Liver Tumor Grade Detection using CNN Based LSTM Model with Corelated Feature Set from CT Images

Every year, liver cancer claims the lives of more than 1.5 million people around the world. Liver cancer can be detected early with Computed Tomography (CT). This could save millions of lives per year. Additionally, there is a pressing need for a computerized system to consistently, quickly, and accurately identify and evaluate CT scans. Because of their small size, shape, intensity, and low contrast, it is extremely difficult to accurately segment minute tumours from the surrounding liver tissue. Deep learning algorithms are utilized to conduct a study of several ways for the prior detection of liver cancer in abdominal imaging. In a computer-aided liver disease diagnostic and liver surgery planning system, such as a liver transplantation system, the ability to accurately and automatically segment the liver parenchyma is essential. However, it is particularly difficult to delineate the liver on CT pictures. A more difficult task in computer-aided diagnosis, stage classification of liver tumours was addressed in this research. It is now being proposed that a modified computer-aided diagnosis can lessen the radiologists' heavy workloads and second opinions. A Convolution Neural Network (CNN) based long short-term memory (LSTM) with correlated feature set (CNN-LSTM-CFS) model is proposed in this paper in order to develop an automated method for liver tumor stage detection. The proposed approach makes use of correlated features from feature set using labeled feature set. The relevance of features appeared to represent the most important imaging criteria for each class, and feature maps have indeed been consistent with the original image features. The difference between cancerous and non-cancerous lesions must be made in addition to the CT-based lesion-stage description in order to arrive at an accurate diagnosis and treatment plan. It necessitates a high level of knowledge, resources, and previous experience. Colorectal cancer metastases in the liver can be distinguished from benign cysts in abdominal CT scans of the liver using a deep end-to-end learning technique. The proposed model is contrasted with the existing methods and the results indicate that the proposed model performance is superior to traditional methods.