AERMG: An effective attention embedded recurrent mobilenet with GRU model and adaptive segmentation mechanism for diagnosing the prostate cancer lesions

Prostate cancer is one of the most commonly diagnosed cancers in males. The prostate biopsy is the experiment that excludes or ensures the existence of tumor present in the tissue. The samples drawn out from the biopsies are executed and digitalized, attaining whole slide images that are estimated by the medical experts. The deep learning techniques support the conventional clinical decision support device by securing the time of the medical experts in localizing the Region of Interest (ROI) and support in offering better care to patients. Therefore, the research study implements a new deep learning-assisted prostate cancer diagnosis framework to detect prostate cancer tissues at an early stage for reducing misdiagnosis and mortality rates. The normal and malignant prostate images are gathered from online databases. The collected images are applied to the Adaptive Vision Transformers (ViT)-aided Mobile Unet3+ (AViT-MUnet3 + ) model to segment the cancer tissues. It effectively segments the prostate region and background from collected images. Here, the Fitness Condition of Lyrebird Optimization (FCLO) is utilized to optimize the parameters from AViT-MUnet3 + to precise the segmentation performance. Then, the segmented images are further fed into diagnosis phase. Moreover, the Attention Embedded Recurrent Mobilenet with Gated Recurrent Unit (GRU) (AERMG) model is employed to identify the prostate cancer cells effectively. This model effectively detects cancer cells and provides higher accuracy than manual diagnosis and other detection methods. The performance of the prostate cancer diagnosis system is contrasted over multiple existing frameworks with several performance factors.