Systems and methods for deep-learning-based segmentation of composite images
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).
1. A system for automatically processing 3D images to identify 3D volumes corresponding to a particular target tissue region within a subject, the system comprising:
a processor of a computing device; and
a memory having instructions stored thereon, wherein the instructions, when executed by the processor, cause the processor to:
(a) receive a 3D anatomical image of the subject obtained using an anatomical imaging modality, wherein the 3D anatomical image comprises a graphical representation of tissue within the subject;
(b) receive a 3D functional image of the subject obtained using a functional imaging modality;
(c) automatically generate using a machine learning module, a 3D segmentation mask that identifies, within the 3D anatomical image and/or the 3D functional image, a 3D volume of interest (VOI) corresponding to the target tissue region,
wherein the machine learning module: (i) receives at least two channels of input, said input channels comprising a first input channel corresponding to at least a portion of the 3D anatomical image and a second input channel corresponding to at least a portion of the 3D functional image and (ii) utilizes both
the portion of the 3D anatomical image and the portion of the 3D functional image to perform image segmentation; and
(d) store and/or provide for display and/or further processing, the 3D segmentation mask, and
wherein the instructions cause the processor to resample the 3D functional image to match a resolution of the 3D anatomical image and use, at step (c), a portion of the resampled 3D functional image as the second input channel.
2. A method for automatically processing 3D images to identify 3D volumes corresponding to a particular target tissue region within a subject, the method comprising:
(a) receiving, by a processor of a computing device, a 3D anatomical image of the subject obtained using an anatomical imaging modality, wherein the 3D anatomical image comprises a graphical representation of tissue within the subject;
(b) receiving, by the processor, a 3D functional image of the subject obtained using a functional imaging modality;
(c) automatically generating, by the processor, using a machine learning module, a 3D segmentation mask that identifies, within the 3D anatomical image and/or the 3D functional image, a 3D volume of interest (VOI) corresponding to the target tissue region,
wherein the machine learning module: (i) receives at least two channels of input, said input channels comprising a first input channel corresponding to at least a portion of the 3D anatomical image and a second input channel corresponding to at least a portion of the 3D functional image and (ii) utilizes both the portion of the 3D anatomical image and the portion of the 3D functional image to perform image segmentation; and
(d) storing and/or providing, for display and/or further processing, by the processor, the 3D segmentation mask,
wherein the method comprises resampling, by the processor, the 3D functional image to match a resolution of the 3D anatomical image and using, at step (c), a portion of the resampled 3D functional image as the second input channel.