D Yang^1,2,3*, R Alfano⁴, G Bauman^3,4, J Chin^3,4, S Pautler^3,4, K Chung^2,3, A Ward⁴, T Lee^2,3, (1) Fox Chase Cancer Center, Philadelphia, PA, (2) Robarts Research Institute, University of Western Ontario, London, ON, CA, (3) Lawson Imaging Research Program, Lawson Health Research Institute, London Health Science Centre, London, ON, CA, (4) London Regional Cancer Program, London Health Science Centre, London, ON, CA

• 
  D Yang

• 
  

MO-B-TRACK 6-7 (Monday, 7/26/2021) 11:30 AM - 12:30 PM [Eastern Time (GMT-4)]

Purpose: Identifying the dominant prostate cancer nodule (dominant intraprostatic lesion, DIL) may enable more accurate diagnosis and therapy through more precise targeting of biopsy, radiotherapy, and focal ablative therapies. Our goal is to validate the performance of [¹⁸F]DCFPyL positron-emission tomography (PET) and CT Perfusion (CTP) for localizing DIL against digital histopathology images.

Methods: Multi-modality image sets: in-vivo T2-weighted (T2w)-magnetic resonance imaging (MRI), 22-min dynamic [¹⁸F]DCFPyL PET/CT, 2-hr post-injection PET/MR, CTP were acquired in patients prior to radical prostatectomy. The explanted gland with implanted fiducial markers was imaged with T2w-MRI. All PET, CT, and MR images were co-registered to the pathologist-annotated whole-mount mid-gland prostatectomy histology using fiducial markers and anatomical landmarks. Regions of interest encompassing DIL and non-DIL tissue were drawn on the digital histopathology images and superimposed on PET and CTP parametric maps. Logistic regression with backward elimination of parameters was used to select the most sensitive parameter set to distinguish DIL from non-DIL voxels. Leave-one-patient-out cross-validation was performed to determine diagnostic performance.

Results: 49,256 voxels from [¹⁸F]DCFPyL PET and CTP parametric maps of 15 patients were analyzed. 2-hr post-injection [¹⁸F]DCFPyL SUV (SUV_late) and a model combining Kᵢ (net uptake rate constant) and k₄ (dissociation rate constant) of [¹⁸F]DCFPyL achieved the most accurate performance distinguishing DIL from non-DIL voxel. Both detection models achieved an AUC of 0.90 and an error rate of <10% from leave-one-patient-out cross-validation. Compared to digital histopathology, the detected DILs had a mean dice similarity coefficient of 0.76 for the Kᵢ and k₄ model and 0.73 for SUV_late.

Conclusion: We have validated using co-registered digital histopathology images that parameters from kinetic analysis of 22-min dynamic [¹⁸F]DCFPyL PET can accurately localize DIL in prostate cancer for targeting of biopsy, radiotherapy, and focal ablative therapies. Short duration dynamic [¹⁸F]DCFPyL PET was not inferior to SUV_late in this diagnostic task.

Funding Support, Disclosures, and Conflict of Interest: Author Ting-Yim Lee licenses the CT Perfusion software to GE Healthcare. All other authors have nothing to disclose.

Handouts

Keywords

PET, Quantitative Imaging, Pharmacokinetic Modeling

Taxonomy

IM- PET : Uptake kinetics

Contact Email