Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-2076
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-21T06:54:05.142985Z
status: published
lens: trace
sector: health
headline: Wavelet Scalogram-Based Nystagmus Classification for Automated Benign Paroxysmal Positional Vertigo Subtype Diagnosis
dek: Benign paroxysmal positional vertigo (BPPV) is the most common cause of vestibular vertigo, yet accurate diagnosis remains challenging without specialist expertise. This study aimed to develop and evaluate an artificial intelligence (AI)-based diagnostic algorithm for BPPV using video-nystagmography (VNG) recordings. A total of 1352 VNG recordings were collected from patients with posterior semicircular canal (PSCC) BPPV (n = 234), lateral semicircular canal (LSCC) BPPV (canalolithiasis, n = 188; cupulolithiasis…
gain_title: An AI model converting eye-tracking signals into wavelet scalograms reached recording-level accuracy of about 0.94 for posterior and lateral canal BPPV, suggesting automated diagnosis without specialist expertise is feasible.
problem_title: For lateral canal BPPV, subtype F1-scores ranged from 0.59 to 0.98, with cupulolithiasis variants classified less well than canalolithiasis variants.
trace_subject: AI classification of BPPV subtypes from video-nystagmography recordings
gain_reading: An AI model converting eye-tracking signals into wavelet scalograms reached recording-level accuracy of about 0.94 for posterior and lateral canal BPPV, suggesting automated diagnosis without specialist expertise is feasible.
gain_evidence: recording-level accuracy was 0.935 for the PSCC and 0.938 for the LSCC evaluation
problem_reading: For lateral canal BPPV, subtype F1-scores ranged from 0.59 to 0.98, with cupulolithiasis variants classified less well than canalolithiasis variants.
problem_evidence: ranged from 0.59 to 0.98 for LSCC classes, with canalolithiasis variants outperforming cupulolithiasis variants
quick_read: Researchers built a diagnostic algorithm for benign paroxysmal positional vertigo using video-nystagmography. Eye trajectories were extracted by a deep learning tracker, turned into two-dimensional wavelet scalogram images, and classified by neural network backbones chosen through systematic comparison.

The work matters because diagnosing the most common vestibular cause of vertigo is described as difficult without specialist expertise, so automated support could widen access to subtype diagnosis. The evidence is still a feasibility finding, and the weaker results for cupulolithiasis show the tool is not uniformly reliable across subtypes.
limitation: All data came from a single tertiary referral hospital, and the authors describe the results only as suggesting feasibility. Lateral canal subtype performance was uneven, especially for cupulolithiasis.
tag: Dual reading
key_points: The dataset contained 1352 recordings from patients with posterior canal BPPV, lateral canal BPPV, and healthy controls at one tertiary referral hospital. | Two otolaryngologists annotated 1312 nystagmus segments, alongside 426 normal segments. | Wavelet-based 2D representations significantly outperformed raw 1D time-series baselines in ablation analysis. | AUROC was 0.974 for horizontal and 0.961 for torsional nystagmus classification. | Posterior canal classes all had F1-scores above 0.84.
rundown: The cohort included 234 posterior canal BPPV patients, 188 lateral canal canalolithiasis, 58 cupulolithiasis, and 224 healthy controls. Recordings yielded 1738 signal segments, and models were evaluated under a patient-level data split.

Performance was reported at two levels: nystagmus classification by direction type, and recording-level BPPV subtype diagnosis. The nystagmus classifiers scored AUROC values of 0.974 horizontal and 0.961 torsional, while subtype accuracy was 0.935 for posterior canal and 0.938 for lateral canal evaluation.

Within lateral canal disease, results varied widely by class, with F1-scores between 0.59 and 0.98. Canalolithiasis variants outperformed cupulolithiasis variants, the smaller group in the dataset.
sources:
- peer_reviewed | Journal of Imaging Informatics in Medicine | https://doi.org/10.1007/s10278-026-02417-w | 2026-10-19
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

5bdfdcb3ba314fcc79db6ba6fc65307d127fb9bdaafc64ee4c6885cd7808e31f

previous

0000000000000000000000000000000000000000000000000000000000000000