Multimodal Analysis of Structural Voice Disorders Based on Speech and Stroboscopic Laryngoscope Video

Multimodal Analysis of Structural Voice Disorders Based on Speech and Stroboscopic Laryngoscope Video

This study intends to collect clinical data such as strobary laryngoscope images and vowel audio data of patients with structural voice disorders and healthy individuals, and to establish a multimodal voice disorder diagnosis system model by using deep learning algorithms. Multi-classification of diseases that cause voice disorders can be applied to patients with voice disorders but undiagnosed in clinical practice, thereby assisting clinicians in diagnosing diseases and reducing misdiagnosis and missed diagnosis. In addition, some patients with voice disorders can be managed remotely through the audio diagnosis model, and better follow-up and treatment suggestions can be given to them. Remote voice therapy can alleviate the current situation of the shortage of speech therapists in remote areas of our country, and increase the number of patients who need voice therapy. opportunity. Remote voice therapy is more cost-effective, more flexible in time, and more cost-effective.

1. Detection and Classification of Acoustic Lesions Based on Speech Deep Learning
2. Detection and Classification of Acoustic Lesions Based on Deep Learning of Images
3. Detection and Classification of Acoustic Lesions Based on Deep Learning Based on Multimodality

Inclusion Criteria:

Laryngeal cancer, laryngeal precancerous lesions, benign laryngeal lesions with voice disorders, healthy people without throat diseases

Exclusion Criteria:

1. A history of laryngeal surgery
2. Patients with voice disorders caused by various causes except laryngeal cancer, laryngeal precancerous lesions, and benign laryngeal lesions
3. The audio quality is not clear, the stroboscopic laryngoscope does not clearly display the anatomical area related to the glottis, and it is underexposed and blocked;