Systems and methods for artificial intelligence-based image analysis for cancer assessment
Presented herein are systems and methods that provide for automated analysis of medical images to determine a predicted disease status (e.g., prostate cancer status) and/or a value corresponding to predicted risk of the disease status for a subject. The approaches described herein leverage artificial intelligence (AI) to analyze intensities of voxels in a functional image, such as a PET image, and determine a risk and/or likelihood that a subject's disease, e.g., cancer, is aggressive. The approaches described herein can provide predictions of whether a subject that presents a localized disease has and/or will develop aggressive disease, such as metastatic cancer. These predictions are generated in a fully automated fashion and can be used alone, or in combination with other cancer diagnostic metrics (e.g., to corroborate predictions and assessments or highlight potential errors). As such, they represent a valuable tool in support of improved cancer diagnosis and treatment.
1. A method for determining a predicted disease status and/or a value corresponding to predicted risk of the disease status based on automated analysis of intensities of a three-dimensional (3D) functional image, the method comprising:
(a) receiving, by a processor of a computing device, a 3D anatomical image of a subject obtained using an anatomical imaging modality, wherein the 3D anatomical image comprises a graphical representation of tissue within the subject;
(b) identifying, by the processor, within the 3D anatomical image, a target volume of interest (VOI) corresponding to a prostate region of the subject;
(c) receiving, by the processor, a 3D functional image of the subject obtained using a functional imaging modality; and
(d) determining, by the processor, a predicted disease status of the subject and/or a value corresponding to predicted risk of the disease status of the subject using a machine learning module that receives, as input, intensities of voxels of the 3D functional identified as corresponding to the target VOI of the anatomical image.
2. The method of claim 1 , wherein the method comprises identifying the voxels of the 3D functional image corresponding to the target VOI of the anatomical image by:
identifying, by the processor, within the 3D functional image, a 3D prostate volume corresponding to the target VOI identified within the 3D anatomical image and identifying voxels of the 3D functional image lying within the 3D prostate volume as corresponding to the target VOI of the anatomical image.
3. The method of claim 1 , wherein the machine learning module further receives, as input, one or more clinical variables.
4. The method of claim 3 , wherein the one or more clinical variables comprise one or more members selected from the group consisting of:
a race/ethnicity;
a prostate specific antigen (PSA) level and/or velocity;
a hemoglobin level;
a lactate dehydrogenase level;
an albumin level;
a clinical T stage;
a biopsy Gleason score; and
a percentage positive core score.
5. The method of claim 1 , comprising determining the predicted disease status, wherein the predicted disease status is a classification corresponding to a prediction of aggressive disease status.
6. The method of claim 5 , wherein the classification comprises one or more of the following classes corresponding to predictions of whether the subject has and/or likely will develop one or more metastases:
an overall metastases class, wherein assignment to the overall metastases class corresponds to a prediction that the subject has and/or will likely develop one or more metastases;
one or more particular metastases class(es), each corresponding to a particular type of metastases wherein assignment to the particular metastases class corresponds to a prediction that the subject has and/or will likely develop the particular type of metastases; and
a no metastases class, wherein assignment to the no metastases class corresponds to a prediction that the subject had not and/or is not likely to develop one or more metastases.
7. The method of claim 5 , wherein the machine learning module generates, as output, one or more likelihood values representing likelihood(s) of overall metastases and/or one or more particular types of metastases, and wherein determining the classification comprises comparing the one or more likelihood values with one or more thresholds.
8. The method of claim 1 , wherein the method comprises determining the value corresponding to predicted risk of the disease status of the subject, and wherein the disease status is an aggressive disease status.
9. The method of claim 8 , wherein the value represents a likelihood that the subject has and/or will develop one or more metastases.
10. The method of claim 9 , wherein the value represents a likelihood that the subject has and/or will develop one or more of a particular type of metastases.
11. The method of claim 8 , comprising determining a plurality of values, each corresponding to a particular type of metastases and representing a likelihood that the subject has and/or will develop one or more of the particular type of metastases.
12. The method of claim 1 , wherein the disease is prostate cancer.
13. The method of claim 1 , wherein the machine learning module comprises a convolutional neural network (CNN).
14. The method of claim 1 , wherein the anatomical image is a CT scan.
15. The method of claim 1 , wherein the functional image is a PET image obtained following administration of a radiopharmaceutical to the subject.
16. The method of claim 15 , wherein the radiopharmaceutical comprises a prostate specific membrane antigen (PSMA) binding agent.
17. The method of claim 16 , wherein the PSMA binding agent is [18F]DCFPyL.
18. The method of claim 1 , wherein the functional image is a SPECT image.
19. A system for determining a predicted disease status and/or a value corresponding to predicted risk of the disease status based on automated analysis of intensities of a three-dimensional (3D) functional image, 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 a subject obtained using an anatomical imaging modality, wherein the 3D anatomical image comprises a graphical representation of tissue within the subject;
(b) identify, within the 3D anatomical image, a target volume of interest (VOI) corresponding to a prostate region of the subject;
(c) receive a 3D functional image of the subject obtained using a functional imaging modality; and
(d) determine a predicted disease status of the subject and/or a value corresponding to predicted risk of the disease status of the subject using a machine learning module that receives, as input, intensities of voxels of the 3D functional identified as corresponding to the target VOI of the anatomical image.
20. A method for determining a predicted disease status of a subject and/or a value corresponding to predicted risk of the disease status of the subject based on automated analysis of intensities of a three-dimensional (3D) functional image, 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) identifying, by the processor, within the 3D anatomical image, a first volume corresponding to a target tissue volume of interest (VOI) of the subject;
(c) receiving, by the processor, a 3D functional image of the subject obtained using a functional imaging modality; and
(d) determining, by the processor, a predicted disease status and/or a value corresponding to a predicted risk of the disease status using a machine learning module that receives, as input, intensities of voxels of the 3D functional image identified as corresponding to the first volume of the 3D anatomical image and one or more clinical variables selected from the group consisting of:
a race/ethnicity;
a prostate specific antigen (PSA) level and/or velocity;
a hemoglobin level;
a lactate dehydrogenase level;
an albumin level;
a clinical T stage;
a biopsy Gleason score; and
a percentage positive core score.
21. A system for determining a predicted disease status of a subject and/or a value corresponding to predicted risk of the disease status of the subject based on automated analysis of intensities of a three-dimensional (3D) functional image, 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) identify, within the 3D anatomical image, a first volume corresponding to a target tissue volume of interest (VOI) of the subject;
(c) receive a 3D functional image of the subject obtained using a functional imaging modality; and
(d) determine a predicted disease status and/or a value corresponding to a predicted risk of the disease status using a machine learning module that receives, as input, intensities of voxels of the 3D functional image identified as corresponding to the first volume of the 3D anatomical image and one or more clinical variables selected from the group consisting of:
a race/ethnicity;
a prostate specific antigen (PSA) level and/or velocity;
a hemoglobin level;
a lactate dehydrogenase level;
an albumin level;
a clinical T stage;
a biopsy Gleason score; and
a percentage positive core score.