基于深度学习的甲状腺癌病理图像分级方法

Automatic grading of pathological images of thyroid cancer based on deep learning

针对日益增长的甲状腺癌早期诊断的需求,基于深度学习方法,在EfficientNet网络的基础上结合CA注意力机制,进行甲状腺癌病理图像自动分级方法研究。实验结果显示,CA-EfficientNet网络模型的精确率达到96.6%,证明了基于CA-EfficientNet网络的甲状腺癌病理图像自动分级算法的先进性,基于该算法实现的自动辅助诊断系统具有实际应用性,可有效降低病理医生工作负担,并降低因疲劳等主观因素造成的人工诊断误诊率。

In response to the increasing demand for early diagnosis of thyroid cancer,a deep learning based method is proposed for the automatic grading of the pathological images of thyroid cancer through EfficientNet combined with CA-Net.The experimental results show that the accuracy of CA-EfficientNet model is up to 96.6%,which proves the algorithm superiority in the automatic grading of the pathological images of thyroid cancer.The automatic auxiliary diagnosis system implemented based on the proposed algorithm is applicable in practice for it can effectively reduce the workload of pathologists and reduce the rate of misdiagnosis caused by subjective factors such as fatigue.