Review on ultrasonographic diagnosis of thyroid diseases based on deep learning / 生物医学工程学杂志

In recent years, the incidence of thyroid diseases has increased significantly and ultrasound examination is the first choice for the diagnosis of thyroid diseases. At the same time, the level of medical image analysis based on deep learning has been rapidly improved. Ultrasonic image analysis has made a series of milestone breakthroughs, and deep learning algorithms have shown strong performance in the field of medical image segmentation and classification. This article first elaborates on the application of deep learning algorithms in thyroid ultrasound image segmentation, feature extraction, and classification differentiation. Secondly, it summarizes the algorithms for deep learning processing multimodal ultrasound images. Finally, it points out the problems in thyroid ultrasound image diagnosis at the current stage and looks forward to future development directions. This study can promote the application of deep learning in clinical ultrasound image diagnosis of thyroid, and provide reference for doctors to diagnose thyroid disease.

A review on multi-modal human motion representation recognition and its application in orthopedic rehabilitation training / 生物医学工程学杂志

Human motion recognition (HAR) is the technological base of intelligent medical treatment, sports training, video monitoring and many other fields, and it has been widely concerned by all walks of life. This paper summarized the progress and significance of HAR research, which includes two processes: action capture and action classification based on deep learning. Firstly, the paper introduced in detail three mainstream methods of action capture: video-based, depth camera-based and inertial sensor-based. The commonly used action data sets were also listed. Secondly, the realization of HAR based on deep learning was described in two aspects, including automatic feature extraction and multi-modal feature fusion. The realization of training monitoring and simulative training with HAR in orthopedic rehabilitation training was also introduced. Finally, it discussed precise motion capture and multi-modal feature fusion of HAR, as well as the key points and difficulties of HAR application in orthopedic rehabilitation training. This article summarized the above contents to quickly guide researchers to understand the current status of HAR research and its application in orthopedic rehabilitation training.

A novel pixel value space statistics map of the pulmonary nodule for classification in computerized tomography images.

Accurate assessment of pulmonary nodules can help to diagnose the serious degree of lung cancer. In most computed aided diagnosis (CADx) systems, the feature extraction module plays quite an important role in classifying pulmonary nodules based on different attributes of them. To precisely evaluate the malignancy of an unknown pulmonary nodule, this paper first proposes a novel pixel value space statistics map (PVSSM) for pulmonary nodules classification. By means of PVSSM this study can transform an original two-dimensional (2D) or three-dimensional (3D) pulmonary nodule into a 2D feature matrix, which contributes to better classifying a pulmonary nodule. To validate the proposed method, this study assembled 5385 valid 3D nodules from 1006 cases in LIDC-IDRI database. This study extracts sets of features from the created feature matrixes by singular value decomposition (SVD) method. Using several popular classifiers including KNN, random forest and SVM, we acquire the classification accuracies of 77.29%, 80.07% and 84.21%, respectively. Moreover, this study also utilizes the convolutional neural network (CNN) to assess the malignancy of nodules and the sensitivity, specificity and area under the curve (AUC) reach up to 86.0%, 88.5% and 0.913, respectively. Experiments demonstrate that the PVSSM has a benefit for nodules classification.

Content-based retrieval for lung nodule diagnosis using learned distance metric.

Similarity metric of the lung nodules can be useful in differentiating between benign and malignant lung nodule lesions on computed tomography (CT). Unlike previous computerized schemes, which focus on the features extracting, we concentrate on similarity metric of the lung nodules. In this study, we first assemble a lung nodule dataset which is from LIDC-IDRI lung CT images. This dataset includes 746 lung nodules in which 375 domain radiologists identified malignant nodules and 371 domain radiologists-identified benign nodules. Each nodule is represented by a vector of 26 texture features. We then propose a content-based image retrieval (CBIR) scheme to classify between benign and malignant lung nodules with a learned Mahalanobis distance metric. The Mahalanobis distance metric as a similarity metric can preserve semantic relevance and visual similarity of lung nodules. The CBIR approach uses this Mahalanobis distance to search for most similar reference nodules for each queried nodule. The majority of votes are then computed to predict the likelihood of the queried nodule depicting a malignant lesion. For the classification accuracy, the area under the ROC curve (AUC) can achieve as 0.942±0.008. The recall and precision of benign nodules are 0.860 and 0.889, respectively. The recall and precision of malignant nodules are 0.893 and 0.866, respectively.