Calculation of Myometrial Invasion Depth of Early Endometrial Cancer MRI Images Based on Deep Learning

The application of artificial intelligence (AI) assisted analysis in the diagnosis of endometrial cancer using magnetic resonance images (MRI) can assist doctors in accurately evaluating the depth of myometrial invasion (MI). We studied the MRI images of 68 patients with early endometrial cancer, and successfully developed a sophisticated computerized method for accurately segmenting endometrial cancer. A comparative evaluation was performed on three datasets: a dataset consisting of dynamic contrast-enhanced (DCE) sequences, a combined dataset comprising diffusion weighted imaging (DWI) and T2 weighted imaging (T2WI) sequences, and another combined dataset including DWI, T2WI sequences, and DCE sequence. Among these datasets, superior segmentation performance was observed in the DCE sequence dataset. The improved method based on ellipse fitting algorithm to calculate the MI depth effectively solves the problem of MI depth calculation error caused by endometrial tumor deviating from the center of the uterus.