Comparison of Modern Convolutional Neural Networks for Segmentation of Vascular Defects in Ventilation/Perfusion (V/Q) Scans
Bibliographic record
Abstract
Interpretation of ventilation-perfusion (V/Q) scintigraphy, primarily used to diagnose pulmonary embolism (PE), requires identification of V/Q mismatch defects, which is prone to interobserver variability. To overcome these challenges, we investigated three AI models for automated segmentation of vascular perfusion defects in planar V/Q scans and compared their performance with human annotators. We trained and validated three 2D neural networks models (a nnUnet, SwinUnet, and Bottleneck Transformer Unet [BTUnet]) on 1,313 patients and 329 patients, respectively. These data were previously annotated for V/Q mismatch defects by nuclear medicine physicians. Training was accomplished by minimizing a combined cross-entropy and Dice loss. BT-Unet featured a lightweight transformer within the U-Net bottleneck. Performance was measured on 46 high probability patients according to the original clinical report via Dice score and freeresponse receiver operating characteristic (FROC) curves relative to annotations by 8 readers. Segments whose centroid were within 1.95 cm of the ground truth were scored as true positives. BTUnet, SwinUnet, and Unet mean ± std Dice scores were$0.51 \pm 0.37$,$0.48 \pm 0.39,0.48 \pm 0.40$, respectively. FROC analysis showed large inter-observer variability among readers with BTUnet the only machine observer performing on par with human readers. Machine observers can provide rapid and consistent segmentations consistent with that of experienced human readers, potentially enhancing diagnostic efficiency and supporting non-expert readers for evaluation of$\mathbf{P E}$.
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 | 0.006 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| 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) | 0.001 | 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 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".