Background Targeted therapy and immunotherapy put forward higher demands for accurate lung cancer classification, as well as benign versus malignant disease discrimination. Digital whole slide images (WSIs) witnessed the transition from traditional histopathology to computational approaches, arousing a hype of deep learning methods for histopathological analysis. We aimed at exploring the potential of deep learning models in the identification of lung cancer subtypes and cancer mimics from WSIs. Methods We initially obtained 741 WSIs from the First Affiliated Hospital of Sun Yat-sen University (SYSUFH) for the deep learning model development, optimization, and verification. Additional 318 WSIs from SYSUFH, 212 from Shenzhen People’s Hospital, and 422 from The Cancer Genome Atlas were further collected for multi-centre verification. EfficientNet-B5- and ResNet-50-based deep learning methods were developed and compared using the metrics of recall, precision, F1-score, and areas under the curve (AUCs). A threshold-based tumour-first aggregation approach was proposed and implemented for the label inferencing of WSIs with complex tissue components. Four pathologists of different levels from SYSUFH reviewed all the testing slides blindly, and the diagnosing results were used for quantitative comparisons with the best performing deep learning model. Results We developed the first deep learning-based six-type classifier for histopathological WSI classification of lung adenocarcinoma, lung squamous cell carcinoma, small cell lung carcinoma, pulmonary tuberculosis, organizing pneumonia, and normal lung. The EfficientNet-B5-based model outperformed ResNet-50 and was selected as the backbone in the classifier. Tested on 1067 slides from four cohorts of different medical centres, AUCs of 0.970, 0.918, 0.963, and 0.978 were achieved, respectively. The classifier achieved high consistence to the ground truth and attending pathologists with high intraclass correlation coefficients over 0.873. Conclusions Multi-cohort testing demonstrated our six-type classifier achieved consistent and comparable performance to experienced pathologists and gained advantages over other existing computational methods. The visualization of prediction heatmap improved the model interpretability intuitively. The classifier with the threshold-based tumour-first label inferencing method exhibited excellent accuracy and feasibility in classifying lung cancers and confused nonneoplastic tissues, indicating that deep learning can resolve complex multi-class tissue classification that conforms to real-world histopathological scenarios.

The forecast of tropical cyclone trajectories is crucial for the protection of people and property. Although forecast dynamical models can provide high-precision short-term forecasts, they are computationally demanding, and current statistical forecasting models have much room for improvement given that the database of past hurricanes is constantly growing. Machine learning methods, that can capture non-linearities and complex relations, have only been scarcely tested for this application. We propose a neural network model fusing past trajectory data and reanalysis atmospheric images (wind and pressure 3D fields). We use a moving frame of reference that follows the storm center for the 24h tracking forecast. The network is trained to estimate the longitude and latitude displacement of tropical cyclones and depressions from a large database from both hemispheres (more than 3000 storms since 1979, sampled at a 6 hour frequency). The advantage of the fused network is demonstrated and a comparison with current forecast models shows that deep learning methods could provide a valuable and complementary prediction. Moreover, our method can give a forecast for a new storm in a few seconds, which is an important asset for real-time forecasts compared to traditional forecasts.

Background:Joint line (JL) is a very important factor for total knee arthroplasty (TKA) to restore. The objective of this study was to evaluate the early clinical and kinematic results of TKAs with posterior-stabilized (PS) or cruciate retaining (CR) implants in which the JL was elevated postoperatively.Methods:Data were collected from patients who underwent TKA in our department between April 2011 and April 2014. The patients were divided into two groups based on the prosthesis they received (PS or CR). At 1-year postoperatively, clinical outcomes were evaluated by the American Knee Society (AKS) knee score, AKS function score, and patella score. In vivo kinematic analysis after TKA was performed on all patients and a previously validated three-dimensional to two-dimensional image registration technique was used to obtain the kinematic data. Anteroposterior (AP) translation of the medial and lateral femoral condyles, and axial rotation relative to the tibial plateau, were analyzed. The data were assessed using the Mann–Whitney test.Results:At time of follow-up, there were differences in the AKS knee scores (P = 0.005), AKS function scores (P = 0.025), patella scores (P = 0.015), and postoperative range of motions (P = 0.004) between the PS group and the CR group. In the PS group, the magnitude of AP translation for the medial and lateral condyle was 4.9 ± 3.0 mm and 12.8 ± 3.3 mm, respectively. Axial rotation of the tibial component relative to the femoral component was 12.9 ± 4.5°. In the CR group, the magnitude of AP translation for the medial and lateral condyle was 4.3 ± 3.5 mm and 7.9 ± 4.2 mm, respectively. The axial rotation was 6.7 ± 5.9°. There were statistically different between PS group and CR group in kinematics postoperatively.Conclusion:Our results demonstrate that postoperative JL elevation had more adverse effects on the clinical and kinematic outcomes of CR TKAs than PS TKAs.

No abstract

Coronavirus disease 2019 (COVID-19) pneumonia has erupted worldwide, causing massive population deaths and huge economic losses. In clinic, lung ultrasound (LUS) plays an important role in the auxiliary diagnosis of COVID-19 pneumonia. However, the lack of medical resources leads to the low using efficiency of the LUS, to address this problem, a novel automated LUS scoring system for evaluating COVID-19 pneumonia based on the two-stage cascaded deep learning model was proposed in this paper. 18,330 LUS images collected from 26 COVID-19 pneumonia patients were successfully assigned scores by two experienced doctors according to the designed four-level scoring standard for training the model. At the first stage, we made a secondary selection of these scored images through five ResNet-50 models and five-fold cross validation to obtain the available 12,949 LUS images which were highly relevant to the initial scoring results. At the second stage, three deep learning models including ResNet-50, Vgg-19, and GoogLeNet were formed the cascaded scored model and trained using the new dataset, whose predictive result was obtained by the voting mechanism. In addition, 1000 LUS images collected another 5 COVID-19 pneumonia patients were employed to test the model. Experiments results showed that the automated LUS scoring model was evaluated in terms of accuracy, sensitivity, specificity, and F1-score, being 96.1%, 96.3%, 98.8%, and 96.1%, respectively. They proved the proposed two-stage cascaded deep learning model could automatically score an LUS image, which has great potential for application to the clinics on various occasions.