Interpretable CNN for ischemic stroke subtype classification with active model adaptation

BMC Medical Informatics and Decision Making

Runzhi Li¹  Honghua Dai²  Shuo Zhang³  Jing Wang³  Rui Zhang⁴  Yuming Xu⁴  Shilei Sun⁴  Bo Song⁴  Jun Wu⁴  Yuan Gao⁴  Kai Liu⁴  Hui Fang⁴  Lu Zhao⁴  Lulu Pei⁴ 

[1] Cooperative Innovation Center of Internet Healthcare, Zhengzhou University, Zhengzhou, China;Cooperative Innovation Center of Internet Healthcare, Zhengzhou University, Zhengzhou, China;Institute of Intelligent Systems, Deakin University, Burwood, Australia;School of Information Engineering, Zhengzhou University, Zhengzhou, China;Cooperative Innovation Center of Internet Healthcare, Zhengzhou University, Zhengzhou, China;The Department of Neurology, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China;

关键词: Interpretability;    Ischemic Stroke;    Active learning;    Classification algorithm;    Loss function;   

DOI  :  10.1186/s12911-021-01721-5

来源: Springer

【 摘 要 】

BackgroundTOAST subtype classification is important for diagnosis and research of ischemic stroke. Limited by experience of neurologist and time-consuming manual adjudication, it is a big challenge to finish TOAST classification effectively. We propose a novel active deep learning architecture to classify TOAST.MethodsTo simulate the diagnosis process of neurologists, we drop the valueless features by XGB algorithm and rank the remaining ones. Utilizing active learning framework, we propose a novel causal CNN, in which it combines with a mixed active selection criterion to optimize the uncertainty of samples adaptively. Meanwhile, KL-focal loss derived from the enhancement of Focal loss by KL regularization is introduced to accelerate the iterative fine-tuning of the model.ResultsTo evaluate the proposed method, we construct a dataset which consists of totally 2310 patients. In a series of sequential experiments, we verify the effectiveness of each contribution by different evaluation metrics. Experimental results show that the proposed method achieves competitive results on each evaluation metric. In this task, the improvement of AUC is the most obvious, reaching 77.4.ConclusionsWe construct a backbone causal CNN to simulate the neurologist process of that could enhance the internal interpretability. The research on clinical data also indicates the potential application value of this model in stroke medicine. Future work we would consider various data types and more comprehensive patient types to achieve fully automated subtype classification.

【 授权许可】

CC BY