The Applicability of Artificial Intelligence in Predicting the Depth of Myometrial Invasion on MRI Studies-A Systematic Review

• Published in: Diagnostics (Basel) Vol. 13; no. 15; p. 2592
• Main Authors: Petrila, Octavia, Stefan, Anca-Elena, Gafitanu, Dumitru, Scripcariu, Viorel, Nistor, Ionut
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 01-08-2023; MDPI
• Subjects: Accuracy; Artificial intelligence; Cancer; Deep learning; Endometrial cancer; endometrial cancer AI; endometrial cancer artificial intelligence; endometrial cancer machine learning; endometrial cancer machine learning MRI; endometrial cancer MRI artificial intelligence; Fibroids; Gynecology; Lymphatic system; Machine learning; Magnetic resonance imaging; Medical imaging equipment; Medical prognosis; Oncology, Experimental; Pathology; Prognosis; Radiomics; Software; Systematic Review; Romania; endometrial cancer AI; endometrial cancer artificial intelligence; endometrial cancer machine learning; endometrial cancer MRI artificial intelligence; endometrial cancer machine learning MRI
• ISSN: 2075-4418; 2075-4418
• DOI: 10.3390/diagnostics13152592

(1) Objective: Artificial intelligence (AI) has become an important tool in medicine in diagnosis, prognosis, and treatment evaluation, and its role will increase over time, along with the improvement and validation of AI models. We evaluated the applicability of AI in predicting the depth of myometrial invasion in MRI studies in women with endometrial cancer. (2) Methods: A systematic search was conducted in PubMed, SCOPUS, Embase, and clinicaltrials.gov databases for research papers from inception to May 2023. As keywords, we used: "endometrial cancer artificial intelligence", "endometrial cancer AI", "endometrial cancer MRI artificial intelligence", "endometrial cancer machine learning", and "endometrial cancer machine learning MRI". We excluded studies that did not evaluate myometrial invasion. (3) Results: Of 1651 screened records, eight were eligible. The size of the dataset was between 50 and 530 participants among the studies. We evaluated the models by accuracy scores, area under the curve, and sensitivity/specificity. A quantitative analysis was not appropriate for this study due to the high heterogeneity among studies. (4) Conclusions: High accuracy, sensitivity, and specificity rates were obtained among studies using different AI systems. Overall, the existing studies suggest that they have the potential to improve the accuracy and efficiency of the myometrial invasion evaluation of MRI images in endometrial cancer patients.