A REVIEW ON UTERINE FIBROID DETECTION, PREDICTION AND PREVENTION USING DATA SCIENCE TECHNIQUES

Abstract

Uterine fibroids are the most common benign tumors in women at reproductive age, the Uterine fibroids which nevertheless remain a clinically challenging diagnosis to detect, predict and treat. This review combines the use of machine learning, deep learning, radiomics and hybrid models to the diagnostics of fibroid. Methods are subdivided into classical classifiers like SVM and decision trees or advanced methods like YOLOv7, DenseUNet and 3D DA-VNet. The combination of imaging (MRI, ultrasound) and patient reported information and omics-based features has permitted better segmentation, classification and prediction of outcome. An accuracy of over 90% has always been obtained with the appearance of models with accuracy greater than 0.87, as far as the Dice accuracy metric is concerned and a result value stronger than 0.90 as far as the AUC accuracy metric is concerned. The multimodal fusion and protection-oriented modeling are still problematic despite the improvements. This review points out the salient gaps and future directions like explainable AI, integration of health through mobile devices and real-time support in surgery. The results aim at educating the proper formulation of proactive, individualized fibroid treatment.