Deep Learning for Automated Measures of SUV and Molecular Tumor Volume in [ <sup>68</sup> Ga]PSMA-11 or [ <sup>18</sup> F]DCFPyL, [ <sup>18</sup> F]FDG, and [ <sup>177</sup> Lu]Lu-PSMA-617 Imaging with Global Threshold Regional Consensus Network
Price Jackson, James P Buteau, Lachlan McIntosh, Y. Sun, Raghava Kashyap, Sebastián Casanueva, Aravind S. Ravi Kumar, Shahneen Sandhu, Arun Azad, Ramin Alipour, Javad Saghebi, Grace Kong, Kerry Jewell, Michal Eifer, Neeraja Bollampally, Michael S. Hofman
Abstract
Delineation of disease extent and tracer avidity can be performed with a high degree of accuracy using automated deep learning methods. By incorporating threshold-based postprocessing, the tools can closely match the output of manual workflows. Pretrained models and scripts to adapt to institutional data are provided for open use.
Topics & Concepts
Artificial intelligenceGround truthDiceSegmentationNuclear medicineSørensen–Dice coefficientWorkflowDeep learningMedicineComputer sciencePositron emission tomographyVolume (thermodynamics)Pattern recognition (psychology)Similarity (geometry)Automated methodProstate cancerImage segmentationMedical physicsMedical imagingMachine learningJaccard indexStage (stratigraphy)Image registrationMedical Imaging Techniques and ApplicationsRadiopharmaceutical Chemistry and ApplicationsProstate Cancer Treatment and Research