aPROMISE

aPROMISE is a cloud-based software platform designed to assist healthcare professionals and researchers in the comprehensive analysis of PSMA PET/CT images for oncological assessment. It employs AI-driven segmentation and hotspot detection, enabling users to mark regions of interest and obtain automated quantitative metrics like SUV, thereby aiming to provide more reproducible and sensitive characterization of potential metastatic disease compared to manual methods.

aPROMISE is a cloud-based software platform designed for healthcare professionals and researchers to manage and process PSMA PET/CT medical images for oncological applications. It facilitates the acceptance, transfer, storage, display, manipulation, quantification, and reporting of these DICOM 3 compliant images. The system allows users to upload multiple scans, review them in a comprehensive 4-panel view (PET, CT, PET/CT fusion, MIP), and utilize image visualization and analysis tools. A key function is its AI-enabled automated segmentation of skeleton and soft tissue organs, along with an algorithm to detect high-intensity regions of interest (hotspots) indicative of tracer uptake. Users interact with aPROMISE by reviewing all algorithm-detected hotspots and can either select system-suggested pre-defined segmentations or manually draw ROIs. These selected or drawn hotspots then undergo automatic quantitative analysis, providing metrics such as Standard Uptake Values (SUV-max, -mean, -peak), volume, Lesion Index (LI), and Intensity-weighted Tissue Lesion Volume (ITLV). The user ultimately reviews these quantitative results to identify suspicious lesions for reporting, which requires quality control confirmation and electronic signature. Performance data from bench testing, including digital phantom validation for accuracy and comparison to a predicate device for functional equivalence, supports its specifications. Notably, an analytical performance study demonstrated that aPROMISE's automated quantification of tracer uptake in reference organs is more reproducible and consistent than manual methods, and it showed high sensitivity in pre-selecting ROIs suspicious for metastatic disease.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the AI-enabled automated segmentation of soft tissue organs in aPROMISE compare in accuracy and robustness to its skeletal segmentation, particularly across diverse patient anatomies and image qualities?; What is the specific algorithm or methodology used for "hotspot detection" in PET series, and how does its performance (sensitivity/specificity) vary with different levels of PSMA tracer uptake and lesion sizes?; The "Analytical Performance in Clinical Study" showed aPROMISE had "high sensitivity in pre-selection of regions of interest." What was the corresponding specificity, and what were the cohort size and characteristics for this study?; How does aPROMISE handle variations in PSMA PET/CT acquisition protocols (e.g., scanner differences, reconstruction parameters, tracer injection to scan time) and their potential impact on quantitative SUV measurements and hotspot detection?; Beyond SUV, LI, and ITLV, are there plans to incorporate other advanced quantitative metrics or radiomic features to further enhance the characterization of PSMA-avid lesions?