A Novel Machine Learning Algorithm for Prostate Cancer Image Segmentation using mpMRI

Recently, the advancements in technology and the changes in lifestyle behaviors of people leads to a sedentary routine of everyday habits. For this reason, numerous cancers have been developed and causes death for millions of people every year. Although, cancer is a deadly disease, early detection can help for survival. Especially for prostate cancer (PCa), early detection helps to cure the disease. Several researches have been done in medical image processing using Artificial Intelligence (AI) algorithms, yet accuracy and computational complexities limits the performance. With the intension of introducing a novel model for PCa detection from multi-parametric Magnetic Resonance Imaging (mpMRI), this study introduces an enhanced image segmentation model using the efficiency of Machine Learning (ML) algorithm together with Moth Flame Optimization (MFO) Algorithm to eradicate the previous issues. Generally, segmentation of an image is a partition of the image into multiple regions which enhances the classification performances. The major phases in this research includes 1. Data Pre-processing, 2. Feature Extraction, and finally,3. Segmentation. In data pre-processing, noises in the input images are eliminated using Gaussian filtering. The efficiency of MFO is employed to extract the optimal features from the images, and the extracted images are further subjected for U-Net segmentation. Moreover, the performance of the proposed model is validated through a comparative analysis over state-of the-art models in terms of DSC.

I. Introduction

Medical experts stated that PCa develops once the cells in the prostate start to modify the DNA and alters the normal activities [1]. Moreover, the major risk is cancer cells grow rapidly and survive, meanwhile the normal cells die early. Moreover, in America, PCa becomes the second major reason for cancer deaths. Although cancer is a deadly disease, most men recovered with proper diagnosis and treatment. For this reason, early detection of PCa is necessary to save lives. Most specifically, early detection and appropriate treatment at right time can help to 90% successful cure [2]. However, detecting PCa at an early stage is a challenging one as it needs medical experts or experienced professionals for diagnosis. Artificial Intelligence based deep learning method [3] with Computer-Aided Design (CAD) may help to reduce manual burden to identify cancer [4]. A lot of researches have been performed and proposed to develop effective PCa detection methodologies to solve the issues and to help the PCa affected patients. Previously, several ML algorithms were introduced for PCa detection [5].