Dilated TransUNet + +-Based Segmentation with Multi-Scale Adaptive DenseNet with Bi-LSTM Layer-Aided Prostate Cancer Classification Model

Abstract

In common, prostate cancer is regarded as the type of cancer, which occurs over the small walnut-shaped gland in men termed as the prostate. In addition to that, the prostate is considered as the most generally identified type of cancer among men. Here, the gland has been aided in the production of seminal fluid that has been utilized for transporting and nourishing the sperm. In order to exclude the existence of cancer in the tissues, prostate biopsy techniques are utilized. Moreover, the mortality rate due to this disease may be low in the last few years, but it is regarded as the leading cause of cancer. In this case, the automated intelligent techniques are helpful for aiding the pathologists in minimizing fatigue and enhancing the routing process. Moreover, there are some limitations in the traditional model, and it is tackled with the help of a new prostate cancer segmentation and classification approach. Firstly, images related to prostate cancer are attained from standard resources and offered as input to lesion segmentation. Here, lesion segmentation is performed with the help of Adaptive Dilated TransUNet[Formula: see text] to get the segmented image features. The parameters of Dilated TransUNet[Formula: see text] are tuned by utilizing a hybrid approach named Position-aided Pelican-Sea Lion Optimization (PPSLO). Then, the segmented images are offered as the input to Region-of-Interest (ROI) cropping, and the ROI cropped image is attained as the output. Further, the ROI cropped image is fed as the input to the prostate cancer classification phase. In this phase, prostate cancer is classified using Multiscale Adaptive DenseNet with Bi-directional Long Short Term Memory (MAD-Bi-LSTM) Layer, in which the parameters in the network are tuned by the developed approach PPSLO. Hence, the developed prostate cancer segmentation and classification model helps for securing an enhanced disease classification rate than other experimental observations.