[Paper Review] Comp2Comp: Open-Source Software with FDA-Cleared Artificial Intelligence Algorithms for Computed Tomography Image Analysis

Adrit Rao, Malte Jensen+29 more|arXiv (Cornell University)|Feb 10, 2026

Aortic aneurysm repair treatments0 citations

TL;DR

Comp2Comp provides open-source, FDA-510(k)-cleared AI pipelines for opportunistic CT analysis (AAQ and BMD), with transparent validation across multiple institutions.

ABSTRACT

Artificial intelligence allows automatic extraction of imaging biomarkers from already-acquired radiologic images. This paradigm of opportunistic imaging adds value to medical imaging without additional imaging costs or patient radiation exposure. However, many open-source image analysis solutions lack rigorous validation while commercial solutions lack transparency, leading to unexpected failures when deployed. Here, we report development and validation for two of the first fully open-sourced, FDA-510(k)-cleared deep learning pipelines to mitigate both challenges: Abdominal Aortic Quantification (AAQ) and Bone Mineral Density (BMD) estimation are both offered within the Comp2Comp package for opportunistic analysis of computed tomography scans. AAQ segments the abdominal aorta to assess aneurysm size; BMD segments vertebral bodies to estimate trabecular bone density and osteoporosis risk. AAQ-derived maximal aortic diameters were compared against radiologist ground-truth measurements on 258 patient scans enriched for abdominal aortic aneurysms from four external institutions. BMD binary classifications (low vs. normal bone density) were compared against concurrent DXA scan ground truths obtained on 371 patient scans from four external institutions. AAQ had an overall mean absolute error of 1.57 mm (95% CI 1.38-1.80 mm). BMD had a sensitivity of 81.0% (95% CI 74.0-86.8%) and specificity of 78.4% (95% CI 72.3-83.7%). Comp2Comp AAQ and BMD demonstrated sufficient accuracy for clinical use. Open-sourcing these algorithms improves transparency of typically opaque FDA clearance processes, allows hospitals to test the algorithms before cumbersome clinical pilots, and provides researchers with best-in-class methods.

Motivation & Objective

• Address the transparency gap in FDA-cleared AI for opportunistic CT analysis by providing open-source pipelines and full pivotal data.
• Develop and validate two FDA 510(k)-cleared modules (AAQ and BMD) for CT-based body composition assessment.
• Enable hospitals and researchers to test and adopt best-in-class AI methods with open code and models.

Proposed method

• AAQ pipeline uses nnU-Net to segment the abdominal aorta and output the maximal axial diameter plus QC outputs.
• BMD pipeline uses a custom nnU-Net to segment L1–L4 vertebrae and estimate vertebral trabecular BMD via radiodensity normalization and a binary threshold against DXA-derived T-scores.
• Validation datasets comprise multi-institutional CT scans with diverse scanners and protocols to mirror real-world variation.
• Pivotal endpoints and statistical analyses from FDA filings are reported in full in the manuscript to maximize transparency.
• The platform itself is open source under Apache License 2.0; AAQ and BMD modules are FDA-cleared but the Comp2Comp platform is not itself cleared.

Figure 1: (A) Abdominal aortic quantification (AAQ) pipeline architecture from input DICOM files of CT scan to output of aortic maximum diameter with associated statistics and visualizations. (B) Study layout, consisting of data acquisition for AAQ training, integration within the AAQ pipeline, and

Experimental results

Research questions

• RQ1 Can AAQ accurately measure maximal abdominal aortic diameter on CT across diverse scanners and patient anatomies?
• RQ2 Can BMD reliably classify low vs normal vertebral bone mineral density from CT-derived vertebral ROIs compared to DXA?
• RQ3 How do model performance metrics vary across subgroups such as age, sex, scanner manufacturer, kernel, and slice thickness?
• RQ4 Does open-sourcing the FDA-cleared modules improve local validation and adoption compared with closed-source solutions?

Key findings

• AAQ achieved a mean absolute error of 1.58 mm (95% CI 1.375–1.797) and radiologist–model ICC of 0.985 (95% CI 0.979–0.989).
• BMD classification had 81.0% sensitivity and 78.4% specificity (DXA T-score threshold -1.0) with PPV 73.6% and NPV 84.8%.
• Continuous BMD score correlated with DXA T-score at r = 0.791 (95% CI 0.752–0.830); continuous score AUROC = 0.883 and binary AUROC = 0.797.
• AAQ and BMD received FDA 510(k) clearances (AAQ K243779 on 2025-07-01; BMD K242295 on 2025-04-08).
• AAQ performance remained robust across most subgroups but showed reduced accuracy in endograft patients (MAE 3.964 mm).
• The open-source Comp2Comp repository (Apache 2.0) enables testing and replication of these FDA-cleared methods.

Figure 2: Overview of the BMD Algorithm Development and Output. (A) Processing pipeline from DICOM series input to BMD classification (normal or low density). The pipeline segments the L1-L4 vertebrae and generates ROIs in the body of the vertebrae. The Hounsfield Unit (HU) values are then calibrate