Artificial intelligence software for detecting unsuspected lung cancer on chest radiographs in an asymptomatic population
Bibliographic record
Abstract
Abstract Background Detecting clinically unsuspected lung cancer on chest radiographs is challenging. Artificial intelligence (AI) software that performs comparably to radiologists may serve as a useful tool. Purpose To evaluate the lung cancer detection performance of a commercially available AI software and to that of humans in a healthy population. Materials and Methods This retrospective study used chest radiographs from the Prostate, Lung, Colorectal, and Ovarian (PLCO) cancer screening trial in the United States between November 1993 and July 2001 with pathological cancer diagnosis follow-up to 2009 (median 11.3 years). The software's predictions were compared to the PLCO radiologists' reads. A reader study was performed with a subset comparing the software to 3 experienced radiologists. Results The analysis included 24 370 individuals (mean age 62.6±5.4; median age 62; cancer rate 2%), with 213 individuals (mean age 63.6±5.5; median age 63; cancer rate 46%) for the reader study. AI achieved higher specificity (0.910 for AI vs. 0.803 for radiologists, P < .001), positive predictive value (0.054 for AI vs. 0.032 for radiologists, P < .001), but lower sensitivity (0.326 for AI vs. 0.412 for radiologists, P = .001) than the PLCO radiologists. When we calibrated the sensitivity of AI to match it with the PLCO radiologists, AI had higher specificity (0.815 for AI vs. 0.803 for radiologists, P < .001). In the reader study, AI achieved higher sensitivity than readers 1 and 3 (0.608 for AI vs. 0.588 for reader 1, P = .789 vs. 0.588 for reader 3, P = .803) but lower specificity than reader 1 (0.888 for AI vs. 0.905 for reader 1, P = .814). Compared to reader 2, AI showed higher specificity (0.888 for AI vs. 0.819 for reader 2, P = .153) but lower sensitivity (0.888 for AI vs. 0.905 for reader 1, P = .814). Conclusion AI detects lung cancer on chest radiographs among asymptomatic individuals with comparable performance to experienced radiologists.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 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 teacher head, 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".