Background and purpose. This study evaluated a modified specialized convolutional neural network (CNN) to improve the accuracy of medical images. Materials and Methods. We defined computed tomography (CT) images as belonging to one of the following 10 classes: head, neck, chest, abdomen, and pelvis with and without contrast media, with 10,000 images per class. We modified the CNN based on the AlexNet with an input size of 512 × 512. We resized the filter sizes of the convolution layer and max pooling. Using these modified CNNs, various models were created and evaluated. The improved CNN was evaluated to classify the presence or absence of the pancreas in the CT images. We compared the overall accuracy, which was calculated from images not used for training, to that of the ResNet. Results. The overall accuracies of the most improved CNN and ResNet in the 10 classes were 94.8% and 89.3%, respectively. The filter sizes of the improved CNN for the convolution layer were (13, 13), (7, 7), (5, 5), (5, 5), and (5, 5) in order from the first layer, and that of max-pooling was (7, 7). The calculation times of the most improved CNN and ResNet were 56 and 120 min, respectively. Regarding the classification of the pancreas, the overall accuracies of the most improved CNN and ResNet were 75.75% and 58.25%, respectively. The calculation times of the most improved CNN and ResNet were 36 and 55 min, respectively. Conclusion. By optimizing the filter size of the convolution layer and max-pooling of 512 × 512 images, we quickly obtained a highly accurate medical image classification model. This improved CNN can be useful for classifying lesions and anatomies for related diagnostic aid applications.

Purpose: A deep learning technique was used to analyze the triceps surae muscle. The devised interpolation method was used to determine muscle’s volume and verify the usefulness of the method. Materials and Methods: Thirty-eight T1-weighted cross-sectional magnetic resonance images of the triceps of the lower leg were divided into three classes, i.e., gastrocnemius lateralis (GL), gastrocnemius medialis (GM), and soleus (SOL), and the regions of interest (ROIs) were manually defined. The supervised images were classified as per each patient. A total of 1199 images were prepared. Six different datasets separated patient-wise were prepared for K-fold cross-validation. A network model of the DeepLabv3+ was used for training. The images generated by the created model were divided as per each patient and classified into each muscle types. The model performance and the interpolation method were evaluated by calculating the Dice similarity coefficient (DSC) and error rates of the volume of the predicted and interpolated images, respectively. Results: The mean DSCs for the predicted images were >0.81 for GM and SOL and 0.71 for GL. The mean error rates for volume were approximately 11% for GL, SOL, and total error and 23% for GL. DSCs in the interpolated images were >0.8 for all muscles. The mean error rates of volume were <10% for GL, SOL, and total error and 18% for GM. There was no significant difference between the volumes obtained from the supervised images and interpolated images. Conclusions: Using the semantic segmentation of the deep learning technique, the triceps were detected with high accuracy and the interpolation method used in this study to find the volume was useful.