Interpretation of deep learning using attributions: application to ophthalmic diagnosis
Bibliographic record
Abstract
Optical coherence tomography (OCT) and retinal fundus images are widely used for detecting retinal pathology. In particular, these images are used by deep learning methods for classification of retinal disease. The main hurdle for widespread deployment of AI-based decision making in healthcare is a lack of interpretability of the cutting-edge deep learning-based methods. Conventionally, decision making by deep learning methods is considered to be a black box. Recently, there is a focus on developing techniques for explaining the decisions taken by deep neural networks, i.e. Explainable AI (XAI) to improve their acceptability for medical applications. In this study, a framework for interpreting the decision making of a deep learning network for retinal OCT image classification is proposed. An Inception-v3 based model was trained to detect choroidal neovascularization (CNV), diabetic macular edema (DME) and drusen from a dataset of over 80,000 OCT images. We visualized and compared various interpretability methods for the three disease classes. The attributions from various approaches are compared and discussed with respect to clinical significance. Results showed a successful attribution of the specific pathological regions of the OCT that are responsible for a given condition in the absence of any pixel-level annotations.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.009 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".