Motivation: Traditional image attribution methods struggle to satisfactorily explain predictions of neural networks. Prediction explanation is important, especially in medical imaging, for avoiding the unintended consequences of deploying AI systems when false positive predictions can impact patient care. Thus, there is a pressing need to develop improved models for model explainability and introspection. 
Specific problem: A new approach is to transform input images to increase or decrease features which cause the prediction. However, current approaches are difficult to implement as they are monolithic or rely on GANs. These hurdles prevent wide adoption.
Our approach: Given an arbitrary classifier, we propose a simple autoencoder and gradient update (Latent Shift) that can transform the latent representation of a specific input image to exaggerate or curtail the features used for prediction. We use this method to study chest X-ray classifiers and evaluate their performance. We conduct a reader study with two radiologists assessing 240 chest X-ray predictions to identify which ones are false positives (half are) using traditional attribution maps or our proposed method.
Results: We found low overlap with ground truth pathology masks for models with reasonably high accuracy. However, the results from our reader study indicate that these models are generally looking at the correct features.
We also found that the Latent Shift explanation allows a user to have more confidence in true positive predictions compared to traditional approaches (0.15±0.95 in a 5 point scale with p=0.01) with only a small increase in false positive predictions (0.04±1.06 with p=0.57).

Accompanying webpage: https://mlmed.org/gifsplanation/
Source code: https://github.com/mlmed/gifsplanation

Gifsplanation via Latent Shift: A Simple Autoencoder Approach to Counterfactual Generation for Chest X-rays

"Too Long; Didn't Read": a short sentence describing your paper

Abstract of paper. Add TeX formulas using the following formats: $In-line Formula$ or $$Block Formula$$

Search author profile by first, middle and last name or email address. If the profile is not found, you can add the author by completing first, middle, and last names as well as author email address.

Before uploading, please carefully read the author instructions at https://2021.midl.io/author-instructions.html and in the call for papers https://2021.midl.io/call-for-papers.html. Full papers are strictly limited to 8 PAGES with as many extra pages as needed for the references and appendix section. Reviews are SINGLE-BLIND, which means you do not have to anonymize your submission.

What best describes the contribution of your paper?

Title of paper. Add TeX formulas using the following formats: $In-line Formula$ or $$Block Formula$$

Gifsplanation via Latent Shift: A Simple Autoencoder Approach to Counterfactual Generation for Chest X-rays

Joseph Paul Cohen, Rupert Brooks, Sovann En, Evan Zucker, Anuj Pareek, Matthew P. Lungren, Akshay Chaudhari