Updated automated PROMISE assessment : Treatment response evaluation based on PSMA PET/CT.
(2024) In Journal of Clinical Oncology 42(4_suppl). p.48-48- Abstract
Background: PSMA PETs can track treatment response in patients with prostate cancer. Recent integration of the RECIP framework with PROMISE v2 allows quantitative assessment of PSMA PETs. While useful for assessing differentially responding lesions, manual measurement and tracking is cumbersome and PSA reliance can be misleading. We evaluated an automated AI-enabled platform, aPROMISE, for longitudinal lesion tracking to evaluate treatment response based solely upon tumor volume. Methods: Patients with castration sensitive or resistant disease undergoing 18F-DCFPyL PET/CTs during workup and at ≥3 months after treatment were included. Tracer-avid lesions were identified using the aPROMISE algorithm. Detected lesions were manually... (More)
Background: PSMA PETs can track treatment response in patients with prostate cancer. Recent integration of the RECIP framework with PROMISE v2 allows quantitative assessment of PSMA PETs. While useful for assessing differentially responding lesions, manual measurement and tracking is cumbersome and PSA reliance can be misleading. We evaluated an automated AI-enabled platform, aPROMISE, for longitudinal lesion tracking to evaluate treatment response based solely upon tumor volume. Methods: Patients with castration sensitive or resistant disease undergoing 18F-DCFPyL PET/CTs during workup and at ≥3 months after treatment were included. Tracer-avid lesions were identified using the aPROMISE algorithm. Detected lesions were manually approved by an experienced nuclear medicine physician. After lesion identification and approval, the total tumor volume was calculated. The corresponding low-dose CT was auto-segmented, allowing anatomic compartment (prostate/bed, node, bone, or visceral) annotation of 18F-DCFPyL-avid lesions. Baseline and follow-up PETs were deformably co-registered to automatically match individual lesions if they were in the same anatomic compartment, and had overlap <10 mm center-to-center (<50 mm for rib lesions). Unmatched lesions on follow-up PET were noted as new. Response was assessed as CR (no avidity in follow-up), PR (volume decrease ≥30% without new lesions), SD (volume decrease ≥30% with new lesions), or PD (volume increase ≥20% plus new lesions). Volume changes at the total tumor and individual lesion levels were also assessed independent of RECIP imaging criteria. Results: The automated algorithm assigned the correct RECIP classification in 97% of the 67 cases with 14 as CR, 14 as PR, 20 as SD, and 19 as PD. 1027 of the 1686 lesions identified on baseline PET were seen on follow-up. 673 of the 1700 lesions identified on follow-up PET were new. Conclusions: For response assessment using 18F-DCFPyL PET/CT and RECIP imaging criteria, the automated algorithm and manual detection achieved the same clinical classifications. Automated approaches may allow more granular lesion analysis The precision and accuracy this automated process is being validated in prospective clinical studies.
(Less)
- author
- Benitez, Cecil Mayra ; Sahlstedt, Hannicka LU ; Brynolfsson, Johan LU ; Berenji, Gholam Reza ; Juarez, Jesus Eduardo ; Kane, Nathanael ; Tsai, Sonny ; Rettig, Matthew ; Nickols, Nicholas George and Duriseti, Sai
- publishing date
- 2024
- type
- Contribution to journal
- publication status
- published
- subject
- in
- Journal of Clinical Oncology
- volume
- 42
- issue
- 4_suppl
- pages
- 1 pages
- publisher
- Lippincott Williams & Wilkins
- external identifiers
-
- scopus:105023971020
- ISSN
- 0732-183X
- DOI
- 10.1200/JCO.2024.42.4_suppl.48
- language
- English
- LU publication?
- no
- additional info
- Publisher Copyright: © 2024 by American Society of Clinical Oncology
- id
- f3a548b8-9b35-4dd1-beb6-776eb6620f46
- date added to LUP
- 2026-03-09 13:22:54
- date last changed
- 2026-03-25 15:55:58
@article{f3a548b8-9b35-4dd1-beb6-776eb6620f46,
abstract = {{<p>Background: PSMA PETs can track treatment response in patients with prostate cancer. Recent integration of the RECIP framework with PROMISE v2 allows quantitative assessment of PSMA PETs. While useful for assessing differentially responding lesions, manual measurement and tracking is cumbersome and PSA reliance can be misleading. We evaluated an automated AI-enabled platform, aPROMISE, for longitudinal lesion tracking to evaluate treatment response based solely upon tumor volume. Methods: Patients with castration sensitive or resistant disease undergoing 18F-DCFPyL PET/CTs during workup and at ≥3 months after treatment were included. Tracer-avid lesions were identified using the aPROMISE algorithm. Detected lesions were manually approved by an experienced nuclear medicine physician. After lesion identification and approval, the total tumor volume was calculated. The corresponding low-dose CT was auto-segmented, allowing anatomic compartment (prostate/bed, node, bone, or visceral) annotation of <sup>18</sup>F-DCFPyL-avid lesions. Baseline and follow-up PETs were deformably co-registered to automatically match individual lesions if they were in the same anatomic compartment, and had overlap <10 mm center-to-center (<50 mm for rib lesions). Unmatched lesions on follow-up PET were noted as new. Response was assessed as CR (no avidity in follow-up), PR (volume decrease ≥30% without new lesions), SD (volume decrease ≥30% with new lesions), or PD (volume increase ≥20% plus new lesions). Volume changes at the total tumor and individual lesion levels were also assessed independent of RECIP imaging criteria. Results: The automated algorithm assigned the correct RECIP classification in 97% of the 67 cases with 14 as CR, 14 as PR, 20 as SD, and 19 as PD. 1027 of the 1686 lesions identified on baseline PET were seen on follow-up. 673 of the 1700 lesions identified on follow-up PET were new. Conclusions: For response assessment using <sup>18</sup>F-DCFPyL PET/CT and RECIP imaging criteria, the automated algorithm and manual detection achieved the same clinical classifications. Automated approaches may allow more granular lesion analysis The precision and accuracy this automated process is being validated in prospective clinical studies.</p>}},
author = {{Benitez, Cecil Mayra and Sahlstedt, Hannicka and Brynolfsson, Johan and Berenji, Gholam Reza and Juarez, Jesus Eduardo and Kane, Nathanael and Tsai, Sonny and Rettig, Matthew and Nickols, Nicholas George and Duriseti, Sai}},
issn = {{0732-183X}},
language = {{eng}},
number = {{4_suppl}},
pages = {{48--48}},
publisher = {{Lippincott Williams & Wilkins}},
series = {{Journal of Clinical Oncology}},
title = {{Updated automated PROMISE assessment : Treatment response evaluation based on PSMA PET/CT.}},
url = {{http://dx.doi.org/10.1200/JCO.2024.42.4_suppl.48}},
doi = {{10.1200/JCO.2024.42.4_suppl.48}},
volume = {{42}},
year = {{2024}},
}