Machine Learning Algorithms to Detect Sex in Myocardial Perfusion Imaging

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Myocardial perfusion imaging (MPI) is crucial for diagnosing coronary artery disease.
• Machine learning (ML) offers transformative potential in healthcare applications.
• The General Data Protection Regulation (GDPR) emphasizes data security and individual privacy.

Purpose of the Study:

• To develop and evaluate ML models for sex determination using MPI data.
• To assess the performance of various ML algorithms in classifying sex from myocardial scintigraphy.
• To explore the implications of these findings on patient data privacy and GDPR.

Main Methods:

• Trained seven ML models (CART, NB, KNN, SVM, AB, RF, GB) on 260 MPI polar maps (140 male, 120 female).
• Utilized a cross-validation strategy for model evaluation.
• Assessed model performance using accuracy, standard deviation, precision, and area under the ROC curve.

Main Results:

• All ML models achieved >82% accuracy in sex determination from MPI.
• Support Vector Machine (SVM) reached 90% accuracy, with KNN, RF, AB, and GB also showing high performance.
• SVM and Random Forests (RF) demonstrated the best area under the ROC curve (0.93).

Conclusions:

• ML algorithms are effective for assessing patient sex from myocardial scintigraphy.
• High accuracy in sex determination presents challenges for patient data confidentiality under GDPR.
• The study contributes to the ongoing debate on defining sensitive data in the context of data protection regulations.