TY - GEN
T1 - Optimizing Multi-expert Consensus for Classification and Precise Localization of Barrett's Neoplasia
AU - Kusters, Koen
AU - Boers, Tim
AU - Jaspers, Tim J.M.
AU - Jong, Martijn R.
AU - van Eijck van Heslinga, Rixta A.H.
AU - de Groof, Albert J. (Jeroen)
AU - Bergman, Jacques J.G.H.M.
AU - van der Sommen, Fons
AU - de With, Peter H.N.
PY - 2024/10/9
Y1 - 2024/10/9
N2 - Recognition of early neoplasia in Barrett’s Esophagus (BE) is challenging, despite advances in endoscopic technology. Even with correct identification, the subtle nature of lesions leads to significant inter-observer variability in placing targeted biopsy markers and delineation of lesions. Computer-Aided Detection (CADe) systems may assist endoscopists, however, compliance of endoscopists with CADe is often suboptimal, reducing joint performance below CADe stand-alone performance. Improved localization performance of CADe could enhance compliance. These systems often use fused consensus ground-truths (GT), which may not capture subtle neoplasia gradations, affecting classification and localization. This study evaluates five consensus GT strategies from multi-expert segmentation labels and four loss functions for their impact on classification and localization performance. The dataset includes 7,995 non-dysplastic BE images (1,256 patients) and 2,947 neoplastic images (823 patients), with each neoplastic image annotated by two experts. Classification, localization for true positives, and combined detection performance are assessed and compared with 14 independent Barrett’s experts. Results show that using multiple consensus GT masks with a compound Binary Cross-Entropy and Dice loss achieves the best classification sensitivity and near-expert level localization, making it the most effective training strategy. The code is made publicly available at: https://github.com/BONS-AI-VCA-AMC/BE-CADe-GT.
AB - Recognition of early neoplasia in Barrett’s Esophagus (BE) is challenging, despite advances in endoscopic technology. Even with correct identification, the subtle nature of lesions leads to significant inter-observer variability in placing targeted biopsy markers and delineation of lesions. Computer-Aided Detection (CADe) systems may assist endoscopists, however, compliance of endoscopists with CADe is often suboptimal, reducing joint performance below CADe stand-alone performance. Improved localization performance of CADe could enhance compliance. These systems often use fused consensus ground-truths (GT), which may not capture subtle neoplasia gradations, affecting classification and localization. This study evaluates five consensus GT strategies from multi-expert segmentation labels and four loss functions for their impact on classification and localization performance. The dataset includes 7,995 non-dysplastic BE images (1,256 patients) and 2,947 neoplastic images (823 patients), with each neoplastic image annotated by two experts. Classification, localization for true positives, and combined detection performance are assessed and compared with 14 independent Barrett’s experts. Results show that using multiple consensus GT masks with a compound Binary Cross-Entropy and Dice loss achieves the best classification sensitivity and near-expert level localization, making it the most effective training strategy. The code is made publicly available at: https://github.com/BONS-AI-VCA-AMC/BE-CADe-GT.
KW - my_intconf
KW - Barrett’s Esophagus
KW - Detection
KW - Inter-observer variability
UR - http://www.scopus.com/inward/record.url?scp=85206984616&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-73376-5_8
DO - 10.1007/978-3-031-73376-5_8
M3 - Conference contribution
SN - 978-3-031-73375-8
T3 - Lecture Notes in Computer Science (LNCS)
SP - 83
EP - 92
BT - Cancer Prevention, Detection, and Intervention - 3rd MICCAI Workshop, CaPTion 2024, Held in Conjunction with MICCAI 2024, Proceedings
A2 - Ali, Sharib
A2 - van der Sommen, Fons
A2 - Papież, Bartłomiej Władysław
A2 - Ghatwary, Noha
A2 - Jin, Yueming
A2 - Kolenbrander, Iris
PB - Springer
CY - Cham
T2 - 3rd International Workshop on Cancer Prevention, detection and intervenTion - Satellite event at MICCAI 2024
Y2 - 6 October 2024 through 10 October 2024
ER -