Hybrid Quantum Image Classification and Federated Learning for Hepatic Steatosis Diagnosis

Area of Science:

• Medical Imaging
• Quantum Machine Learning
• Hepatology

Background:

• Accurate hepatic steatosis assessment is vital in liver transplantation.
• Current diagnostic methods require improvement for efficiency and precision.
• Data privacy concerns hinder collaborative development of automated solutions.

Purpose of the Study:

• To develop advanced algorithms for enhanced liver biopsy image classification.
• To address data privacy challenges in developing automated diagnostic tools.
• To improve diagnostic accuracy and efficiency in assessing non-alcoholic liver steatosis.

Main Methods:

• Utilized quantum machine learning techniques for superior generalization.
• Implemented privacy-conscious collaborative machine learning with federated learning.
• Developed a hybrid quantum neural network model using real-world clinical data.

Main Results:

• Achieved 97% accuracy in liver biopsy image classification, surpassing traditional methods by 1.8%.
• Maintained over 90% accuracy using federated learning for privacy-preserving data sharing.
• Demonstrated a scalable, collaborative, and efficient computational framework.

Conclusions:

• The hybrid quantum neural network model offers a significant advancement in diagnosing non-alcoholic liver steatosis.
• Federated learning effectively addresses privacy concerns in collaborative AI development for medical diagnostics.
• This framework supports clinical pathologists by providing dependable and efficient diagnostic assistance.