Classification of biological tissue by multi-mode data registration, segmentation and characterization

A non-transitory computer readable medium including computer program instructions configured to cause a computer processor to perform the steps of a method for classifying multiple voxels within dynamic contrast enhanced (DCE) magnetic resonance imaging (MRI) images acquired over a series of time periods, the method comprising:

• receiving a 2D or 3D image for each time period from among the corresponding set of DCE MRI images, the image representing a plurality of voxels having a respective plurality of voxel locations, the plurality of voxels including multiple voxels in a region of interest (ROI) being representative of the ROI in the corresponding time period;
• determining 2D or 3D textural features for at least the multiple voxels in the ROI in each image at each time period in the series of time periods;
• determining kinetic texture features for each of the voxel locations in the ROI across all time periods in the series of time periods to represent spatio-temporal changes in the textural features of each of the voxel locations in the ROI over the series of time periods;
• fitting a mathematical model to the determined kinetic texture features representing the changes in the textural features of each of the voxel locations in the ROI;
• determining parameter values of the mathematical model that is fitted to the changes in texture for each of the voxel locations in the ROI by selecting individual unpaired coefficients from the mathematical model as the determined parameters;
• applying a classifier to the determined kinetic texture features and to the determined parameters of the mathematical model of changes in spatio-temporal texture to identify an abnormality in the images and to classify the abnormality into one of at least two classes;
• and
• visually indicating the classified abnormalities as determined by the classifier for at least a portion of voxel locations in the ROI.

A method for classifying voxel locations in a region of interest (ROI) of a body from sets of dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) images acquired over a series of time periods, the method comprising:

• receiving image data representing a ROI slice for each time period, from among the corresponding set of DCE MRI images, the image data representing the ROI slice including a plurality of voxels having a respective plurality of voxel locations, being representative of the ROI in the corresponding time period;
• determining 2D or 3D textural features for the plurality of voxels in the ROI slice for each image at each time period in the series of time periods;
• determining kinetic texture features for each of the voxel locations across all time periods in the series of time periods to represent spatio-temporal changes in the textural features at each voxel location across all of the time periods;
• fitting a mathematical model to the determined kinetic texture features representing the changes in the textural features of each of the voxel locations;
• determining parameter values of the mathematical model that is fitted to the spatio-temporal changes in texture for each of the voxel locations by selecting individual unpaired coefficients from the mathematical model as the determined parameters;
• applying a classifier to the determined kinetic texture features and to the determined parameters of the mathematical model of changes in the spatio-temporal texture features to identify voxel locations corresponding to atherosclerotic plaques in the image and to classify the voxel locations corresponding to the plaques as stable or vulnerable;
• and
• visually indicating the classified voxel locations as determined by the classifier for at least a portion of voxel locations in the image.

A non-transitory computer readable medium including computer program instructions configured to cause a processor to perform a method for classifying a region of interest (ROI) of a body from sets of dynamic contrast enhanced (DCE) magnetic resonance imaging (MRI) images acquired over a series of time period, the method comprising the steps of:

• receiving image data representing a ROI slice for each time period, from among the corresponding set of DCE MRI images, the image data representing the ROI slice including a plurality of voxels having a respective plurality of voxel locations, being representative of the ROI in the corresponding time period;
• determining 2D or 3D textural features for the plurality of voxels in the ROI slice for each image at each time period in the series of time periods;
• determining kinetic texture features for each of the voxel locations across all time periods in the series of time periods to represent temporal changes in the textural features at each voxel location across all of the time periods;
• fitting a mathematical model to the determined kinetic texture features representing the changes in the textural features of each of the voxel locations;
• determining parameter values of the mathematical model that is fitted to these changes in texture for each of the voxel locations by selecting individual unpaired coefficients from the mathematical model as the determined parameters;
• applying a classifier to the determined kinetic texture features and to the determined parameters of the mathematical model of changes in the spatio-temporal texture to identify voxel locations corresponding to atherosclerotic plaques in the image and to classify the voxel locations corresponding to the plaques as stable or vulnerable;
• and
• visually indicating the classified voxel locations as determined by the classifier for at least a portion of voxel locations in the image.

A system for classifying voxel locations in a region of interest (ROI) in a body, comprising:

• a processor, having an input port configured to receive dynamic contrast enhanced (DCE) magnetic resonance imaging (MRI) images acquired over a series of time periods, the processor being configured to:
• receive image data representing an ROI slice for each time period, from among the corresponding set of DCE MRI images, the image data representing the ROI slice including a plurality of voxels having a respective plurality of voxel locations in the corresponding time period;
• automatically extract a boundary of the ROI from each respective received ROI slice in the series of time periods,
• determine 2D or 3D textural features for the plurality of voxels in the ROI slice for each image at each time period in the series of time periods;
• determine kinetic texture features for each of the voxel locations across all time periods in the series of time periods to represent spatio-temporal changes in the textural features at each voxel location across all of the time periods, based on the respective extracted boundary of the ROI in each ROI slice;
• fit a mathematical model to the determined kinetic texture features representing the changes in the textural features of each of the voxel locations;
• determine parameter values of the mathematical model that is fitted to the spatio-temporal changes in texture for each of the voxel locations by selecting individual unpaired coefficients from the mathematical model as the determined parameters;
• and
• apply a classifier to the determined kinetic texture features and to the determined parameters of the mathematical model of changes in the spatio-temporal texture features to identify voxel locations corresponding to atherosclerotic plaques in the image and to classify the voxel locations as stable or vulnerable;
• and
• a display device for visually indicating the classified voxel locations as determined by the classifier for at least a portion of voxel locations in the ROI.