Optimizing the Use of Radiologist Seed Points for Improved Multiple Sclerosis Lesion Segmentation
Bibliographic record
Abstract
Many current methods for multiple sclerosis (MS) lesion segmentation require radiologist seed points as input, but do not necessarily allow the expert to work in an intuitive or efficient way. Ironically, most methods also assume that the points are placed optimally. This paper examines how seed points can be processed with intuitive heuristics, which provide improved segmentation accuracy while facilitating quick and natural point placement. Using a large set of MRIs from an MS clinical trial, two radiologists are asked to seed the lesions while unaware that the points would be fed into a classifier, based on Parzen windows, that automatically delineates each marked lesion. To evaluate the impact of the new heuristics, an interactive region-growing method is used to provide ground truth and the Dice coefficient (DC) and Spearman’s rank correlation are used as the primary measures of agreement. A stratified analysis is performed to determine the effect on scans with low-, medium-, and high lesion loads. Compared to the unenhanced classifier, the heuristics dramatically improve the DC (+32.91 pt.) and correlation (+0.50) for the scans with low lesion loads, and also improve the DC (+14.55 pt.) and correlation (+0.15) for the scans with medium lesion loads, while having aminimal effect for the scans with high lesion loads, which are already segmented accurately by Parzen windows.With the heuristics, the DC is close to 80% and the correlation is above 0.9 for all three load categories.
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 | 0.004 | 0.020 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
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".