A Benchmark for Endoluminal Scene Segmentation of Colonoscopy Images

By David Vázquez, Jorge Bernal, F. Sánchez, G. Fernández-Esparrach, Antonio M. López, Adriana Romero, Michal Drozdzal, Aaron C. Courville. Journal of Healthcare Engineering, 2016

921 citations recorded by the source corpus.

Colorectal cancer (CRC) is the third cause of cancer death worldwide. Currently, the standard approach to reduce CRC-related mortality is to perform regular screening in search for polyps and colonoscopy is the screening tool of choice. The main limitations of this screening procedure are polyp miss rate and the inability to perform visual assessment of polyp malignancy. These drawbacks can be reduced by designing decision support systems (DSS) aiming to help clinicians in the different stages of the procedure by providing endoluminal scene segmentation. Thus, in this paper, we introduce an extended benchmark of colonoscopy image segmentation, with the hope of establishing a new strong benchmark for colonoscopy image analysis research. The proposed dataset consists of 4 relevant classes to inspect the endoluminal scene, targeting different clinical needs. Together with the dataset and taking advantage of advances in semantic segmentation literature, we provide new baselines by training standard fully convolutional networks (FCNs). We perform a comparative study to show that FCNs significantly outperform, without any further postprocessing, prior results in endoluminal scene segmentation, especially with respect to polyp segmentation and localization.

Subfields: Computer Vision and Pattern Recognition, Oncology

View Paper: https://www.semanticscholar.org/paper/8b008beb62007504aed4234980b5481edda644b0

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Oncology

Citations in the corpus

Segment Anything Model for Medical Image Segmentation: Current Applications and Future Directions (2024, background): “Zhou et al. [51] evaluate the performance of SAM with unprompted setting in segmenting polyps from colonoscopy images on five benchmark datasets [52, 53, 54, 55, 56].”
DS-TransUNet: Dual Swin Transformer U-Net for Medical Image Segmentation (2021, methodology): “Polyp Segmentation: For polyp segmentation task, we select five public polyp datasets including Kvasir [33], CVCColonDB (ColonDB) [34], EndoScene [35], ETIS [36], and CVC-ClinicDB (ClinicDB) [37].”
Automatic Polyp Segmentation via Multi-scale Subtraction Network (2021, methodology): “We evaluate the proposed model on five benchmark datasets: CVC-ColonDB [22], ETIS [21], Kvasir [11], CVC-T [24] and CVC-ClinicDB [2].”
UNCERTAINTY MODELING AND INTERPRETABILITY IN CONVOLUTIONAL NEURAL NETWORKS FOR POLYP SEGMENTATION (2018, methodology): “Missed polyps can have fatal consequences, as they are potential precursors to colon cancer, which causes the third most cancer deaths globally [2].”
EMCAD: Efficient Multi-Scale Convolutional Attention Decoding for Medical Image Segmentation (2024, methodology): “We resize images to 352 × 352 and use a multi-scale { 0.75, 1.0, 1.25 } training strategy with a gradient clip limit of 0.5 for ClinicDB [3], Kvasir [29], ColonDB [51], ETIS [51], BKAI [40], ISIC17 [15], and ISIC18 [15], while we resize images to 256 × 256 for BUSI [1], EM [6], and DSB18 [4].”
Shallow Attention Network for Polyp Segmentation (2021, methodology): “To evaluate the performance of the proposed SANet, ﬁve polyp segmentation datasets are adopted, including Kvasir [10], CVC-ClinicDB [2], CVC-ColonDB [3], EndoScene [18] and ETIS [16].”
TransFuse: Fusing Transformers and CNNs for Medical Image Segmentation (2021, methodology): “…the eﬀectiveness of TransFuse, four segmentation tasks with diﬀerent imaging modalities, disease types, target objects, target sizes, etc. are considered: 1) Polyp Segmentation , where ﬁve public polyp datasets are used: Kvasir [14], CVC-ClinicDB [2], CVC-ColonDB [24], EndoScene [27] and ETIS [21].”
Polyp-PVT: Polyp Segmentation with Pyramid Vision Transformers (2021, methodology): “Following the experimental setups in PraNet [5], we adopt ﬁve challenging public datasets, including Kvasir-SEG [13], ClinicDB [8], ColonDB [10], Endoscene [14] and ETIS [9] to verify the effectiveness of our framework.”
A Comprehensive Study on Colorectal Polyp Segmentation With ResUNet++, Conditional Random Field and Test-Time Augmentation (2021, background): “Lumen segmentation helps clinicians navigate through the colon during the examination, and it can be useful to establish a quality metric for the explored colon wall [9].”
Cross-level Feature Aggregation Network for Polyp Segmentation (2023, methodology): “• CVC-300 [50] : This dataset contains 60 polyp images with a size of 500 × 574 .”
CaraNet: context axial reverse attention network for segmentation of small medical objects (2021, methodology): “We test our CaraNet on five polyp segmentation datasets: ETIS [38], CVC-ClinicDB [39], CVC-ColonDB [40], EndoScene [41], and Kvasir [42].”
Transformers in Medical Image Analysis: A Review (2022, background): “Abbreviation: ACT-MOS, abdominal CT multi-organ segmentation; COPLE, COVID-19 Pneumonia Lesion segmentation; ACDC, Automated cardiac diagnosis challenge; MSD-01, Medical Segmentation Decathlon dataset Task01; KCCEE, Kvasir [63], CVC-ClinicDB [64], CVC-ColonDB [65], EndoScene [66] and ETIS [67].”
PraNet: Parallel Reverse Attention Network for Polyp Segmentation (2020, methodology): “Table 2: Quantitative results on CVC-ColonDB [22], ETIS [21], and test set (CVCT) of EndoScene [23] datasets.”
Adaptive Context Selection for Polyp Segmentation (2020, background): “The first is the EndoScene Dataset [17], which contains 912 images and each of which has at least one polyp region.”