HmsU-Net: A hybrid multi-scale U-net based on a CNN and transformer for medical image segmentation.

Accurate medical image segmentation is of great significance for subsequent diagnosis and analysis. The acquisition of multi-scale information plays an important role in segmenting regions of interest of different sizes. With the emergence of Transformers, numerous networks adopted hybrid structures incorporating Transformers and CNNs to learn multi-scale information. However, the majority of research has focused on the design and composition of CNN and Transformer structures, neglecting the inconsistencies in feature learning between Transformer and CNN. This oversight has resulted in the hybrid network's performance not being fully realized. In this work, we proposed a novel hybrid multi-scale segmentation network named HmsU-Net, which effectively fused multi-scale features. Specifically, HmsU-Net employed a parallel design incorporating both CNN and Transformer architectures. To address the inconsistency in feature learning between CNN and Transformer within the same stage, we proposed the multi-scale feature fusion module. For feature fusion across different stages, we introduced the cross-attention module. Comprehensive experiments conducted on various datasets demonstrate that our approach surpasses current state-of-the-art methods.

TGMIL: A hybrid multi-instance learning model based on the Transformer and the Graph Attention Network for whole-slide images classification of renal cell carcinoma.

BACKGROUND AND OBJECTIVES: The pathological diagnosis of renal cell carcinoma is crucial for treatment. Currently, the multi-instance learning method is commonly used for whole-slide image classification of renal cell carcinoma, which is mainly based on the assumption of independent identical distribution. But this is inconsistent with the need to consider the correlation between different instances in the diagnosis process. Furthermore, the problem of high resource consumption of pathology images is still urgent to be solved. Therefore, we propose a new multi-instance learning method to solve this problem. METHODS: In this study, we proposed a hybrid multi-instance learning model based on the Transformer and the Graph Attention Network, called TGMIL, to achieve whole-slide image of renal cell carcinoma classification without pixel-level annotation or region of interest extraction. Our approach is divided into three steps. First, we designed a feature pyramid with the multiple low magnifications of whole-slide image named MMFP. It makes the model incorporates richer information, and reduces memory consumption as well as training time compared to the highest magnification. Second, TGMIL amalgamates the Transformer and the Graph Attention's capabilities, adeptly addressing the loss of instance contextual and spatial. Within the Graph Attention network stream, an easy and efficient approach employing max pooling and mean pooling yields the graph adjacency matrix, devoid of extra memory consumption. Finally, the outputs of two streams of TGMIL are aggregated to achieve the classification of renal cell carcinoma. RESULTS: On the TCGA-RCC validation set, a public dataset for renal cell carcinoma, the area under a receiver operating characteristic (ROC) curve (AUC) and accuracy of TGMIL were 0.98±0.0015,0.9191±0.0062, respectively. It showcased remarkable proficiency on the private validation set of renal cell carcinoma pathology images, attaining AUC of 0.9386±0.0162 and ACC of 0.9197±0.0124. Furthermore, on the public breast cancer whole-slide image test dataset, CAMELYON 16, our model showed good classification performance with an accuracy of 0.8792. CONCLUSIONS: TGMIL models the diagnostic process of pathologists and shows good classification performance on multiple datasets. Concurrently, the MMFP module efficiently diminishes resource requirements, offering a novel angle for exploring computational pathology images.