Development of AI-Based Laryngeal Cancer Diagnostic Platform Using Laryngoscope Images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Laryngeal cancer diagnosis relies on visual inspection of laryngoscope images.
• Automated detection systems can improve diagnostic accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate AI models for laryngeal cancer detection using laryngoscope images.
• To assess the performance of deep learning models in identifying and localizing laryngeal cancer.

Main Methods:

• Two FCN-ResNet101 deep learning models were designed for vocal cord selection and laryngeal cancer localization.
• Datasets were annotated by otolaryngologists and preprocessed using cropping, normalization, and augmentation techniques.
• Performance metrics included Intersection over Union (IoU), Dice score, accuracy, precision, recall, F1 score, and inference time.

Main Results:

• The vocal cord selection model achieved a mean IoU of 0.6534 and Dice score of 0.7692, with 0.9972 accuracy.
• The laryngeal cancer detection model achieved a mean IoU of 0.6469 and Dice score of 0.7515, with 0.9860 accuracy.
• Real-time inference was achieved with processing times between 0.0244-0.0284 seconds per image.

Conclusions:

• An integrated AI platform combining vocal cord selection and lesion detection enables accurate and rapid laryngeal cancer detection.
• The developed models demonstrate high performance in identifying laryngeal cancer from laryngoscope images.
• This AI-driven approach shows promise for improving the diagnostic workflow in otolaryngology.