Medical image segmentation methods are generally designed as fully-supervised to guarantee model performance, which requires a significant amount of expert annotated samples that are high-cost and laborious. Semi-supervised image segmentation can alleviate the problem by utilizing a large number of unlabeled images along with limited labeled images. However, learning a robust representation from numerous unlabeled images remains challenging due to potential noise in pseudo labels and insufficient class separability in feature space, which undermines the performance of current semi-supervised segmentation approaches. To address the issues above, we propose a novel semi-supervised segmentation method named as Rectified Contrastive Pseudo Supervision (RCPS), which combines a rectified pseudo supervision and voxel-level contrastive learning to improve the effectiveness of semi-supervised segmentation. Particularly, we design a novel rectification strategy for the pseudo supervision method based on uncertainty estimation and consistency regularization to reduce the noise influence in pseudo labels. Furthermore, we introduce a bidirectional voxel contrastive loss in the network to ensure intra-class consistency and inter-class contrast in feature space, which increases class separability in the segmentation. The proposed RCPS segmentation method has been validated on two public datasets and an in-house clinical dataset. Experimental results reveal that the proposed method yields better segmentation performance compared with the state-of-the-art methods in semi-supervised medical image segmentation. The source code is available at https://github.com/hsiangyuzhao/RCPS .

Objective: This study aimed to develop a predictive model to detect osteoporosis using radiomic features from lumbar spine computed tomography (CT) images. Methods: A total of 133 patients were included in this retrospective study, 41 men and 92 women, with a mean age of 65.45 ± 9.82 years (range: 31–94 years); 53 had normal bone mineral density, 32 osteopenia, and 48 osteoporosis. For each patient, the L1–L4 vertebrae on the CT images were automatically segmented using SenseCare and defined as regions of interest (ROIs). In total, 1,197 radiomic features were extracted from these ROIs using PyRadiomics. The most significant features were selected using logistic regression and Pearson correlation coefficient matrices. Using these features, we constructed three linear classification models based on the random forest (RF), support vector machine (SVM), and K-nearest neighbor (KNN) algorithms, respectively. The training and test sets were repeatedly selected using fivefold cross-validation. The model performance was evaluated using the area under the receiver operator characteristic curve (AUC) and confusion matrix. Results: The classification model based on RF had the highest performance, with an AUC of 0.994 (95% confidence interval [CI]: 0.979–1.00) for differentiating normal BMD and osteoporosis, 0.866 (95% CI: 0.779–0.954) for osteopenia versus osteoporosis, and 0.940 (95% CI: 0.891–0.989) for normal BMD versus osteopenia. Conclusions: The excellent performance of this radiomic model indicates that lumbar spine CT images can effectively be used to identify osteoporosis and as a tool for opportunistic osteoporosis screening.

Background: To develop a magnetic resonance imaging (MRI)-based radiomics predictive model for the identification of knee osteoarthritis (OA), based on the tibial and femoral subchondral bone, and compare with the trabecular structural parameter-based model. Methods: Eighty-eight consecutive knees were scanned with 3T MRI and scored using MRI osteoarthritis Knee Scores (MOAKS), in which 56 knees were diagnosed to have OA. The modality of sagittal three-dimensional balanced fast-field echo sequence (3D BFFE) was used to image the subchondral bone. Four trabecular structural parameters (bone volume fraction [BV/TV], trabecular thickness [Tb.Th], trabecular separation [Tb.Sp], and trabecular number) and 93 radiomics features were extracted from four regions of the lateral and medial aspects of the femur condyle and tibial plateau. Least absolute shrinkage and selection operator (LASSO) was used for feature selection. Machine learning-based support vector machine models were constructed to identify knee OA. The performance of the models was assessed by area under the curve (AUC) of the receiver operator characteristic (ROC). The correlation between radiomics features and trabecular structural parameters was analyzed using Pearson’s correlation coefficient. Results: Our radiomics-based classification model achieved the AUC score of 0.961 (95% confidence interval [CI], 0.912–1.000) when distinguishing between normal and knee OA, which was higher than that of the trabecular parameter-based model (AUC, 0.873; 95% CI, 0.788–0.957). The first-order, texture, and Laplacian of Gaussian-based radiomics features correlated positively with Tb.Th and BV/TV, but negatively with Tb.Sp (P < 0.05). Conclusions: Our results suggested that our MRI-based radiomics models can be used as biomarkers for the classification of OA and are superior to the conventional structural parameter-based model.

Background: Intracranial aneurysm (IA) is a nodular protrusion of the arterial wall caused by the localized abnormal enlargement of the lumen of a brain artery, which is the primary cause of subarachnoid hemorrhage. Accurate rupture risk prediction can effectively aid treatment planning, but conventional rupture risk estimation based on clinical information is subjective and time-consuming.Methods: We propose a novel classification method based on the CTA images for differentiating aneurysms that are prone to rupture. The main contribution of this study is that the learning-based method proposed in this study leverages deep learning and radiomics features and integrates clinical information for a more accurate prediction of the risk of rupture. Specifically, we first extracted the provided aneurysm regions from the CTA images as 3D patches with the lesions located at their centers. Then, we employed an encoder using a 3D convolutional neural network (CNN) to extract complex latent features automatically. These features were then combined with radiomics features and clinical information. We further applied the LASSO regression method to find optimal features that are highly relevant to the rupture risk information, which is fed into a support vector machine (SVM) for final rupture risk prediction.Results: The experimental results demonstrate that our classification method can achieve accuracy and AUC scores of 89.78% and 89.09%, respectively, outperforming all the alternative methods.Discussion: Our study indicates that the incorporation of CNN and radiomics analysis can improve the prediction performance, and the selected optimal feature set can provide essential biomarkers for the determination of rupture risk, which is also of great clinical importance for individualized treatment planning and patient care of IA.

Resting-state functional magnetic resonance imaging has revealed disrupted brain network connectivity in adults and teenagers with cerebral palsy. However, the specific brain networks implicated in neonatal cases remain poorly understood. In this study, we recruited 14 term-born infants with mild hypoxic ischemic encephalopathy and 14 term-born infants with severe hypoxic ischemic encephalopathy from Changzhou Children’s Hospital, China. Resting-state functional magnetic resonance imaging data showed efficient small-world organization in whole-brain networks in both the mild and severe hypoxic ischemic encephalopathy groups. However, compared with the mild hypoxic ischemic encephalopathy group, the severe hypoxic ischemic encephalopathy group exhibited decreased local efficiency and a low clustering coefficient. The distribution of hub regions in the functional networks had fewer nodes in the severe hypoxic ischemic encephalopathy group compared with the mild hypoxic ischemic encephalopathy group. Moreover, nodal efficiency was reduced in the left rolandic operculum, left supramarginal gyrus, bilateral superior temporal gyrus, and right middle temporal gyrus. These results suggest that the topological structure of the resting state functional network in children with severe hypoxic ischemic encephalopathy is clearly distinct from that in children with mild hypoxic ischemic encephalopathy, and may be associated with impaired language, motion, and cognition. These data indicate that it may be possible to make early predictions regarding brain development in children with severe hypoxic ischemic encephalopathy, enabling early interventions targeting brain function. This study was approved by the Regional Ethics Review Boards of the Changzhou Children’s Hospital (approval No. 2013-001) on January 31, 2013. Informed consent was obtained from the family members of the children. The trial was registered with the Chinese Clinical Trial Registry (registration number: ChiCTR1800016409) and the protocol version is 1.0.