Avenda Health AI Prostate Cancer Planning Software

The Avenda Health AI Prostate Cancer Planning Software aims to improve prostate cancer management by using artificial intelligence to assist physicians in accurately segmenting prostate MR images, characterizing lesions, and pre-planning interventions. Its methodology relies on AI analysis of multimodal patient data to generate probabilistic cancer maps and an 'Encapsulation Confidence Score' for lesion boundaries, which has demonstrated superior lesion encapsulation and tissue sparing in reader studies compared to standard of care.

The Avenda Health AI Prostate Cancer Planning Software is an artificial intelligence (AI)-based decision support tool designed to assist physicians in the prostate oncological workflow. Its primary purpose is to aid in the review of magnetic resonance (MR) prostate images and biopsy findings by facilitating the segmentation of MR image features, including the prostate gland itself, evaluating and quantifying lesions, documenting findings, and pre-planning for diagnostic and interventional procedures like biopsies or soft tissue ablation. The software utilizes AI algorithms to process prostate MR images, biopsy results, pathology reports, and clinical data to create a 'cancer map.' This map assigns a probability to each voxel within the prostate, indicating its likelihood of containing clinically significant prostate cancer (csPCa), defined as Gleason Grade Group (GGG) ≥ 2. Users can then select a threshold on this map to define a lesion boundary, which is subsequently assigned an 'Encapsulation Confidence Score' derived from a lookup table based on a database of cases with known ground-truth. The software also functions as a medical image application for viewing, manipulating, and 3D-visualizing MR prostate images. A multi-reader, multi-case (MRMC) study demonstrated the software's efficacy, showing that lesion contours produced with its assistance had superior sensitivity (mean 97.4% vs 38.2% for standard of care) and superior specificity compared to hemi-gland contours (mean 72.1% vs 53.4%). The software also led to superior balanced accuracy and 'clinical quality' scores. Notably, readers achieved a complete csPCa encapsulation rate of 72.8% using the device, compared to only 1.6% with standard methods. Standalone performance testing validated the prostate segmentation algorithm on 137 patients and the lesion contouring algorithm on 50 patients against clinically valid ground truths. Software verification, validation, and usability testing confirmed its performance according to design specifications, indicating it can help users create lesion contours that better encapsulate csPCa while sparing non-cancerous tissue, addressing the underestimation common with manual methods.

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 'Encapsulation Confidence Score' quantitatively correlate with long-term patient outcomes or recurrence rates post-intervention when using the planned treatment?; What specific AI architectural models are employed for the prostate segmentation and lesion contouring tools, and how were they trained and validated against diverse patient populations, imaging protocols, and MRI scanner vendors/field strengths?; How does the software's performance in defining the 'extent of known disease' compare when biopsy coordinate information is sparse or less precise?; Beyond improved contouring accuracy, what is the quantifiable impact of the simulated interventional tool placement feature on actual procedure times, precision, and oncological outcomes in a real-world clinical setting?; Are there mechanisms within the software to update its AI models and lookup tables as new ground-truth data becomes available, and how is this validated post-update?