Multi-scale deep reinforcement machine learning for n-dimensional segmentation in medical imaging

Abstract

Multi-scale deep reinforcement machine learning for N-dimensional segmentation in medical imaging. Multi-scale deep reinforcement learning generates a multi-scale deep reinforcement model (22) for multi-dimensional (e.g., 3D) segmentation of an object. In this context, segmentation is formulated as learning an image-driven policy (38) for shape evolution (40) that converges to the object boundary.The segmentation is treated as a reinforcement learning problem, and scale-space theory is used to enable robust and efficient multi-scale shape estimation. By learning an iterative strategy to findthe segmentation, the learning challenges of end-to-end regression systems may be addressed.

Description

[0001] Related application [0002] This patent document claims the rights to the filing date of provisional U.S. Patent Application Serial No. 62 / 500,604 filed on May 3, 2017 under 35 U.S.C. §119(e), which is incorporated herein by reference. Background technique [0003] This embodiment relates to segmentation in medical imaging. Accurate and rapid segmentation of anatomical structures is a basic task in medical image analysis that realizes real-time guidance, quantification and processing of diagnosis and intervention procedures. Previous solutions for 3D segmentation are based on machine learning-driven active shape models, forward propagation theory, Markov random field methods, or depth image-to-image regression models. [0004] The active shape model and the forward propagation theory solution propose a parametric surface model, which is deformed to fit the boundary of the target object. Machine learning technology uses image databases to learn complex parametric models. Th...

AI Technical Summary

Problems solved by technology

These methods can suffer from several limitations such as suboptimal local convergence and limited scalability

For high-resolution volumetric data, the use of scan patterns for boundary fitting leads to significant computational challenges

Images