A dual-layer quality assurance approach leveraging dose prediction for efficient review of automated contours of organs at risk in the brain in radiotherapy

Sep 29, 2026

Uploaded by Adam

Background and Purpose: Despite widespread adoption of automatic segmentation (AS), manual review and adjustment of generated contours remain essential and time-consuming, with clinically relevant corrections still difficult to identify. Inter-observer variability and the risk of missing significant errors further complicate routine radiotherapy workflows. This study aims to identify critical segmentation errors while reducing unnecessary manual review, enabling efficient AS integration. Materials and Methods: A contour evaluation assistant used geometric Dice similarity and Hausdorff distance, combined with dose prediction to estimate clinical relevance and guide human review.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages8

Reading Time20 min

AccessFree to download

Contents lists

Physics and Imaging in Radiation Oncology

Article information

Article

Introduction

Materials and Methods

Results

Conclusions

Physics and Imaging in Radiation Oncology 36 (2025) 100888 Contents lists available at ScienceDirect Physics and Imaging in Radiation Oncology journal [homepage:](homepage: www.sciencedirect.com/journal/physics-and-imaging-in-radiation-oncology)[ www.sciencedirect.com/journal/physics-and-imaging-in-radiation-oncology](homepage: www.sciencedirect.com/journal/physics-and-imaging-in-radiation-oncology) | Original Research Article
A dual-layer quality assurance approach leveraging dose prediction for efficient review of automated contours of organs at risk in the brainin radiotherapy | | | | | --- | --- | --- | --- | | Robert Poela,*, Amith Kamathb, Ekin Ermis¸ a, Jonas Willmannc,d, Elias Rüfenacht b, Nicolaus Andratschkec, Peter Mansera,e, Daniel M. Aebersold a, Mauricio Reyes a,b
a Department of Radiation Oncology, Inselspital, Bern University Hospital, and University of Bern, Bern, Switzerland b ARTORG Center for Biomedical Research, University of Bern, Bern, Switzerland
c Department of Radiation Oncology, University Hospital Zurich, University of Zurich, Switzerland d Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, USA e Division of Medical Radiation Physics, Bern University Hospital, and University of Bern, Bern, Switzerland | | | | | A R T I C L E I N F O | | A B S T R A C T | | | Keywords:
Autosegmentation
Quality assurance Dose prediction OARs
Brain | | Background and Purpose: Despite widespread adaptation of automatic segmentation (AS), manual review and adjustment of generated contours are still essential. This process is time-consuming and identifying clinically relevant corrections remains challenging. Inter-observer variability and the risk of overlooking significant errors further complicate the workflow. A dedicated quality assurance tool is highly relevant to assure quality and speed up the manual review task. The primary aim of this work is to identify critical segmentation errors while reducing unnecessary manual review, enabling efficient integration of AS into routine radiotherapy.
Materials and Methods: We developed an evaluation assistant that assesses contour quality through the geometric measures Dice similarity coefficient and the Hausdorff distance. This was combined with a dose prediction model to determine the clinical relevance. The system was validated on 30 glioblastoma cases with ground truth and manually modified organ at risk (OAR) contours. A traffic light decision matrix classified contours based on geometric and dose parameters, flagging structures for human review.
Results: Out of 507 analyzed OARs, 180 were classified as critical. Our approach identified 173 of these critical structures (sensitivity: 0.96, specificity: 0.55). The system flagged 317 organs (61%) as critical, effectively ruling out 39% as non-critical with only 7 false negatives comprising structures.
Conclusions: Our dual-layer QA approach effectively identifies critical OAR segmentations with high sensitivity and acceptable specificity, potentially reducing manual review requirements significantly. By focusing on clinically relevant dose/volume metric endpoints, this method assures the quality of brain AS results in clinical radiotherapy practice. | | 1. Introduction In the radiotherapy workflow, accurately defining target volumesand organs at risk (OARs) is a critical step. Delineating the volume to be irradiated while sparing surrounding healthy tissues is essential for optimizing tumor control and minimizing toxicity. Manual contouring is time-consuming [1,2] and exhibits high inter-observer variability depending on user experience, interpretation and image quality [3,4]. Consequently, inter-observer variability remains a persistent issue in radiotherapy, with inconsistencies that may result in clinically significant errors [5,6], especially in advanced high-precision radiotherapy modalities [7–11]. Notably, human errors in contouring are

FAQs

Why is manual review still needed after automatic segmentation in radiotherapy?

Manual review and adjustment remain necessary because automatic segmentation can produce errors that are not fully preventable, and clinically relevant corrections are hard to identify. Inter-observer variability and the risk of overlooking significant errors further increase the need for verification.

How does the proposed dual-layer QA approach evaluate contour quality?

It combines geometric measures (Dice similarity coefficient and Hausdorff distance) with a dose prediction model to assess the clinical relevance of contour differences. A traffic light decision matrix uses geometric and dose parameters to flag structures for human review.

What were the main performance outcomes of the approach?

Among 507 analyzed organs at risk (OARs), 180 were classified as critical. The method identified 173 critical structures, achieving sensitivity of 0.96 and specificity of 0.55, while flagging 317 organs (61%) as critical and ruling out 39% as non-critical with only 7 false negatives.

Sep 29, 2026

Uploaded by Adam