NeuroNasal: Advanced AI-Driven Self-Supervised Learning Approach for Enhanced Sinonasal Pathology Detection.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Machine Learning

Background:

• Sinus diseases significantly impact quality of life, causing symptoms like facial pain and reduced smell.
• Accurate diagnosis of sinus diseases is challenging due to factors like poor patient adherence to protocols.
• Artificial Intelligence (AI) offers a promising avenue to improve diagnostic precision for sinonasal pathology.

Purpose of the Study:

• To develop and evaluate a novel AI-based approach for detecting sinonasal pathology.
• To leverage Self-Supervised Learning (SSL) and Random Forest (RF) algorithms for improved classification accuracy.
• To introduce a new, expert-labeled dataset of CT and MRI images for sinonasal pathology research.

Main Methods:

• Utilized a new dataset of 137 CT and MRI images, labeled by expert radiologists into healthy and unhealthy classes.
• Employed the Deep InfoMax (DIM) model within a self-supervised framework to extract global and local image features.
• Integrated extracted features into a Random Forest (RF) classifier for distinguishing between healthy and pathological sinus cases.

Main Results:

• The AI-based approach demonstrated high efficacy in classifying sinonasal pathology.
• Achieved a mean classification accuracy of 92.62% in distinguishing healthy from diseased sinus cases.
• The combination of Deep InfoMax (DIM) and Random Forest (RF) proved effective for feature learning and classification.

Conclusions:

• The proposed AI-driven method shows significant potential for enhancing the accuracy and effectiveness of sinonasal pathology diagnosis.
• The developed dataset serves as a valuable resource for future research in AI-based medical image analysis for sinus diseases.
• This study highlights the successful application of Self-Supervised Learning (SSL) and Random Forest (RF) in improving diagnostic outcomes for sinus conditions.