Abstract 97: Development And Validation Of A Deep Machine Learning Tool For Automated Intraventricular Hemorrhage Segmentation And Volume Measurement Using 3d Convolutional Neural Networks
Bibliographic record
Abstract
Introduction: Development of new intraventricular hemorrhage (IVH) within 24 hours of presentation or expansion of 1mL or more independently predicts poor outcome. Current methods for IVH volume estimation lack sufficient precision to detect this level of change. The best method of quantification with manual/semi-automated segmentation remain clinically impractical. We aimed to train and validate a convolutional neural network (CNN) to autonomously segment and quantify IVH volume in acute intracerebral hemorrhage (ICH) patients. Methods: Imaging data for training and validation was obtained from 3 international multicenter studies. IVH volumes were assessed using CNN, IVH Score (IVHS), Original Graeb Score (oGS), Modified Graeb Score (mGS), and manual segmentation using ITK-SNAP. Agreement between CNN volume and IVHS estimated volume compared to manual segmentation was assessed using intraclass correlation coefficient (ICC) and Bland-Altman charts. Accuracy of CNN segmentation was compared to manual segmentation with Dice similarity coefficient (DSC). Accuracy of hematoma expansion detection was assessed using receiver operating characteristic curves. Results: A total of 172 patients and 311 CT scans were included. Intra-rater reliability was significantly greater for CNN versus IVHS method (ICC 0.99 [95%CI 0.99 - 1.00] vs 0.76 [95%CI 0.66 - 0.83]). Accuracy of CNN segmentation was satisfactory (DSC 0.76 [95%CI 0.75 - 0.78]). Accuracy of hematoma expansion detection was substantially greater in CNN (AUC 0.91 [95%CI 0.85 - 0.97]) versus IVHS, oGS and mGS. Conclusion: Our results demonstrate that a fully automated CNN algorithm is capable of segmenting IVH volumes on multi-center data with higher intra-rater reliability and satisfactory accuracy over the current methods of IVH volume measurement tools. The algorithm has proven to be capable of detecting hematoma expansion with substantially greater accuracy over existing methods of IVH measurement.
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 | 0.003 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 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".