Automatic identification and segmentation of target regions in pet imaging using dynamic protocol and modeling

Abstract

A continuous dynamic positron emission tomography (PET) assembly for imaging a target region of a subject. The assembly includes a radioactive tracer isotope injector configured to administer a radioactive isotope into the subject and a scintillator crystal configured to absorb ionizing radiation from the subject and emit scintillator light. The scintillator crystal undertakes the absorption substantially at the same time of the start of administering the radioactive isotope. The assembly also includes a photo detector in communication with the scintillator crystal, wherein the photodetector is configured to detect the emitted scintillation light as input and provide electrical signals as output. The assembly further includes a signal digitizing circuitry converting the output electrical signals into digital data. Moreover, the assembly includes a processor configured to receive the digital data and implement a model to convert the digital data into a three dimensional, tomographic image reconstruction.

We claim: 
     
         1 . A continuous dynamic positron emission tomography (PET) assembly for imaging a target region of a subject, said assembly comprising:
 a radioactive tracer isotope injector configured to administer a radioactive isotope into the subject;   a scintillator crystal configured to absorb ionizing radiation from the subject and emit scintillator light, wherein said scintillator crystal undertakes the absorption substantially at the same time of the start of administering the radioactive isotope;   a photo detector in communication with said scintillator crystal, wherein the photodetector is configured to detect the emitted scintillation light and provide electrical signals as output;   a signal digitizing circuitry converting the output electrical signals into digital data; and   a processor configured to receive the digital data and implement a model to convert the digital data into a three dimensional, tomographic image reconstruction.   
     
     
         2 . The PET assembly of  claim 1 , wherein said model comprises an artificial Neural Network (ANN). 
     
     
         3 . The PET assembly of  claim 1 , wherein undertaking the absorption, by the scintillator crystal, substantially at same time of the start of administering the radioactive isotope includes at least one of the following ranges of timing:
 about 1 minute to about 60 minutes after the start of administering;   about 5 minutes to about 30 minutes after the start of administering;   about 1 minute to about 25 minutes after the start of administering;   about 10 to about 30 minutes after the start of administering;   about 15 to about 20 minutes after the start of administering;   about 25 minutes after the start of administering;   about 30 to about 60 minutes after the start of administering; or   about 0 to 30 minutes after the start of administering;   about 25 to 30 minutes after the start of administering;   about 1 minute to about 10 minutes after the start of administering; or   about 60 minutes after the start of administering.   
     
     
         4 . The PET assembly of  claim 1 , wherein the target region comprises: breast, brain, head/neck, heart, liver, prostate, or lower extremities. 
     
     
         5 . The PET assembly of  claim 1 , wherein implementing the model comprises using a trained artificial neural network (ANN) model for image segmentation of at least a part of the target region of the subject. 
     
     
         6 . The PET assembly of  claim 5 , wherein the ANN is trained at least in part based on time activity curves associated with pixels, from regions of interest in PET images, in tumor regions and non-tumor regions. 
     
     
         7 . A method for continuous dynamic positron emission tomography (PET) imaging of a target region of a subject, said method comprising:
 administering a radioactive tracer isotope into the subject;   absorbing ionizing radiation from the subject and emitting scintillator light, wherein the absorption is substantially undertaken at the same time of the start of administering the radioactive isotope;   detecting the emitted scintillation light and providing electrical signals as output;   converting the output electrical signals into digital data; and   receiving the digital data and implementing a model to convert the digital data into a three dimensional, tomographic image reconstruction.   
     
     
         8 . The PET imaging method of  claim 7 , wherein said model comprises an artificial Neural Network (ANN). 
     
     
         9 . The PET imaging method of  claim 7 , wherein undertaking the absorption, by the scintillator crystal, substantially at same time of the start of administering the radioactive isotope includes at least one of the following ranges of timing:
 about 1 minute to about 60 minutes after the start of administering;   about 5 minutes to about 30 minutes after the start of administering;   about 1 minute to about 25 minutes after the start of administering;   about 10 to about 30 minutes after the start of administering;   about 15 to about 20 minutes after the start of administering;   about 25 minutes after the start of administering;   about 30 to about 60 minutes after the start of administering; or   about 60 minutes after the start of administering.   
     
     
         10 . The PET imaging method of  claim 7 , wherein the target region comprises: breast, brain, head/neck, heart, liver, prostate, or lower extremities. 
     
     
         11 . The PET imaging method of  claim 7 , wherein implementing the model comprises using a trained artificial neural network (ANN) model for image segmentation of at least a part of the target region of the subject. 
     
     
         12 . The PET imaging method of  claim 11 , wherein the ANN is trained at least in part based on time activity curves associated with pixels, from regions of interest in PET images, in tumor regions and non-tumor regions. 
     
     
         13 . A non-transitory computer readable medium having computer program logic that when implemented enables one or more processors in a positron emission tomography (PET) assembly to generate continuous dynamic positron emission tomography (PET) images of a target region of a subject, said computer program logic comprising:
 administering a radioactive tracer isotope into the subject;   absorbing ionizing radiation from the subject and emitting scintillator light, wherein the absorption is substantially undertaken at the same time of the start of administering the radioactive isotope;   detecting the emitted scintillation light and providing electrical signals as output;   converting the output electrical signals into digital data; and   receiving the digital data and implementing a model to convert the digital data into a three dimensional, tomographic image reconstruction.   
     
     
         14 . The non-transitory computer readable medium of  claim 13 , wherein said model comprises an artificial Neural Network (ANN). 
     
     
         15 . The non-transitory computer readable medium of  claim 13 , wherein undertaking the absorption, by the scintillator crystal, substantially at same time of the start of administering the radioactive isotope includes at least one of the following ranges of timing:
 about 1 minute to about 60 minutes after the start of administering;   about 5 minutes to about 30 minutes after the start of administering;   about 1 minute to about 25 minutes after the start of administering;   about 10 to about 30 minutes after the start of administering;   about 15 to about 20 minutes after the start of administering;   about 25 minutes after the start of administering;   about 30 to about 60 minutes after the start of administering; or   about 60 minutes after the start of administering.   
     
     
         16 . The non-transitory computer readable medium of  claim 13 , wherein the target region comprises: breast, brain, head/neck, heart, liver, prostate, or lower extremities. 
     
     
         17 . The non-transitory computer readable medium of  claim 13 , wherein implementing the model comprises using a trained artificial neural network (ANN) model for image segmentation of at least a part of the target region of the subject. 
     
     
         18 . The non-transitory computer readable medium of  claim 17 , wherein the ANN is trained at least in part based on time activity curves associated with pixels, from regions of interest in PET images, in tumor regions and non-tumor regions.