Rxivist logo

Assessing the (Un)Trustworthiness of Saliency Maps for Localizing Abnormalities in Medical Imaging

By Nishanth Arun, Nathan Gaw, Praveer Singh, Ken Chang, Mehak Aggarwal, Bryan Chen, Katharina Hoebel, Sharut Gupta, Jay Patel, Mishka Gidwani, Julius Adebayo, Matthew D Li, Jayashree Kalpathy-Cramer

Posted 30 Jul 2020
medRxiv DOI: 10.1101/2020.07.28.20163899

Saliency maps have become a widely used method to make deep learning models more interpretable by providing post-hoc explanations of classifiers through identification of the most pertinent areas of the input medical image. They are increasingly being used in medical imaging to provide clinically plausible explanations for the decisions the neural network makes. However, the utility and robustness of these visualization maps has not yet been rigorously examined in the context of medical imaging. We posit that trustworthiness in this context requires 1) localization utility, 2) sensitivity to model weight randomization, 3) repeatability, and 4) reproducibility. Using the localization information available in two large public radiology datasets, we quantify the performance of eight commonly used saliency map approaches for the above criteria using area under the precision-recall curves (AUPRC) and structural similarity index (SSIM), comparing their performance to various baseline measures. Using our framework to quantify the trustworthiness of saliency maps, we show that all eight saliency map techniques fail at least one of the criteria and are, in most cases, less trustworthy when compared to the baselines. We suggest that their usage in the high-risk domain of medical imaging warrants additional scrutiny and recommend that detection or segmentation models be used if localization is the desired output of the network.

Download data

  • Downloaded 1,904 times
  • Download rankings, all-time:
    • Site-wide: 11,167
    • In radiology and imaging: 26
  • Year to date:
    • Site-wide: 5,390
  • Since beginning of last month:
    • Site-wide: 13,389

Altmetric data

Downloads over time

Distribution of downloads per paper, site-wide