Poster — Wed Eve—43: A Maximum Likelihood/ Simulated Annealing‐Based Validation Method for Tumor Segmentation Techniques
Bibliographic record
Abstract
Introduction: The performance of tumor segmentation methods can only be evaluated by comparison with observers' contours when the true pathologic extent is unknown. All observers' contours contain bias and it is unclear how to best combine segmentations to estimate “truth”. Objective: To develop a method that optimally combines observers' contours to estimate a “true” reference for evaluating tumor segmentation techniques. Methods: A probabilistic “truth” was estimated from multiple observers' segmentations via maximum‐likelihood analysis using the simulated‐annealing (SA) and Expectation‐Maximum optimization algorithms. A qualitative ranking of the performance levels of observers' segmentations, was introduced to steer the method when the relative qualities of input contours are known. The SA‐based method was evaluated first using digital phantoms which were “true” tumors and simulated observers' contours by shifting, shrinking and expanding the “true” tumors. It, secondly, was evaluated using clinical data. The tumor volumes of 12 head and neck cancer (HNC) patients were contoured by three radiation oncologists using CT alone and subsequently, using PET/CT. Results: The SA‐based method exactly determined the “truth” in digital phantom studies. In the clinical study of HNC patients the mean and the range of sensitivity of the SA method were 0.90 and 0.70–0.98, respectively, when compared to a pre‐defined “probabilistic” reference. Conclusions: This work suggests that the SA method can accurately estimate a “truth” to validate image segmentation methods. It also provides a logical means of combining segmentations from different sources, which could potentially improve accuracy of radiation therapy or surgical target localization.
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 | 0.006 | 0.010 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.001 | 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) | 0.007 | 0.002 |
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".