一种基于改进EfficientNet的骨肉瘤病理图像分级方法、系统、存储介质
By introducing a dual-gating module and a transfer learning strategy into EfficientNetV2, the spatial structural information preservation capability of the osteosarcoma pathological image grading model is enhanced, solving the accuracy and generalization problems of osteosarcoma pathological grading in existing technologies, and achieving high-precision and high-generalization grading results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TIANJIN POLYTECHNIC UNIV
- Filing Date
- 2026-06-16
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies struggle to effectively capture the complex pathological morphological features of osteosarcoma, especially under small sample conditions. The loss of spatial structural information in traditional models limits their performance in osteosarcoma pathological grading tasks.
In the EfficientNetV2 network, a dual-gated module (DPG) is introduced, which, combined with a transfer learning strategy, enhances the preservation and utilization of spatial structural information. Feature map processing is performed through depthwise separable convolution and GELU activation function, and the SE module in the middle layer is replaced to improve the model's ability to grade osteosarcoma pathological images.
It achieves high-precision and high-generalization osteosarcoma pathological image grading with an accuracy of 98.67% and an F1 score of 98.68%. It performs well under small sample conditions and achieves an accuracy of 89.85% and a recall of 92.13% on public datasets, which is significantly better than other models.
Smart Images
Figure CN122416162A_ABST