<p>Cysts are the most common structural injury observed in adult kidneys. Rarely, these are the result of hereditary conditions such as von Hippel Lindau disease, autosomal dominant polycystic kidney disease, and tuberous sclerosis; however, most of the time, cysts are isolated or limited in number and have no known cause. The development of cysts, stones, and tumors causes chronic kidney disorders, which frequently impede renal function. At first, kidney disorders don't exhibit any noticeable signs and are asymptomatic. To avoid renal failure and the loss of kidney function, kidney illnesses must be diagnosed at an earlier stage. In this paper, renal cyst segmentation and classification is performed using AMS-PAN and hybrid RFN-PINN approaches. Renal ultrasound image is provided as a source for this proposed approach. These raw pictures entered are blurred and poor calibre. The supplied image is pre-processed to improve image quality and eliminate noise. Pre-processing techniques such as Dualistic Sub-Image Histogram Equalization and Logical-Pool Recurrent Neural Network are used to improve and reduce noise in images. After processing, the images are input into the Pyramid Attention Network, which uses a combination of attention mechanisms and a segmentation technique based on Multi-Scale features (AMS-PAN) to identify the affected areas of the image. Finally, this segmented images are given to Hybrid ReducedFireNet and Physics Informed Neural Network algorithm based classifier to predict the renal cyst. The proposed algorithm achieves 97% accuracy, 96.7% specificity, and 91.8% NPV. Consequently, our suggested model is the most effective way to divide and categorize renal cyst sickness.</p>