149 Impact of AI in diagnostic imaging: establishing its role in contributing to or mitigating harms from overdiagnosis
Bibliographic record
Abstract
Computational science in medicine is no longer a futuristic premise. The medical device industry has engineered Artificial Intelligence (AI) and Machine Learning (ML) tools to derive insights from large volumes of structured and unstructured data and innovate products that have advanced the delivery of healthcare. Also, in the area of diagnostic imaging (DI), the power of AI and ML is soon to be harnessed to its full extent. It is anticipated that its widespread use will replace the intellectual role of a radiologist in places where radiologists are rare, provide greater sensitivity and specificity in cases where physicians have limited experience and sub-optimal ability to pick up rare but significant findings, and potentially reduce physician workload. We know that the use of AI/ML models in medical imaging may not be without error, in that there will likely be false positive findings. However, it is unknown whether these tools would exacerbate the potential risks of making incidental findings that require additional investigation or treatment OR would mitigate them as part of its intelligent design. The aim of our proposal is to highlight the need to have more studies that fully evaluate the downstream effects of utilizing AI-assisted DI modalities, with respect to overdiagnosis. It is anticipated that having this conversation at the Preventing Overdiagnosis conference this year will provide interesting perspectives that are not only timely but are also critical as we launch the AI train in diagnostic radiology. Suggested topics for discussion include: What is the role of AI in preventing overdiagnosis in DI? Can AI be used to better identify incidental findings that are so benign and carry such a good prognosis that they may not even need to be communicated in an imaging report? If the use of AI in diagnostic radiology becomes widespread, what factors within healthcare may arise and contribute to overdiagnosis (outside of the AI system itself)? Given the power of AI to detect incidental and/or rare findings with greater sensitivity and specificity, will there be a shift in the way radiologists interpret findings and make diagnoses? Will there be new diseases created that might potentially lead to increased patient stigmatization and overdiagnosis? Projected outcomes of proposed workshop: Attendees will learn and benefit from perspectives provided by the panel and subject-matter experts on the use of AI in diagnostic imaging. Conclusions derived from the conversation will be collated and written up as a commentary. This will contribute to the ongoing conversation on AI-assisted healthcare and hopefully provide fresh insight on overdiagnosis in that space. The proceedings from the workshop will also be written up to define the challenges, potential solutions and as such a research agenda.
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 | 0.041 | 0.169 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.003 | 0.014 |
| Scholarly communication | 0.011 | 0.010 |
| Open science | 0.003 | 0.004 |
| Research integrity | 0.008 | 0.006 |
| Insufficient payload (model declined to judge) | 0.016 | 0.004 |
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".