Add MV-AFA seizure detection algorithm#89
Open
yuhangYH wants to merge 6 commits into
Open
Conversation
added 6 commits
June 12, 2026 22:09
Author
|
Hi @jon-dan, thanks for maintaining SzCORE! This PR adds MV-AFA — the Docker image ( |
Member
|
Hello, thank you for your contribution! I cannot pull the image (see CI logs) Can you check:
|
Author
|
Thanks @jon-dan, and sorry about that! Both issues are now fixed:
You can verify: Should be good to re-run on the benchmark now — thank you for your patience! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Algorithm
MV-AFA (Multi-View Adaptive Fusion Attention) — topology-aware EEG seizure detection fusing four views of each 2-second window (temporal Transformer, frequency 2-D CNN, statistical features, TDA persistent homology) via a gated mixture.
v1.3.0 — multi-dataset (expanded TUSZ)
Trained across CHB-MIT + Siena + TUH-Sz (TUSZ, 1500 recordings) with subject-level splits, EEG augmentation and per-dataset balancing. SzCORE scores event-F1 across 5 datasets, so a model that has seen multiple corpora beats a single-dataset specialist; leave-one-dataset-out experiments confirmed zero-shot transfer to an unseen corpus is near chance, so the model is trained on each target corpus. Referential 10-20 inputs are auto-remontaged to 18 bipolar pairs.
Expanding TUSZ training data (400→1500 recordings) improved held-out window AUROC across all datasets (overall 0.732→0.745, TUSZ 0.658→0.675).
Docker image
docker.io/mellow99/mv-afa-szcore:v1.3.0Training datasets
CHB-MIT, Siena, TUH-Sz (TUSZ)
Repository
https://github.com/yuhangYH/mv-afa-szcore
Author
Yuhang Guo — Khalifa University