Volume visualization using tissue mix

A non-transitory computer-readable medium that includes a program that, when executed by a processor, causes the processor to:

• receive a data set that includes a plurality of voxels, each voxel corresponding to a location within a volume, each voxel including a measure of a physical property at the location of the voxel,
• determine a likelihood of each tissue type or tissue mix of a plurality of tissue types or tissue mix contributing to the measure at each voxel, the likelihood of each tissue type forming a plurality of likelihoods associated with each voxel,
• determine a classification of each voxel based on the associated plurality of likelihoods, and
• provide a display based on the classification of select voxels;
• wherein determining the likelihood includes:
• determining a first gradient magnitude at the voxel,
• determining an average measure and a second gradient magnitude in a neighborhood around the voxel, and
• fitting the measure, average measure, first gradient, and second gradient to a model of expected behavior for select tissue transitions to provide a best-lit estimate of the likelihood of the select tissue transition.

A method of mapping voxel values to object elements by a classification machine, comprising:

• identifying a plurality of object elements,
• creating a plurality of models of transitions between pairs of object elements, defining how parameters of each model change when crossing an edge between the object elements of each pair,
• receiving a set of voxel values corresponding to a set of measurements of a physical property at locations in a volume of space,
• determining a set of gradient values corresponding to changes of voxel values between adjacent locations,
• for each location:
• classifying the location as corresponding to a mix of a particular pair of object elements if the voxel values and gradient values in a vicinity of the location correspond to the parameters of the model of transitions between the particular pair of object elements;
• and
• classifying the location as corresponding to a single object element based on the voxel values and gradient values if the voxel values and gradient values do not correspond to the parameters of any of the plurality of models of transitions;
• and
• providing a representation of the classification of each location.

A non-transitory computer-readable medium that includes a program that, when executed by a processor, causes the processor to:

• receive a plurality of models of transitions between pairs of object elements of a plurality of object elements, defining how parameters of each model change when crossing an edge between the object elements of each pair,
• receive a set of voxel values corresponding to a set of measurements of a physical property at locations in a volume of space,
• determine a set of gradient values corresponding to changes of voxel values between adjacent locations,
• for each location:
• classify the location as corresponding to a mix of a particular pair of object elements if the voxel values and gradient values in a vicinity of the location correspond to the parameters of the model of transitions between the particular pair of object elements;
• and
• classify the location as corresponding to a single object element based on the voxel values and gradient values if the voxel values and gradient values do not correspond to the parameters of any of the plurality of models of transitions;
• and
• provide a representation of the classification of each location.