Kidney image segmentation from CT for disease diagnosis based on deep extreme cut and NASNet-Bi-LSTM model using generative AI for improved resolution

Kidney disease technically referred to as nephropathy, which is a broad term used to describe a variety of disorders that affect the structure and function of the kidneys. Even a slight deviation in kidney function and structure measurements are linked to a higher chance of death more frequent than kidney failure. The patient's kidney condition doesn't appear severe in its initial stages, but recovery becomes difficult as the illness advances. To preserve the patient's life, doctors must be able to diagnose the illness early. Several machine learning algorithms are some of the commonly used automated models to predict for diagnosing various diseases. But achieving accurate illness prediction with a low error probability is difficult due to inadequate data training, poor image quality, and incorrect segmentation. So, a hybrid deep learning system is created to detect kidney illness based on CT scans in order to allay these worries. The input images of the kidney stone, cysts, normal and tumor are collected and pre-processed using a modified Gen AI enabled super resolution conversion algorithm to replace the distorted pixels in the input image. Then for enhancing the contrast level of the super resolution image, Dandelion based CLAHE algorithm is developed. At last, hybrid NASNet-BiLSTM is utilized for detecting the kidney disease whether it is normal, stone, cysts and tumor. The suggested method provides 94 % precision, 93 % specificity, and 96 % accuracy. Consequently, by employing this automated approach for detecting the kidney disease diagnosis can be facilitated and treatment can be started early to reduce the death rate.