Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2023
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2307.02191 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912322739503104 |
|---|---|
| author | Stutz, David Cemgil, Ali Taylan Roy, Abhijit Guha Matejovicova, Tatiana Barsbey, Melih Strachan, Patricia Schaekermann, Mike Freyberg, Jan Rikhye, Rajeev Freeman, Beverly Matos, Javier Perez Telang, Umesh Webster, Dale R. Liu, Yuan Corrado, Greg S. Matias, Yossi Kohli, Pushmeet Liu, Yun Doucet, Arnaud Karthikesalingam, Alan |
| author_facet | Stutz, David Cemgil, Ali Taylan Roy, Abhijit Guha Matejovicova, Tatiana Barsbey, Melih Strachan, Patricia Schaekermann, Mike Freyberg, Jan Rikhye, Rajeev Freeman, Beverly Matos, Javier Perez Telang, Umesh Webster, Dale R. Liu, Yuan Corrado, Greg S. Matias, Yossi Kohli, Pushmeet Liu, Yun Doucet, Arnaud Karthikesalingam, Alan |
| contents | For safety, medical AI systems undergo thorough evaluations before deployment, validating their predictions against a ground truth which is assumed to be fixed and certain. However, this ground truth is often curated in the form of differential diagnoses. While a single differential diagnosis reflects the uncertainty in one expert assessment, multiple experts introduce another layer of uncertainty through disagreement. Both forms of uncertainty are ignored in standard evaluation which aggregates these differential diagnoses to a single label. In this paper, we show that ignoring uncertainty leads to overly optimistic estimates of model performance, therefore underestimating risk associated with particular diagnostic decisions. To this end, we propose a statistical aggregation approach, where we infer a distribution on probabilities of underlying medical condition candidates themselves, based on observed annotations. This formulation naturally accounts for the potential disagreements between different experts, as well as uncertainty stemming from individual differential diagnoses, capturing the entire ground truth uncertainty. Our approach boils down to generating multiple samples of medical condition probabilities, then evaluating and averaging performance metrics based on these sampled probabilities. In skin condition classification, we find that a large portion of the dataset exhibits significant ground truth uncertainty and standard evaluation severely over-estimates performance without providing uncertainty estimates. In contrast, our framework provides uncertainty estimates on common metrics of interest such as top-k accuracy and average overlap, showing that performance can change multiple percentage points. We conclude that, while assuming a crisp ground truth can be acceptable for many AI applications, a more nuanced evaluation protocol should be utilized in medical diagnosis. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2307_02191 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Evaluating AI systems under uncertain ground truth: a case study in dermatology Stutz, David Cemgil, Ali Taylan Roy, Abhijit Guha Matejovicova, Tatiana Barsbey, Melih Strachan, Patricia Schaekermann, Mike Freyberg, Jan Rikhye, Rajeev Freeman, Beverly Matos, Javier Perez Telang, Umesh Webster, Dale R. Liu, Yuan Corrado, Greg S. Matias, Yossi Kohli, Pushmeet Liu, Yun Doucet, Arnaud Karthikesalingam, Alan Machine Learning Computer Vision and Pattern Recognition Methodology For safety, medical AI systems undergo thorough evaluations before deployment, validating their predictions against a ground truth which is assumed to be fixed and certain. However, this ground truth is often curated in the form of differential diagnoses. While a single differential diagnosis reflects the uncertainty in one expert assessment, multiple experts introduce another layer of uncertainty through disagreement. Both forms of uncertainty are ignored in standard evaluation which aggregates these differential diagnoses to a single label. In this paper, we show that ignoring uncertainty leads to overly optimistic estimates of model performance, therefore underestimating risk associated with particular diagnostic decisions. To this end, we propose a statistical aggregation approach, where we infer a distribution on probabilities of underlying medical condition candidates themselves, based on observed annotations. This formulation naturally accounts for the potential disagreements between different experts, as well as uncertainty stemming from individual differential diagnoses, capturing the entire ground truth uncertainty. Our approach boils down to generating multiple samples of medical condition probabilities, then evaluating and averaging performance metrics based on these sampled probabilities. In skin condition classification, we find that a large portion of the dataset exhibits significant ground truth uncertainty and standard evaluation severely over-estimates performance without providing uncertainty estimates. In contrast, our framework provides uncertainty estimates on common metrics of interest such as top-k accuracy and average overlap, showing that performance can change multiple percentage points. We conclude that, while assuming a crisp ground truth can be acceptable for many AI applications, a more nuanced evaluation protocol should be utilized in medical diagnosis. |
| title | Evaluating AI systems under uncertain ground truth: a case study in dermatology |
| topic | Machine Learning Computer Vision and Pattern Recognition Methodology |
| url | https://arxiv.org/abs/2307.02191 |