IC‐P‐139: Accurate automatic segmentation of white matter hyperintensities using a linear regression classifier
Bibliographic record
Abstract
White matter hyperintensities (WMHs) occur in healthy elderly persons as well as patients with Alzheimer's dementia (AD), vascular dementia, and multiple sclerosis. Manual segmentation of WMHs is expensive, time consuming, and vulnerable to inter-rater and test-retest variability. We propose an automated method for segmentation of WMHs from MRI of elderly subjects. Forty adults aged 55 years or older from the PREVENT-AD study had T1-w, T2*, and FLAIR scans. The union of manual WMH labels from two raters were used as “gold standard” for training and validation. All scans were pre-processed using our standard pipeline: image intensity non-uniformity correction, intensity range normalization (0-100), co-registration of T2* and FLAIR to T1-w, and stereotaxic registration of T1-w. The three modalities were then non-linearly warped to an AD template obtained from the ADNI dataset. 17 features served as inputs to the classifier: three MRI intensities (T1-w, T2*, and FLAIR), spatial probability map, intensity distribution probabilities for healthy and WMH tissues (PWM,PWMH) as well as the ratio PWM/PWMH for each modality, and the average intensity of healthy tissue at each voxel for each modality. WMHs were segmented using a linear regression classifier with thresholding yielding low variance, high accuracy and low computation time. Performance was evaluated with 10-fold cross-validation, using Dice Kappa similarity, sensitivity, and specificity to compare automated and manual labels. The mean Dice Kappa, sensitivity, and specificity between automated and manual labels were 0.46, 0.55, and 0.999, respectively. For further validation, total WMH loads were correlated with Fazekas scores of 30 subjects (r=0.80, p=6.44e-7 for periventricular and r=0.55, p=2.8e-3 for deep WMHs) as well as age for the total population (r=0.43, p=1.52e-6). Figures 1.a., 1.b. and 1.c. show the FLAIR images, and automated and manual labels for a subject. Figures 1.d. and 1.e. show Kappa and sensitivity/specificity versus threshold.
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 | 0.002 | 0.004 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.006 | 0.006 |
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".