A semiautomated radiomics model based on multimodal dual-layer spectral CT for preoperative discrimination of the invasiveness of pulmonary ground-glass nodules.

Background: In recent years, spectral computed tomography (CT) has shown excellent performance in the diagnosis of ground-glass nodules (GGNs) invasiveness; however, no research has combined spectral multimodal data and radiomics analysis for comprehensive analysis and exploration. Therefore, this study goes a step further on the basis of the previous research: to investigate the value of dual-layer spectral CT-based multimodal radiomics in accessing the invasiveness of lung adenocarcinoma manifesting as GGNs. Methods: In this study, 125 GGNs with pathologically confirmed preinvasive adenocarcinoma (PIA) and lung adenocarcinoma were divided into a training set (n=87) and a test set (n=38). Each lesion was automatically detected and segmented by the pre-trained neural networks, and 63 multimodal radiomic features were extracted. The least absolute shrinkage and selection operator (LASSO) was used to select target features, and a rad-score was constructed in the training set. Logistic regression analysis was conducted to establish a joint model which combined age, gender, and the rad-score. The diagnostic performance of the two models was compared by the receiver operating characteristic (ROC) curve and precision-recall curve. The difference between the two models was compared by the ROC analysis. The test set was used to evaluate the predictive performance and calibrate the model. Results: Five radiomic features were selected. In the training and test sets, the area under the curve (AUC) of the radiomics model was 0.896 (95% CI: 0.830-0.962) and 0.881 (95% CI: 0.777-0.985) respectively, and the AUC of the joint model was 0.932 (95% CI: 0.882-0.982) and 0.887 (95% CI: 0.786-0.988) respectively. There was no significant difference in AUC between the radiomics model and joint model in the training and test sets (0.896 vs. 0.932, P=0.088; 0.881 vs. 0.887, P=0.480). Conclusions: Multimodal radiomics based on dual-layer spectral CT showed good predictive performance in differentiating the invasiveness of GGNs, which could assist in the decision of clinical treatment strategies.

Enhancing Multi-disease Diagnosis of Chest X-rays with Advanced Deep-learning Networks in Real-world Data.

The current artificial intelligence (AI) models are still insufficient in multi-disease diagnosis for real-world data, which always present a long-tail distribution. To tackle this issue, a long-tail public dataset, "ChestX-ray14," which involved fourteen (14) disease labels, was randomly divided into the train, validation, and test sets with ratios of 0.7, 0.1, and 0.2. Two pretrained state-of-the-art networks, EfficientNet-b5 and CoAtNet-0-rw, were chosen as the backbones. After the fully-connected layer, a final layer of 14 sigmoid activation units was added to output each disease's diagnosis. To achieve better adaptive learning, a novel loss (Lours) was designed, which coalesced reweighting and tail sample focus. For comparison, a pretrained ResNet50 network with weighted binary cross-entropy loss (LWBCE) was used as a baseline, which showed the best performance in a previous study. The overall and individual areas under the receiver operating curve (AUROC) for each disease label were evaluated and compared among different models. Group-score-weighted class activation mapping (Group-CAM) is applied for visual interpretations. As a result, the pretrained CoAtNet-0-rw + Lours showed the best overall AUROC of 0.842, significantly higher than ResNet50 + LWBCE (AUROC: 0.811, p = 0.037). Group-CAM presented that the model could pay the proper attention to lesions for most disease labels (e.g., atelectasis, edema, effusion) but wrong attention for the other labels, such as pneumothorax; meanwhile, mislabeling of the dataset was found. Overall, this study presented an advanced AI diagnostic model achieving a significant improvement in the multi-disease diagnosis of chest X-rays, particularly in real-world data with challenging long-tail distributions.

An Improved HMM-Based Approach for Planning Individual Routes Using Crowd Sourcing Spatiotemporal Data.

With the rapid development of LBSs (location-based services) in recent years, researchers have increasingly taken an interest in trying to make travel routes more practicable and individualized. Despite the fact that many studies have been conducted on routes using LBS data, the specific routes are deficient in dynamic scalability and the correlations between environmental constraints and personal choices have not been investigated. This paper proposes an improved HMM-based (hidden Markov model) method for planning personalized routes with crowd sourcing spatiotemporal data. It tries to integrate the dynamic public preferences, the individual interests and the physical road network space in the same spatiotemporal framework, ensuring that reasonable routes will be generated. A novel dual-layer mapping structure has been proposed to bridge the gap from brief individual preferences to specific entries of POIs (points-of-interest) inside realistic road networks. A case study on Changsha city has proven that the proposed method can not only flexibly plan people's travel routes under different spatiotemporal backgrounds but also is close to people's natural selection by the perception of the group.