Data-driven surrogate respiratory signal generation for medical imaging

A non-transitory, machine-readable storage medium, tangibly embodying a program of instructions executable by a computer processor of a medical imaging system to cause the computer processor to perform operations comprising:

• retrieving, from a memory of the medical imaging system, projection data acquired by one or more gamma detectors of the medical imaging system, the projection data representing detected counts of photons emitted by a radioisotope within a patient, the counts detected at a plurality of projections, the counts for each projection binned into a plurality of time bins to provide respective frames of the projection data;
• computing a local neighborhood for each input of a plurality of input points associated with respective time bins, each input point being an M-dimensional vector representation of the frame corresponding to the associated time bin for one of the projections;
• within the local neighborhood of at least one of the input points, computing a distance between said at least one input point and each other input point in the local neighborhood according to an adaptive distance measure, wherein the adaptive distance measure is dependent on the projection data corresponding to said one projection;
• and
• generating an N-dimensional surrogate respiratory signal based on an optimization of a nonlinear objective function using the computed distances, wherein the surrogate respiratory signal is indicative of respiratory activity by the patient and N<M.

A method of signal processing for medical imaging, the method comprising:

• providing projection data acquired by one or more gamma detectors of a medical imaging system, the projection data representing detected counts of photons emitted by a radioisotope within a patient, the counts detected for a plurality of projections, the counts for each projection binned into a plurality of time bins to provide respective frames of the projection data;
• using a computer processor of the medical imaging system, computing a local neighborhood for each input point of a plurality of input points associated with respective time bins, each input point being an M-dimensional representation of the frame corresponding to the associated time bin for one of the projections, wherein each local neighborhood is adaptively computed based on the plurality of input points;
• using said computer processor, within the local neighborhood of at least one of the input points, computing a distance between said at least one input point and each other input point in the local neighborhood according to a distance measure dependent on the projection data corresponding to said one projection;
• and
• using said computer processor, generating an N-dimensional surrogate respiratory signal based on an optimization of a nonlinear objective function using the computed distances, wherein N<M and the surrogate respiratory signal is indicative of respiratory activity by the patient.