Web-based Application for Cancerous Object Segmentation in Ultrasound Images Using Active Contour Method

Abstract

Segmentation, or the process of separating clinical objects from surrounding tissue in medical images, is an important step in the Computer-Aided Diagnosis (CAD) system. The CAD system is developed to assist radiologists in diagnosing cancer malignancy, which in this research is found in ultrasound (US) medical imaging. The manual segmentation process, which cannot be accessed remotely, is a limitation of the CAD system because cancer objects are screened frequently, continuously, and at all times. Therefore, this research aims to build a user-friendly web application called COSION (Cancerous Object Segmentation) that provides easy access for radiologists to segment cancer objects in US images by adopting an active contour method called HERBAC (Hybrid Edge & Region-Based Active Contour). The waterfall method was used to develop the web application with Django as the web framework. The successfully built web application is named Cosion. Cosion was tested on 114 radiology breast and thyroid US images. Functional, portability, efficiency, reliability, expert validation, and usability testing concluded that Cosion runs well and is suitable for use with a functionality value of 0.9375, an average GTmetrix score of 96.43±0.66%, 100% stress testing percentage, 77.5% expert validation, and 75.8% usability. These quantitative performances indicate that the COSION web application is suitable for implementation in the CAD system for US medical imaging.