V3DQutrit a volumetric medical image segmentation based on 3D qutrit optimized modified tensor ring model

Area of Science:

• Quantum Computing
• Artificial Intelligence
• Medical Imaging

Background:

• Conventional Convolutional Neural Networks (CNNs) face challenges in medical image segmentation, including slow convergence and high computational complexity.
• Existing Quantum-Inspired Neural Networks (QINNs) are often limited to grayscale segmentation, restricting their applicability to complex volumetric data.

Purpose of the Study:

• Introduce 3D-QTRNet, a novel quantum-inspired neural network designed for volumetric medical image segmentation.
• Address limitations of existing CNNs and QINNs by improving accuracy, memory efficiency, and convergence speed.

Main Methods:

• Utilize qutrit encoding for enhanced data representation.
• Employ tensor ring decomposition for efficient model architecture.
• Develop a quantum-inspired neural network for 3D medical image segmentation.

Main Results:

• Demonstrate superior performance on the BRATS19 and Spleen datasets.
• Achieve higher Dice similarity and segmentation precision compared to state-of-the-art CNN and quantum models.
• Showcase improved memory usage and accelerated model convergence.

Conclusions:

• 3D-QTRNet offers a scalable and effective solution for volumetric medical image segmentation.
• This work successfully bridges quantum computing principles with practical medical imaging applications.
• The proposed model represents a significant advancement in the field, paving the way for real-world clinical use.