Deep Learning Segmentation of Composite Nuclear Medicine Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current nuclear medicine imaging techniques face challenges in accurately segmenting and analyzing 3D volumes of interest, particularly in tissues with high tracer uptake, such as the bladder, prostate, and skeletal regions, where existing methods often result in inconclusive or inaccurate results.
Innovation Solution
The use of an artificial intelligence-based deep learning approach that combines anatomical and functional images, such as CT and PET/SPECT images, to generate a 3D segmentation mask, allowing for precise delineation of target tissue regions by leveraging information from both image types, including radiopharmaceutical uptake patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional nuclear medicine imaging techniques are used for segmentation, then the imaging process is simple, but segmentation accuracy deteriorates in tissues with high tracer uptake
Solution Approach 1:
The patent combines anatomical images (CT/MRI) and functional images (PET/SPECT) into a composite image that integrates both structural and metabolic information. This merging allows the segmentation algorithm to distinguish between anatomical structures and areas of high tracer uptake, resolving the contradiction by maintaining simple imaging protocols while achieving accurate segmentation through multi-modal image fusion.
Solution Approach 2:
The patent introduces a composite image as an intermediary representation that bridges anatomical and functional information. This intermediate composite image serves as the input for segmentation algorithms, enabling accurate identification of tissues with high tracer uptake without requiring complex direct analysis of separate modalities.
2Measurement precision
If only anatomical images are used for segmentation, then the imaging protocol is simple, but the ability to delineate target tissue regions deteriorates
Solution Approach 1:
The patent merges anatomical images and functional images into a single composite image structure that contains both types of information in an integrated format. This approach allows the segmentation algorithm to access both anatomical boundaries and functional characteristics without requiring separate processing streams, thus improving tissue region delineation while managing data volume efficiently.
3Measurement precision
If deep learning methods are applied to improve segmentation accuracy, then segmentation precision improves, but computational complexity increases
Solution Approach 1:
The patent performs preliminary processing of anatomical and functional images to create a composite image with integrated features before applying deep learning segmentation. This preliminary action pre-organizes the data in a format that reduces the computational burden during the actual segmentation process, allowing accurate results while managing computational resource consumption.
Data Source
AI summary
Presented herein are systems and methods that provide for improved 3D segmentation of nuclear medicine images using an artificial intelligence-based deep learning approach. For example, in certain embodiments, the machine learning module receives both an anatomical image (e.g., a CT image) and a functional image (e.g., a PET or SPECT image) as input, and generates, as output, a segmentation mask that identifies one or more particular target tissue regions of interest. The two images are interpreted by the machine learning module as separate channels representative of the same volume. Following segmentation, additional analysis can be performed (e.g., hotspot detection/risk assessment within the identified region of interest).