Automatic medical protocol classification using machine learning approaches.

Area of Science:

• Medical Imaging Informatics
• Artificial Intelligence in Healthcare
• Natural Language Processing

Background:

• Assigning medical imaging protocols requires extensive patient data analysis from reports.
• Automating protocol assignment can significantly enhance diagnostic efficiency.
• Natural Language Processing (NLP) is effective for extracting information from clinical text.

Purpose of the Study:

• To develop and evaluate machine learning models for automated medical imaging protocol assignment.
• To leverage NLP techniques for extracting relevant patient data from radiological reports.
• To improve the efficiency and accuracy of diagnostic procedures.

Main Methods:

• Developed machine learning classification models using NLP on a large corpus of CT and MRI reports.
• Utilized traditional machine learning (SVM, Random Forest), neural networks, and transfer learning.
• Trained and validated models on nearly 700,000 clinical imaging examinations.

Main Results:

• The best-performing system achieved 92.2% accuracy on the CT dataset.
• The system demonstrated 86.9% accuracy on the MRI dataset.
• The models successfully performed complex multi-class text classification tasks.

Conclusions:

• The developed machine learning system is efficient, high-quality, and cost-effective.
• The system is currently employed as a decision support tool for radiologists.
• It assists in assigning protocols for CT and MRI studies in real clinical settings.