Construction and value of a vestibular function calibration test recognition model based on dual-stream ViT and ConvNeXt architecture

ObjectiveTo improve the efficiency and accuracy of videonystagmography calibration test results while enabling effective recognition of saccadic undershoot waveform by developing a dual-stream architecture-based deep learning model. MethodsA vestibular function calibration test recognition model wit...

Full description

Saved in:
Bibliographic Details
Main Authors: Xu LUO, Peixia WU, Weiming HAO, Yinhong QU, Han CHEN
Format: Article
Language:English
Published: Shanghai Chinese Clinical Medicine Press Co., Ltd. 2025-04-01
Series:Zhongguo Linchuang Yixue
Subjects:
Online Access:https://www.c-jcm.com/article/doi/10.12025/j.issn.1008-6358.2025.20250219
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850176455494860800
author Xu LUO
Peixia WU
Weiming HAO
Yinhong QU
Han CHEN
author_facet Xu LUO
Peixia WU
Weiming HAO
Yinhong QU
Han CHEN
author_sort Xu LUO
collection DOAJ
description ObjectiveTo improve the efficiency and accuracy of videonystagmography calibration test results while enabling effective recognition of saccadic undershoot waveform by developing a dual-stream architecture-based deep learning model. MethodsA vestibular function calibration test recognition model with cross-modal feature fusion was constructed by integrating vision transformer (ViT) and a modified ConvNeXt convolutional network. The model utilized trajectory pictures and spatial distribution maps as inputs, employed a multi-task learning framework to classify calibration data, and to directly evaluate undershoot waveform. ResultsThe model showed outstanding performance in assessing calibration compliance. The accuracy, sensitivity, specificity of the model in left side, middle, and right side were all greater than 90%, and AUC values were all greater than 0.99, with 97.66% of optimal accuracy (middle), 98.98% of optimal sensitivity (middle), 96.87% of optimal specificity (right side), and 0.997 of AUC (right side). The model also showed promising performance in undershoot waveform recognition with 87.50% of accuracy, 89.66% of sensitivity, 85.71% of specificity, 86.67% of F1 score, and 0.931 of AUC. ConclusionsThe proposed method not only significantly enhances the efficiency and accuracy of calibration test results, but also provides a novel solution for undershoot waveform recognition.
format Article
id doaj-art-52cffd07695549d29f47535cf36b5a80
institution OA Journals
issn 1008-6358
language English
publishDate 2025-04-01
publisher Shanghai Chinese Clinical Medicine Press Co., Ltd.
record_format Article
series Zhongguo Linchuang Yixue
spelling doaj-art-52cffd07695549d29f47535cf36b5a802025-08-20T02:19:15ZengShanghai Chinese Clinical Medicine Press Co., Ltd.Zhongguo Linchuang Yixue1008-63582025-04-0132220721110.12025/j.issn.1008-6358.2025.2025021920250219Construction and value of a vestibular function calibration test recognition model based on dual-stream ViT and ConvNeXt architectureXu LUO0Peixia WU1Weiming HAO2Yinhong QU3Han CHEN4Shanghai ZEHNIT Medical Technology Co., Ltd., Shanghai 201318, ChinaVertigo and Balance Function Disorders Clinical Center, EYE & ENT Hospital of Fudan University, Shanghai 200031, ChinaVertigo and Balance Function Disorders Clinical Center, EYE & ENT Hospital of Fudan University, Shanghai 200031, ChinaShanghai ZEHNIT Medical Technology Co., Ltd., Shanghai 201318, ChinaShanghai ZEHNIT Medical Technology Co., Ltd., Shanghai 201318, ChinaObjectiveTo improve the efficiency and accuracy of videonystagmography calibration test results while enabling effective recognition of saccadic undershoot waveform by developing a dual-stream architecture-based deep learning model. MethodsA vestibular function calibration test recognition model with cross-modal feature fusion was constructed by integrating vision transformer (ViT) and a modified ConvNeXt convolutional network. The model utilized trajectory pictures and spatial distribution maps as inputs, employed a multi-task learning framework to classify calibration data, and to directly evaluate undershoot waveform. ResultsThe model showed outstanding performance in assessing calibration compliance. The accuracy, sensitivity, specificity of the model in left side, middle, and right side were all greater than 90%, and AUC values were all greater than 0.99, with 97.66% of optimal accuracy (middle), 98.98% of optimal sensitivity (middle), 96.87% of optimal specificity (right side), and 0.997 of AUC (right side). The model also showed promising performance in undershoot waveform recognition with 87.50% of accuracy, 89.66% of sensitivity, 85.71% of specificity, 86.67% of F1 score, and 0.931 of AUC. ConclusionsThe proposed method not only significantly enhances the efficiency and accuracy of calibration test results, but also provides a novel solution for undershoot waveform recognition.https://www.c-jcm.com/article/doi/10.12025/j.issn.1008-6358.2025.20250219vestibular functioncalibration testvideonystagmographydeep learning model
spellingShingle Xu LUO
Peixia WU
Weiming HAO
Yinhong QU
Han CHEN
Construction and value of a vestibular function calibration test recognition model based on dual-stream ViT and ConvNeXt architecture
Zhongguo Linchuang Yixue
vestibular function
calibration test
videonystagmography
deep learning model
title Construction and value of a vestibular function calibration test recognition model based on dual-stream ViT and ConvNeXt architecture
title_full Construction and value of a vestibular function calibration test recognition model based on dual-stream ViT and ConvNeXt architecture
title_fullStr Construction and value of a vestibular function calibration test recognition model based on dual-stream ViT and ConvNeXt architecture
title_full_unstemmed Construction and value of a vestibular function calibration test recognition model based on dual-stream ViT and ConvNeXt architecture
title_short Construction and value of a vestibular function calibration test recognition model based on dual-stream ViT and ConvNeXt architecture
title_sort construction and value of a vestibular function calibration test recognition model based on dual stream vit and convnext architecture
topic vestibular function
calibration test
videonystagmography
deep learning model
url https://www.c-jcm.com/article/doi/10.12025/j.issn.1008-6358.2025.20250219
work_keys_str_mv AT xuluo constructionandvalueofavestibularfunctioncalibrationtestrecognitionmodelbasedondualstreamvitandconvnextarchitecture
AT peixiawu constructionandvalueofavestibularfunctioncalibrationtestrecognitionmodelbasedondualstreamvitandconvnextarchitecture
AT weiminghao constructionandvalueofavestibularfunctioncalibrationtestrecognitionmodelbasedondualstreamvitandconvnextarchitecture
AT yinhongqu constructionandvalueofavestibularfunctioncalibrationtestrecognitionmodelbasedondualstreamvitandconvnextarchitecture
AT hanchen constructionandvalueofavestibularfunctioncalibrationtestrecognitionmodelbasedondualstreamvitandconvnextarchitecture