Search

In recent years, artificial intelligence-based computer aided diagnosis (CAD) system for the hepatitis has made great progress. Especially, the complex models such as deep learning achieve better performance than the simple ones due to the nonlinear hypotheses of the real world clinical data. However,complex model as a black box, which ignores why it make a certain decision, causes the model distrust from clinicians. To solve these issues, an explainable artificial intelligence (XAI) framework is proposed in this paper to give the global and local interpretation of auxiliary diagnosis of hepatitis while retaining the good prediction performance. First, a public hepatitis classification benchmark from UCI is used to test the feasibility of the framework. Then, the transparent and black-box machine learning models are both employed to forecast the hepatitis deterioration. The transparent models such as logistic regression (LR), decision tree (DT)and k-nearest neighbor (KNN) are picked. While the black-box model such as the eXtreme Gradient Boosting (XGBoost), support vector machine (SVM), random forests (RF) are selected. Finally, the SHapley Additive exPlanations (SHAP), Local Interpretable Model-agnostic Explanations (LIME) and Partial Dependence Plots (PDP) are utilized to improve the model interpretation of liver disease. The experimental results show that the complex models outperform the simple ones. The developed RF achieves the highest accuracy (91.9%) among all the models. The proposed framework combining the global and local interpretable methods improves the transparency of complex models, and gets insight into the judgments from the complex models, thereby guiding the treatment strategy and improving the prognosis of hepatitis patients. In addition, the proposed framework could also assist the clinical data scientists to design a more appropriate structure of CAD. [ABSTRACT FROM AUTHOR]

Cushing's disease (CD) is a rare disease that occurs in 1.2–1.4 persons per million population per year. Recurrence prediction after transsphenoidal surgery (TSS) is important for determining individual treatment and follow-up strategies. Between 2000 and 2017, 354 CD patients with initial postoperative remission and long-term follow-up data were enrolled from Peking union medical college hospital (PUMCH) to predict recurrence, and PUMCH is one of the largest CD treatment centers in the world. We first investigated the effect of a factorization machine (FM)-based neural network to predict recurrence after TSS for CD. This method could automatically reduce a portion of the cross-feature selection work with acceptable parameters. We conducted a performance comparison of various algorithms on the collected dataset. To address the lack of interpretability of neural network models, we also used the local interpretable model-agnostic explanations approach, which provides an explanation in the form of relevant features of the predicted results by approximating the model behavior of the variables in a local manner. Compared with existing methods, the DeepFM model obtained the highest AUC value (0.869) and the lowest log loss value (0.256). According to the importance of each feature, three top features for the DeepFM model were postoperative morning adrenocorticotropic hormone level, age, and postoperative morning serum cortisol nadir. In the post hoc explanation phase, the above-mentioned importance-leading features made a great contribution to the prediction probability. The results showed that deep learning-based models could better aid neurosurgeons in recurrence prediction after TTS for patients with CD, and could contribute to determining individual treatment strategies. [ABSTRACT FROM AUTHOR]