Osteoarthritis is a common degenerative joint disease, which is often analyzed by X-ray images. However, if there is a lack of clinical experience when reading the films, it is easy to cause misdiagnosis. Although deep learning has made significant progress in the field of medical image processing, existing models still have limitations in capturing subtle lesion features such as joint spaces. This paper proposes an automatic diagnosis method for osteoarthritis based on the improved shifted windows Transformer (Swin Transformer) and graph convolutional network. By enhancing the modeling of joint space features and cross-layer feature fusion, it is expected to effectively improve the accuracy of early diagnosis of osteoarthritis. Firstly, this paper designs the shifted windows horizontal attention mechanism (SW-HAM), which can enhance the feature extraction ability in the horizontal direction. Secondly, the central-attention graphSAGE (CAG-SAGE) is introduced to conduct weighted aggregation of the feature information of the lesion area through the dynamic attention mechanism. Finally, cross-layer connection technology is utilized to achieve efficient fusion of multi-layer features. The experimental results show that the SW-HAM and CAG-SAGE modules and cross-layer connections significantly improve the model performance. The classification accuracy, recall rate, precision rate, F1 score, and area under the curve are 94.59%, 95.14%, 94.05%, 94.41%, and 96.30% respectively, all of which are superior to the classical network and existing methods. It provides a new and effective method for the classification and diagnosis of osteoarthritis.
Breast cancer is a malignancy caused by the abnormal proliferation of breast epithelial cells, predominantly affecting female patients, and it is commonly diagnosed using histopathological images. Currently, deep learning techniques have made significant breakthroughs in medical image processing, outperforming traditional detection methods in breast cancer pathology classification tasks. This paper first reviewed the advances in applying deep learning to breast pathology images, focusing on three key areas: multi-scale feature extraction, cellular feature analysis, and classification. Next, it summarized the advantages of multimodal data fusion methods for breast pathology images. Finally, the study discussed the challenges and future prospects of deep learning in breast cancer pathology image diagnosis, providing important guidance for advancing the use of deep learning in breast diagnosis.
Breast whole slide image (WSI) serves as an essential basis for breast cancer diagnosis and subtype classification. However, its high resolution, multi-scale structural characteristics, and arbitrary orientation pose substantial challenges for automated analysis. Existing deep learning methods typically rely on large amounts of annotated data and struggle to handle the rotational variability and scale differences inherent in pathological images, which limits their generalization ability in cross-center settings. To address these issues, this study proposes a self-supervised learning model, IAM-BYOL, which incorporated geometric priors into the representation learning process. Building upon the BYOL framework, the model introduced an isotropic attention module (IAM). By employing discrete rotation group convolution, IAM enabled weight sharing across different rotated versions of the convolutional kernels, endowing the encoder with structural rotation equivariance. A subsequent group pooling operation converted the equivariant features into rotation-invariant isotropic representations. In addition, a multi-scale attention mechanism adjusted feature weights adaptively according to responses from different receptive fields, allowing the model to capture informative patterns ranging from nuclear-level details to tissue-level organization. Experimental results demonstrated that IAM-BYOL achieved classification accuracies of 98.74%, 99.04%, 99.01%, and 98.63% on the BreakHis dataset at 40×, 100×, 200×, and 400× magnifications, respectively, while attaining an accuracy of 93.02% on the cross-center private clinical dataset BCD. These findings indicate that introducing geometric inductive biases into pathological image representation learning can effectively enhance model robustness and generalization capability.
ObjectiveTo conduct a Meta-analysis to determine the clinical effect of protocol biopsy (PB)-monitored therapy after renal transplantation.MethodsPubMed, Embase, Cochrane Library, Chinese National Knowledge Infrastructure, Wanfang Standards Database and VIP Database for Chinese Technical Periodicals were searched for trials comparing the efficacy of timely intervention under PB surveillance with the conventional treatment. The quality of included studies was assessed and Meta-analysis was conducted by RevMan 5.3 software.ResultsSix randomized controlled trials met our inclusion criteria, including 698 cases. No significant difference was found between the PB group and the control group in 1-year [relative risk (RR)=0.99, 95% confidence interval (CI) (0.97, 1.01), P=0.39] and 2-year recipient survival rate [RR=1.00, 95%CI (0.97, 1.02), P=0.72]. Graft survival rate after 1 year [RR=1.01, 95%CI (0.99, 1.04), P=0.29] and 2 years [RR=1.02, 95%CI (0.99, 1.06), P=0.19] were also statistically similar. No statistical difference was found in glomerular filtration rate between the two groups [mean difference (MD)=0.45 mL/(min·1.73 m2), 95%CI (–3.77, 4.67) mL/(min·1.73 m2), P=0.83]. Renal function of PB group, monitored by serum creatinine, was superior to the control group [MD=–0.46 mg/dL, 95%CI (–0.63, –0.29) mg/dL, P<0.000 01]. No statistical difference was found in infection between the two groups [RR=1.23, 95%CI (0.69, 2.19), P=0.48].ConclusionsOur study did not suggest PB for every kidney transplantation recipient. However, long-term randomized controlled trials with larger sample size would be necessary to determine whether PB was effective for specific populations.