Machine learning framework for simulation of artifacts in paranasal sinuses diagnosis using CT images

Area of Science:

• Medical imaging
• Artificial intelligence in medicine
• Computational pathology

Background:

• Accurate diagnosis is critical for effective patient care in modern medicine.
• Deep neural networks have significantly advanced medical diagnostic tools.
• Computed tomography (CT) scans provide crucial anatomical data for sinus surgery planning.

Purpose of the Study:

• To develop a machine learning algorithm for accurate paranasal sinus disorder detection and diagnosis.
• To automate the evaluation of paranasal sinuses in CT scans, reducing manual review.
• To improve diagnostic efficiency and accuracy in medical imaging analysis.

Main Methods:

• Development of a unique machine learning approach for paranasal sinus analysis.
• Implementation of automatic segmentation to isolate and crop the paranasal sinus region.
• Training the algorithm on CT scan data to identify abnormalities.

Main Results:

• The proposed method achieved 95.16% accuracy in diagnosing sinus irregularities.
• The system demonstrated a high sensitivity of 99.14% in detecting disorders.
• Automated segmentation significantly reduced the data required for model training, enhancing computational efficiency.

Conclusions:

• The developed machine learning method offers an efficient and accurate solution for diagnosing paranasal sinus disorders.
• Automatic segmentation eliminates the need for manual cropping, saving time and reducing errors.
• This technology holds promise for improving diagnostic workflows in medical imaging.