<![CDATA[Acute lymphoblastic leukemia is the most important type of childhood leukemia, and accounts for 25% of childhood cancers. Accurately differentiating normal cell precursors from cancer cells is key to the diagnosis of acute lymphoblastic leukemia (ALL). However, under a microscope, cancer cells are so similar to normal cells that it is difficult to classify them. This article presents a detection of acute lymphoblastic leukemia using Convolutional Neural Network (CNN). The dataset which is obtained from ALL_IDB is 582 color image data which is divided into 482 training image data and 100 testing image data. The image data will be resized to 128x128x3 before being input to the CNN model. The CNN model used in this study is a multi-scale CNN which consists of 3 convolution layers (filter size of 3x3, number of filters for each convolution layer is 32, 64, and 128 respectively, and ReLU activation function), 3 subsampling layers using maxpool with filter size of 2x2 , 1 concatenate layer is used to combine the output of each subsampling layer, 1 fully-connected layer with a softmax activation function and a cross-entropy error function, and finally an output layer with 2 classes, namely normal cells and cancer cells. The CNN model will be trained using the Adam optimizer training algorithm with a training rate of 0.0002 and iterated 20 times. Based on the training results after iterating 20 times, the smallest error value was obtained, namely 0.0001 and the largest accuracy value, namely 100% in the 20th epoch. The CNN model was then tested with 100 testing image data and obtained an accuracy rate of 98% and an error value of 0.0482.]]>
Copyrights © 2024
