← Back to all news
arXiv AI

Risk Control Certifies Rates, Not Magnitudes: An Audit of Selective Fetal Biometry

The paper audits a prenatal ultrasound biometry pipeline that predicts head circumference (HC) while certifying a maximum error rate rather than magnitude. Using two pre‑registered splits of 999 HC18 images, the authors retrain the segmenter three times per split and calibrate a contour‑completeness threshold via exact Clopper–Pearson risk control. Although the certified risk bound (α = 0.10) remains intact, the pipeline still produces a large 271.3 mm error on a single image, and tightening the admissibility region to an anatomically defined range reduces worst‑case errors to about 11 mm while only modestly affecting coverage.

Related stories

arXiv AI
Aug 20

Bound-Aware Per-Organ Recall Risk Control for Multi-Organ CT Segmentation under Clinical Domain Shift

The paper presents a distribution‑free risk‑control framework that provides organ‑specific recall guarantees for frozen multi‑organ CT segmentation models. It calibrates per‑organ thresholds for an AMOS‑trained nnU‑Net, audits its transfer to RAOS, and estimates local re‑certification costs using case‑level voxel false‑negative rates. The study compares Risk‑Controlling Prediction Sets (RCPS) and Conformal Risk Control (CRC), noting that RCPS offers high‑probability control of population‑mean risk while CRC provides weaker expectation control, and evaluates the effectiveness of the Waudby–Smith–Ramdas betting bound versus Hoeffding–Bentkus bounds for re‑certification. "whyItMatters":"The work demonstrates how to maintain organ‑level recall guarantees when deploying segmentation models across different clinical domains, highlighting the trade‑offs between threshold conservatism and re‑certification effort."

By Souraj Adhikary, Negar Chabi, Andre Mastmeyer
computer-vision
More like this →
Hugging Face Trending Papers
Aug 18

Bound-Aware Per-Organ Recall Risk Control for Multi-Organ CT Segmentation under Clinical Domain Shift

The paper introduces a distribution‑free risk control method that provides organ‑specific recall guarantees for frozen segmentation models. It calibrates per‑organ thresholds on an AMOS‑trained nnU‑Net, audits transfer to RAOS, and estimates local re‑certification cost using case‑level voxel false‑negative rates. The study compares Risk‑Controlling Prediction Sets (RCPS) and Conformal Risk Control (CRC), noting that RCPS offers high‑probability control of population‑mean risk while CRC provides weaker expectation control, and evaluates the effectiveness of the Waudby‑Smith‑Ramdas betting bound versus Hoeffding‑Bentkus bounds for re‑certification of Tier‑1 organs.

computer-vision
More like this →
arXiv AI
Sep 11

Safe to Stop? Risk-Constrained Stopping for Sequential Clinical Diagnosis Agents

The paper introduces Cros, a risk‑constrained stopping layer for sequential clinical diagnosis agents that determines when to stop testing and make a diagnosis. Cros combines state‑wise error ranking, policy design on disjoint development splits, and exact tests of selective diagnostic error to provide finite‑sample guarantees. On a MIMIC‑derived abdominal‑pain benchmark, Cros achieves higher state‑error AUROC and lower selective error rates compared to baseline stopping methods, though its performance varies across development resplits.

By Yuexin Wu, Vasile Rus
agentsbenchmarkssafety
More like this →
arXiv Computer Vision
Oct 16

A Decision-Aware Evaluation Protocol for Uncertainty in Medical Image Segmentation

The paper introduces a decision‑aware evaluation protocol for medical image segmentation uncertainty that separates uncertainty estimation from deferral policy and assesses their joint performance under fixed review budgets. It reports metrics such as fixed‑budget error reduction, risk‑coverage behavior, region‑level review proxies, structural failure modes, and sensitivity to review assumptions. Experiments on retinal vessel, skin lesion, and polyp segmentation with various backbones and uncertainty estimators show that estimator rankings vary with review unit, uncertainty target, and cost model, and that boundary and thin‑structure failures are partly due to annotation ambiguity.

By Saket Maganti, Shreyank N Gowda
computer-vision
More like this →