SOFTMAX LOGICAL GATED CONVOLUTIONAL NEURAL NETWORK FOR EFFICIENT AND ACCURATE DETECTION OF MULTIPLE MYELOMA IN MICROSCOPIC IMAGES

Abstract

A malignant tumor of plasma cells and multiple myeloma has complicated the disorders. These multiple myeloma images were taken microscopically. The process is time-consuming and labor-intensive, so it results in some noisy images. Identifying cancer cells is difficult and provides less accuracy due to noise in the images. To overcome the above problem, the Softmax Logical Gated Convolutional Neural Network (SLGCNN) approach was applied.The first step is to filter the multiple myeloma by implementing a Median Image Filter based Edge Detection (MIFED). The second stage divides the multiple myeloma preprocessed images using Screen Cluster Area Segmentation (SCAS) enhancing image clarity by eliminating irrelevant parts. After segmentation, we used Recurrent Feature Elimination (RFE) to achieve the selection of an essential set of features to decrease the number of features making processing faster without a loss of efficiency. Moreover, a SLGCNN is employed for the final classification of the refined features, due to its capacity for logical gating that enables the distinction between cytoplasm and nucleus cells with high accuracy. This methodology solves the problems of the and makes this approach to detecting multiple myeloma computationally efficient and clinically valid. The proposed framework could contribute to early intervention and improve the results of multiple myeloma patients by increasing the reliability of diagnosis. The proposed method archives a high accuracy of 96.8% compared to the other systems.