[论文解读] 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 treatments被引用 0

一句话总结

Comp2Comp 提供开放源代码、FDA-510(k) 批准的 AI 管线用于机会性 CT 分析（AAQ 与 BMD），并在多家机构进行透明验证。

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.

研究动机与目标

• 通过提供开放源代码管线和完整关键性数据，解决 FDA 批准的机会性 CT 分析 AI 的透明度差距。
• 开发并验证两款 FDA 510(k) 批准模块（AAQ 与 BMD），用于基于 CT 的人体组成评估。
• 让医院和研究人员能够在开放代码和模型的条件下测试并采用最先进的 AI 方法。

提出的方法

• AAQ 管线使用 nnU-Net 对腹主动脉进行分割，并输出最大轴向直径及 QC 输出。
• BMD 管线使用定制的 nnU-Net 对 L1–L4 椎体进行分割，并通过放射密度归一化和对 DXA 推导的 T 分数的二值阈值来估算椎体海绵骨密度。
• 验证数据集包含多机构 CT 扫描，涵盖不同扫描仪和协议，以反映真实世界的变异性。
• 在手稿中对 FDA 备案的关键终点和统计分析完整披露，以最大化透明度。
• 该平台本身在 Apache License 2.0 下开源；AAQ 与 BMD 模块已获 FDA 批准，但 Comp2Comp 平台本身尚未获批。

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

实验结果

研究问题

• RQ1 AAQ 能否在不同扫描仪和患者解剖结构下准确测量 CT 上的腹主动脉最大直径？
• RQ2 BMD 是否能够从 CT 推断的椎体 ROI 中可靠区分低于/高于正常的椎体骨密度，与 DXA 相比如何？
• RQ3 模型在不同子组（年龄、性别、扫描仪制造商、内核、切片厚度）之间的性能指标有何变化？
• RQ4 将 FDA 批准的模块开源是否能改善本地验证和采用，与闭源解决方案相比有何差异？

主要发现

• AAQ 的平均绝对误差为 1.58 mm（95% 置信区间 1.375–1.797），放射科医师–模型 ICC 为 0.985（95% CI 0.979–0.989）。
• BMD 分类的敏感性 81.0%、特异性 78.4%（以 DXA T-score 阈值 -1.0；PPV 73.6%、NPV 84.8%）。
• 连续 BMD 得分与 DXA T-score 的相关系数 r = 0.791（95% CI 0.752–0.830）；连续分数的 AUROC = 0.883，二分类 AUROC = 0.797。
• AAQ 与 BMD 已获得 FDA 510(k) 批准（AAQ K243779 于 2025-07-01；BMD K242295 于 2025-04-08）。
• AAQ 在大多数子组中保持稳健的性能，但在内血管支架置入患者中的准确性下降（MAE 3.964 mm）。
• 开源的 Comp2Comp 仓库（Apache 2.0）实现了对这些 FDA 批准方法的测试与复现。

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