Dataset Information

Assessment of deep learning-based auto-contouring on interobserver consistency in target volume and organs-at-risk delineation for breast cancer: Implications for RTQA program in a multi-institutional study.

ABSTRACT:

Purpose

To quantify interobserver variation (IOV) in target volume and organs-at-risk (OAR) contouring across 31 institutions in breast cancer cases and to explore the clinical utility of deep learning (DL)-based auto-contouring in reducing potential IOV.

Methods and materials

In phase 1, two breast cancer cases were randomly selected and distributed to multiple institutions for contouring six clinical target volumes (CTVs) and eight OAR. In Phase 2, auto-contour sets were generated using a previously published DL Breast segmentation model and were made available for all participants. The difference in IOV of submitted contours in phases 1 and 2 was investigated quantitatively using the Dice similarity coefficient (DSC) and Hausdorff distance (HD). The qualitative analysis involved using contour heat maps to visualize the extent and location of these variations and the required modification.

Results

Over 800 pairwise comparisons were analysed for each structure in each case. Quantitative phase 2 metrics showed significant improvement in the mean DSC (from 0.69 to 0.77) and HD (from 34.9 to 17.9 mm). Quantitative analysis showed increased interobserver agreement in phase 2, specifically for CTV structures (5-19 %), leading to fewer manual adjustments. Underlying IOV differences causes were reported using a questionnaire and hierarchical clustering analysis based on the volume of CTVs.

Conclusion

DL-based auto-contours improved the contour agreement for OARs and CTVs significantly, both qualitatively and quantitatively, suggesting its potential role in minimizing radiation therapy protocol deviation.

SUBMITTER: Choi MS

PROVIDER: S-EPMC10700624 | biostudies-literature | 2023 Nov

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Assessment of deep learning-based auto-contouring on interobserver consistency in target volume and organs-at-risk delineation for breast cancer: Implications for RTQA program in a multi-institutional study.

Choi Min Seo MS Chang Jee Suk JS Kim Kyubo K Kim Jin Hee JH Kim Tae Hyung TH Kim Sungmin S Cha Hyejung H Cho Oyeon O Choi Jin Hwa JH Kim Myungsoo M Kim Juree J Kim Tae Gyu TG Yeo Seung-Gu SG Chang Ah Ram AR Ahn Sung-Ja SJ Choi Jinhyun J Kang Ki Mun KM Kwon Jeanny J Koo Taeryool T Kim Mi Young MY Choi Seo Hee SH Jeong Bae Kwon BK Jang Bum-Sup BS Jo In Young IY Lee Hyebin H Kim Nalee N Park Hae Jin HJ Im Jung Ho JH Lee Sea-Won SW Cho Yeona Y Lee Sun Young SY Chang Ji Hyun JH Chun Jaehee J Lee Eung Man EM Kim Jin Sung JS Shin Kyung Hwan KH Kim Yong Bae YB

Breast (Edinburgh, Scotland) 20231115

<h4>Purpose</h4>To quantify interobserver variation (IOV) in target volume and organs-at-risk (OAR) contouring across 31 institutions in breast cancer cases and to explore the clinical utility of deep learning (DL)-based auto-contouring in reducing potential IOV.<h4>Methods and materials</h4>In phase 1, two breast cancer cases were randomly selected and distributed to multiple institutions for contouring six clinical target volumes (CTVs) and eight OAR. In Phase 2, auto-contour sets were generat ...[more]

PMID: 37992527

Similar Datasets

Project description:BackgroundThe delineation of pancreatic tumors on CT is challenging. In this study, we quantified the interobserver variation for pancreatic tumor delineation on 3DCT as well as on 4DCT.MethodsEight observers (radiation oncologists) from six institutions delineated pancreatic tumors of four patients with (borderline) resectable pancreatic cancer. The study consisted of two stages. In the 3DCT-stage, the gross tumor volume (GTV) was delineated on a contrast-enhanced scan. In the 4DCT-stage, the internal GTV (iGTV) was delineated, accounting for the respiratory motion. We calculated the volumes of the (i)GTV, the overlap of the delineated volumes (expressed as generalized conformity index: CIgen), the local observer variation (local standard deviation: SD) and the overall observer variation (overall SD). We compared these results between GTVs and iGTVs. Additionally, observers were asked to fill out a questionnaire concerning the difficulty of the delineation and their experience in delineating pancreatic tumors.ResultsThe ratios of the largest to the smallest delineated GTV and iGTV within the same patient were 6.8 and 16.5, respectively. As the iGTV incorporates the GTV during all respiratory phases, the mean volumes of the iGTV (40.07 cm3) were larger than those of the GTV (29.91 cm3). For all patients, CIgen was larger for the iGTV than for the GTV. The mean overall observer variation (root-mean-square of all local SDs over four patients) was 0.63 cm and 0.80 cm for GTV and iGTV, respectively. The largest local observer variations were seen close to biliary stents and suspicious pathological enlarged lymph nodes, as some observers included them and some did not. This variation was more pronounced for the iGTV than for the GTV. The observers rated the 3DCT-stage and 4DCT-stage equally difficult and treated on average three to four pancreatic cancer patients per year.ConclusionsA considerable interobserver variation in delineation of pancreatic tumors was observed. This variation was larger for 4D than for 3D delineation. The largest local observer variation was found around biliary stents and suspicious pathological enlarged lymph nodes.

Dataset Information

Assessment of deep learning-based auto-contouring on interobserver consistency in target volume and organs-at-risk delineation for breast cancer: Implications for RTQA program in a multi-institutional study.

Purpose

Methods and materials

Results

Conclusion

Publications

Assessment of deep learning-based auto-contouring on interobserver consistency in target volume and organs-at-risk delineation for breast cancer: Implications for RTQA program in a multi-institutional study.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets