Guided undersampling classification for automated radiation therapy quality assurance of prostate cancer treatment
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
PURPOSE: To test the use of well-studied and widely used classification methods alongside newly developed data-filtering techniques specifically designed for imbalanced-data classification in order to demonstrate proof of principle for an automated radiation therapy (RT) quality assurance process on prostate cancer treatment. METHODS: A series of acceptable (majority class, n = 61) and erroneous (minority class, n = 12) RT plans as well as a disjoint set of acceptable plans used to develop features (n = 273) were used to develop a dataset for testing. A series of five widely used imbalanced-data classification algorithms were tested with a modularized guided undersampling procedure that includes ensemble-outlier filtering and normalized-cut sampling. RESULTS: Hybrid methods including either ensemble-outlier filtering or both filtering and normalized-cut sampling yielded the strongest performance in identifying unacceptable treatment plans. Specifically, five methods demonstrated superior performance in both area under the receiver operating characteristics curve and false positive rate when the true positive rate is equal to one. Furthermore, ensemble-outlier filtering significantly improved results in all but one hybrid method (p < 0.01). Finally, ensemble-outlier filtering methods identified four minority instances that were considered outliers in over 96% of cross-validation iterations. Such instances may be considered distinct planning errors and merit additional inspection, providing potential areas of improvement for the planning process. CONCLUSIONS: Traditional imbalanced-data classification methods combined with ensemble-outlier filtering and normalized-cut sampling provide a powerful framework for identifying erroneous RT treatment plans. The proposed methodology yielded strong classification performance and identified problematic instances with high accuracy.
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.
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 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| 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 it