics. Among 10 clinical database studies, two ML models(20%) performed better than LR(accuracy: 0.85 vs. 0.82, p=0.16), although not significant. In pre-operative diagnostic studies, ML algorithms tended to improve the detection of EC on MRI images(accuracy: 0.87 vs. 0.82, p=0.24) compared to traditional statistics. In onne serum biomarker study, ML outperformed LR in predicting extrauterine disease(accuracy: 0.81 vs. 0.61). For survival outcomes, one study reported no difference in concordance index scores between Ensemble Algorithm for Clustering(EACCD) and traditional statistics(EACCD with KM: 83.8% vs. EACCD: 83.1%). Conclusions: ML algorithms generally performed similarly to traditional regression models. More studies and larger datasets are needed to assess it's future role in endometrial cancer.