For fair comparison, different machine learning methods were trained with the same training sets and tested with the same testing sets.For each cross-validation, we calculated the performance values for the five machine learning methods and the doctors. The evaluated methods included both classical feature-based methods and the state-of-the-art deep learning approach. Doctors tended to underestimate the malignant tumors because most of the lymph nodes in this study were small in size. And in 1996 we opened an IC plate workshop and a mould workshop. As all departments work perfectly, we have developed the secondary industry.
For the future, we are planning to collect multi-center data to conduct more generalize evaluation, as well as to explore the potential of deep learning with more training data. The patches were cropped around the lymph node center and resampled into 51 × 51 pixels of 1.0-mm size. From the segmented slices, 3D volumes of lymph nodes were reconstructed. For the classical methods, the texture features were compared with the features used by human doctors for clinical diagnosis, such as tumor size, CT value, SUV, image contrast, and intensity standard deviation.
This study used the well-known AlexNet  architecture implemented using the Keras library for Python. To avoid overfitting to our data, the number of AlexNet layers was reduced to five. For SVM, the AUC and ACC were 0.89 and 80.6%, respectively. Texture analysis on 18F-FDG PET/CT images to differentiate malignant and benign bone and soft-tissue lesions. Med Image Anal. 2014;18:591–604.View ArticlePubMedGoogle ScholarCheng J, Ni D, Chou Y, Qin J, Tiu C, Chang Y, et al. From the 168 patients, 1397 lymph nodes were confirmed cancerous by pathology, and the number of negative and positive samples were 1270 and 127, respectively. Comparison between different feature sets and different methods was mainly performed based on the AUC and ACC values.