PREDICTING THE OCCURRENCE OF COLORECTAL LYMPHOMA USING DEEP LEARNING TECHNIQUES

Abstract

The third most prevalent cause of cancer death in the world is Colorectal Lymphoma (CL). Future disease burden predictions advise health planners and raise awareness about the need for action on cancer control. The lymphoma volume is usually estimated using Magnetic Resonance Imaging (MRI), which analyses mutation during medical diagnosis at advanced stages. The precise segmentation of abnormal tissue and its correct 3D display is key to appropriate treatment. Here, there is an intention to build an intelligent diagnostic system based on human MRI research.The most prevalent metastatic location for Rectal Carcinoma (RC) are Lymph Nodes (LNs), and the nodal status is crucial to treating and forecasting choices. The site and several metastatic LNs should be investigated before treatment guidelines comply with the Nippon Electric Company (NEC) Network and the American Joint Committee on Cancer (AJCC) Stage Standards. Identifying and removing metastatic LNs during the intervention is crucial to prevent tumor repetition, especially in lateral lines. Some studies have shown that stronger Lateral Lymph Nodes (LLNs) can be closer to local recurrence and showed that dissection of LLNs might enhance prognosis and reduce local recurrence for patients with poor RC at these locations. In contrast, Lateral Lymph Node Dissection (LLND)is an autonomous procedure with more surgical implications, including surgery and long-term sexual and urinary problems. Therefore, must indicate the correct number and position of metastatic LNs before surgery in the therapy option. Nonetheless, the lesions' topologies, borders, densities, and diameters differ, making this a difficult task. Therefore, the first stage of work deals with identification, segmentation and 3D visualization method, offering medical specialist expertise in an efficient way for the 3D reconstruction of Colorectal Lymphoma using medical image processing in two-dimensional magnetic resonance images. Here, the rectal MR images are preprocessed, which can be done using the Weighted Adaptive Median Filtering and Uplift Laplacian Partial Differential Equation for further enhancement. Followed by preprocessing, the iterative multi-linear component analysis was used for extracting the features. The extracted features can be input for the CNN-based multiscale phase level set segmentation process. In this suggested segmentation, the abnormal resection margin is automatically analyzed and shows that this is consistent with the traditional segmentation algorithm. Finally, a 3D simulation of the lymphoma of the colon is accomplished using the Logical Frustum Model used for medical data rendering. Then in the second stage of research work deals with segmentation and classification challenges for normal and abnormal lymph nodes. Here,Pre-processing is performed by Curvature Based Shearlet Filter with Contrast Limited Savitzky-Golay Histogram Equalization is used. The Semi-Supervised Fuzzy Logic Clustering Algorithm was then used to segment lymph nodes. Once the lymph region is segmented, the Grey Level Co-Occurrence Matrix extracts functions (GLCM). Then the Scale-down Bee Herd Optimization approach is introduced to minimize the number of measures by features, which increases the classifier detection rate. Deep Residual Boltzmann Convolution Neural Network System of classification creates a pattern for the benign lymph nodes and the malignant identification.