Methods and related aspects for pathology prognosis

A method of generating a model to predict prospective pathology scores of test subjects having a pathology, the method comprising:

• extracting a first plurality of image features from sets of longitudinal single photon emission computed tomography (SPECT) and/or positron emission tomography (PET) images obtained from a plurality of reference subjects having the pathology, wherein the images features extracted from at least first and second sets of the longitudinal SPECT and/or PET images are extracted separate from one another to produce a first feature vector;
• extracting a second plurality of image features from the sets of the longitudinal SPECT and/or PET images when the longitudinal SPECT and/or PET images are in an unprocessed form to produce a second feature vector;
• extracting a plurality of clinical features from clinical data obtained from the plurality of reference subjects having the pathology, which clinical features comprise pathology sub-scores to produce a third feature vector;
• and,
• training one or more layers of an artificial neural network (ANN) using the first, second, and third feature vectors, thereby generating the model to predict prospective pathology scores of test subjects having the pathology.

A system, comprising at least one controller that comprises, or is capable of accessing, computer readable media comprising non-transitory computer executable instructions which, when executed by at least one electronic processor, perform at least:

• extracting a first plurality of image features directly from sets of longitudinal single photon emission computed tomography (SPECT) and/or positron emission tomography (PET) images obtained from a plurality of reference subjects having the pathology, wherein the images features extracted from at least first and second sets of the longitudinal SPECT and/or PET images are extracted separately from one another to produce a first feature vector;
• extracting a second plurality of image features from maximum intensity projections (MIPs) of the sets of the longitudinal SPECT and/or PET images to produce a second feature vector;
• extracting a third plurality of image features from semi-quantitative imaging measures of the sets of the longitudinal SPECT and/or PET images to produce a third feature vector;
• extracting a plurality of non-imaging features from non-imaging data obtained from the plurality of reference subjects having the pathology to produce a fourth feature vector;
• training multiple artificial neural networks (ANNs) using the first, second, third, and fourth feature vectors and a plurality of clinical features from clinical data obtained from the plurality of reference subjects having the pathology to produce an ensemble of ANNs;
• and,
• using the ensemble of ANNs to predict a prospective pathology score of a test subject having the pathology.

A system, comprising at least one controller that comprises, or is capable of accessing, computer readable media comprising non-transitory computer executable instructions which, when executed by at least one electronic processor, perform at least:

• extracting a first plurality of image features from sets of longitudinal single photon emission computed tomography (SPECT) and/or positron emission tomography (PET) images obtained from a plurality of reference subjects having the pathology, wherein the image features extracted from at least first and second sets of the longitudinal SPECT and/or PET images are extracted separately from one another to produce a first feature vector;
• extracting a second plurality of image features from the sets of the longitudinal SPECT and/or PET images when the longitudinal SPECT and/or PET images are in an unprocessed form to produce a second feature vector;
• extracting a plurality of clinical features from clinical data obtained from the plurality of reference subjects having the pathology, which clinical features comprise pathology sub-scores to produce a third feature vector;
• training one or more layers of an artificial neural network (ANN) using the first, second, and third feature vectors to generate a predictive model;
• and,
• using the predictive model to predict a prospective pathology score of a test subject having the pathology.

A computer readable media comprising non-transitory computer executable instruction which, when executed by at least electronic processor perform at least:

• extracting a first plurality of image features directly from sets of longitudinal single photon emission computed tomography (SPECT) and/or positron emission tomography (PET) images obtained from a plurality of reference subjects having the pathology, wherein the images features extracted from at least first and second sets of the longitudinal SPECT and/or PET images are extracted separate from one another to produce a first feature vector;
• extracting a second plurality of image features from maximum intensity projections (MIPs) of the sets of the longitudinal SPECT and/or PET images to produce a second feature vector;
• extracting a third plurality of image features from semi-quantitative imaging measures of the sets of the longitudinal SPECT and/or PET images to produce a third feature vector;
• extracting a plurality of non-imaging features from non-imaging data obtained from the plurality of reference subjects having the pathology to produce a fourth feature vector;
• training multiple artificial neural networks (ANNs) using the first, second, third, and fourth feature vectors and a plurality of clinical features from clinical data obtained from the plurality of reference subjects having the pathology to produce an ensemble of ANNs;
• and,
• using the ensemble of ANNs to predict a prospective pathology score of a test subject having the pathology.

A computer readable media comprising non-transitory computer executable instruction which, when executed by at least electronic processor perform at least:

• extracting a first plurality of image features from sets of longitudinal single photon emission computed tomography (SPECT) and/or positron emission tomography (PET) images obtained from a plurality of reference subjects having the pathology, wherein the images features extracted from at least first and second sets of the longitudinal SPECT and/or PET images are extracted separately from one another to produce a first feature vector;
• extracting a second plurality of image features from the sets of the longitudinal SPECT and/or PET images when the longitudinal SPECT and/or PET images are in an unprocessed form to produce a second feature vector;
• extracting a plurality of clinical features from clinical data obtained from the plurality of reference subjects having the pathology, which clinical features comprise pathology sub-scores to produce a third feature vector;
• training one or more layers of an artificial neural network (ANN) using the first, second, and third feature vectors to generate a predictive model;
• and,
• using the predictive model to predict a prospective pathology score of a test subject having the pathology.