Deep Learning-Based Nystagmus Detection for BPPV Diagnosis

Abstract

In this study, we propose a deep learning-based nystagmus detection algorithm using video oculography (VOG) data to diagnose benign paroxysmal positional vertigo (BPPV). Various deep learning architectures were utilized to develop and evaluate nystagmus detection models. Among the four deep learning architectures used in this study, the CNN1D model proposed as a nystagmus detection model demonstrated the best performance, exhibiting a sensitivity of 94.06 ± 0.78%, specificity of 86.39 ± 1.31%, precision of 91.34 ± 0.84%, accuracy of 91.02 ± 0.66%, and an F₁-score of 92.68 ± 0.55%. These results indicate the high accuracy and generalizability of the proposed nystagmus diagnosis algorithm. In conclusion, this study validates the practicality of deep learning in diagnosing BPPV and offers avenues for numerous potential applications of deep learning in the medical diagnostic sector. The findings of this research underscore its importance in enhancing diagnostic accuracy and efficiency in healthcare.

Keywords: benign paroxysmal positional vertigo; convolutional neural network; horizontal nystagmus; nystagmus detection; pupil tracking.

Figure 1

Pupil detection results using a least-squares fitting algorithm. The red circle represents the edge of the pupil, while the blue dot indicates the center of the pupil.

Figure 2

Pupil tracking and eye movement quantification results.

Figure 3

Correcting for missing values (NA) in horizontal pupil movement. (A) Serrated eye movements in patients with nystagmus, (B) eye movement data with missing values, (C) data with linear interpolation, (D) data calibrated to the last detected position.

Figure 4

The deep learning model architecture for nystagmus detection. (A) The details of the first proposed CNN1D model, (B) the details of the second proposed CNN-LSTM1D model.

Figure 5

Comparison of ROC curves among the four nystagmus detection models.