Medical imaging processing with quantum computing

Medical imaging processing with quantum computing

Seminar author:Laia Domingo

Event date and time:05/16/2024 04:00:pm

Event location:

Event contact:

Classical machine learning faces challenges in handling the complexity and high dimensionality of medical imaging data, often resulting in slower and less accurate diagnoses across diverse patient populations. Ensuring the robustness and generalizability of machine learning models is crucial to their successful deployment in clinical practice In this study, we propose a self-supervised approach for automated image segmentation in breast cancer mammography datasets. We introduce a quantum-inspired method for optimal image representation, which incorporates a quantum transformation to evaluate pixel intensity differentials and a multilevel adaptive sigmoidal activation function. This approach significantly enhances image representation while minimizing computational overhead. The resulting images are inputted into a self-supervised segmentation algorithm, formulated as a min-cut optimization problem, utilizing quantum annealing to identify regions of interest. Notably, our holistic approach obviates the need for annotated datasets or training complex neural networks, mitigating overfitting risks and computational expenses associated with traditional methods, yet yielding comparable results.