Red Flag/Blue Flag visualization of a common CNN for text classification. Issue 1 (16th January 2023)

• Record Type: Journal Article
• Authors: Del Gaizo, John; Obeid, Jihad S; Catchpole, Kenneth R; Alekseyenko, Alexander V
• Is Part Of: JAMIA open. Volume 6:Issue 1(2023)
• Publication Date: 2023-01-16
• Subjects: explainable AI -- X-AI -- CNN -- NLP -- text classification -- clinical NLP; Medical informatics -- Periodicals; 610.285
• DOI: 10.1093/jamiaopen/ooac112
• Languages: English
• ISSNs: 2574-2531

Abstract

Abstract: A shallow convolutional neural network (CNN), TextCNN, has become nearly ubiquitous for classification among clinical and medical text. This research presents a novel eXplainable-AI (X-AI) software, Red Flag/Blue Flag (RFBF), designed for binary classification with TextCNN. RFBF visualizes each convolutional filter's discriminative capability. This is a more informative approach than direct assessment of logit contribution, features that overfit to train set nuances on smaller datasets may indiscriminately activate large logits on validation samples from both classes. RFBF enables model diagnosis, term feature verification, and overfit prevention. We present 3 use cases of (1) filter consistency assessment; (2) predictive performance improvement; and (3) estimation of information leakage between train and holdout sets. The use cases derive from experiments on TextCNN for binary prediction of surgical misadventure outcomes from physician-authored operative notes. Due to TextCNN's prevalence, this X-AI can benefit clinical text research, and hence improve patient outcomes. Lay Summary: The past decade has seen dramatic improvement in the performance of artificial intelligence (AI) algorithms. A relatively simple model, TextCNN, has shown persistently strong results for classification among clinical and medical text. However, it can be challenging to understand how TextCNN makes a given prediction for a given body of text. Though eXplainable-AI (X-AI) techniques exist to inspect AI models, we present software (Red Flag/Blue Flag [RFBF]) that employs a novel technique which incorporates the performance of the different model components. This approach can be more robust for model diagnosis on noisy data. As TextCNN is widely used, RFBF can facilitate medical text research, which translates to improved patient outcomes. Furthermore, this software can be extended in future research to support similar models.