Deep Learning for Histopathological Classification and Prognostication of Gynaecologic Smooth Muscle Tumours

Smooth muscle tumors of the uterus that do not fit the diagnostic criteria of benignity (such as leiomyomas) or malignancy (such as leiomyosarcomas) are called STUMP (smooth muscle tumor of uncertain malignant potential). A potential solution to this problem could be the application of predictive models using artificial intelligence (AI) to aid in the histopathological classification and prognosis of gynecological smooth muscle tumors. Deep learning using convolutional neural networks represents a specific class of machine learning, in which predictive models are trained by considering small groups of pixels in digital images and iteratively identifying salient features. In this study, we aim to develop deep learning models capable of accurately subclassifying and predicting the prognosis of gynecological smooth muscle tumors, based on histopathological features of hematoxylin and eosin (H&E) slides. The aim is to develop a diagnostic and prognostic algorithm to help pathologists better classify and diagnose uterine smooth muscle tumors and predict their clinical course.