Automating the Problem of Predicting Cervical Cancer Recurrence using a Conditional Generative Adversarial Network

• DOI: 10.15514/ispras-2024-36(3)-16
• Publication Date: 2024-08-31T13:39:51Z
• Authors: Petr Andreevich P...; Roman Viacheslavo...; Olga Nikolaevna C...
• Keywords: Generative adversarial network

This paper presents an intelligent model based on the Pix2Pix conditional generative adversarial network that automates process of predicting recurrence cervical malignancy in patients who have not yet undergone surgery. The implemented accepts a pelvic MRI image as input data and provides output probability tumor generated for "post-operative" perspective. presented differs from its basic analogue by modifying loss function problem conditions replacing standard generator with convolutional neural U-Net. Since formulated belongs to class medical diagnostic tasks, presence false negatives was reduced zero slightly increasing number positives. In comparative analysis prognostic real postoperative images, it experimentally proven only accurately predicts disease, but also generates almost identical centers foci their relative areas magnetic resonance tomography image. feasibility version confirmed comparing results two models using common quality metrics – precision, recall harmonic mean. modification developed makes possible obtain prediction shortest time, allowing be used real-time mode.