Deep transfer learning for characterizing chondrocyte patterns in phase contrast X-Ray computed tomography images of the human patellar cartilage

Abstract

Phase contrast X-ray computed tomography (PCI-CT) has been demonstrated to be effective for visualization of the human cartilage matrix at micrometer resolution, thereby capturing osteoarthritis induced changes to chondrocyte organization. This study aims to systematically assess the efficacy of deep transfer learning methods for classifying between healthy and diseased tissue patterns. We extracted features from two different convolutional neural network architectures, CaffeNet and Inception-v3 for characterizing such patterns. These features were quantitatively evaluated in a classification task measured by the area (AUC) under the Receiver Operating Characteristic (ROC) curve as well as qualitative visualization through a dimension reduction approach t-Distributed Stochastic Neighbor Embedding (t-SNE). The best classification performance, for CaffeNet, was observed when using features from the last convolutional layer and the last fully connected layer (AUCs >0.91). Meanwhile, off-the-shelf features from Inception-v3 produced similar classification performance (AUC >0.95). Visualization of features from these layers further confirmed adequate characterization of chondrocyte patterns for reliably distinguishing between healthy and osteoarthritic tissue classes. Such techniques, can be potentially used for detecting the presence of osteoarthritis related changes in the human patellar cartilage.

Current status of use of big data and artificial intelligence in RMDs: a systematic literature review informing EULAR recommendations

“…The mean number of data points was 746 million (2000–5 billion) in RMDs, and 9.1 billion (range 100 000–200 billion) outside of RMDs. Even if the mean number of units of observation in the SLR and in the mirror review was higher than 1 million, small numbers of units of observation were also observed; however, they corresponded to imaging data, which actually provided a huge number of data points (eg, five CT-scans in RMDs provide more than 26 million data points) 24…”

Deep learning in medical imaging and radiation therapy

“…A relatively small labeled dataset can then be used to train a classifier such as an SVM for the problem at hand. A number of studies extracted the outputs of the fully connected layers of a DL network that has been pretrained ImageNet, and used those features as input to SVMs to build classification models, which suggests that a network pretrained on natural images is useful for extracting features for medical image analysis purposes.…”

Grade Prediction of Bleeding Volume in Cesarean Section of Patients With Pernicious Placenta Previa Based on Deep Learning

“…After removing the background from the uterus region image, according to the uterus region image and the corresponding bleeding volume level data (negative cases or positive cases), a bleeding volume level classification model was developed using VGGNet-16 network with Transfer Learning (Abidin et al, 2018;Zhe et al, 2019). The framework is shown in Figure 5 (Lin and Yuan, 2019).…”