Effective semantic segmentation in Cataract Surgery: What matters most?

13 Aug 2021  ·  Theodoros Pissas, Claudio Ravasio, Lyndon Da Cruz, Christos Bergeles ·

Our work proposes neural network design choices that set the state-of-the-art on a challenging public benchmark on cataract surgery, CaDIS. Our methodology achieves strong performance across three semantic segmentation tasks with increasingly granular surgical tool class sets by effectively handling class imbalance, an inherent challenge in any surgical video. We consider and evaluate two conceptually simple data oversampling methods as well as different loss functions. We show significant performance gains across network architectures and tasks especially on the rarest tool classes, thereby presenting an approach for achieving high performance when imbalanced granular datasets are considered. Our code and trained models are available at https://github.com/RViMLab/MICCAI2021_Cataract_semantic_segmentation and qualitative results on unseen surgical video can be found at https://youtu.be/twVIPUj1WZM.

PDF Abstract

Datasets

CaDIS
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
2D Semantic Segmentation task 1 (8 classes) CaDIS OCR-R50-Repeat Factor-Lovasz Mean IoU (test) 86.4 # 1
Compare
Mean IoU (val) 88.97 # 1
Compare
2D Semantic Segmentation task 3 (25 classes) CaDIS UPN-R50-Repeat Factor-Lovasz Mean IoU (test) 70.44 # 3
Compare
Mean IoU (val) 75.75 # 3
Compare
2D Semantic Segmentation task 2 (17 classes) CaDIS UPN-R50-Repeat Factor-Lovasz Mean IoU (test) 76.88 # 2
Compare
Mean IoU (val) 81.54 # 3
Compare
2D Semantic Segmentation task 1 (8 classes) CaDIS UPN-R50-Repeat Factor-Lovasz Mean IoU (test) 86.32 # 2
Compare
Mean IoU (val) 88.85 # 2
Compare
2D Semantic Segmentation task 3 (25 classes) CaDIS DeepLabv3+-R50-Repeat Factor-Lovasz Mean IoU (test) 70.51 # 2
Compare
Mean IoU (val) 77.44 # 2
Compare
2D Semantic Segmentation task 2 (17 classes) CaDIS DeepLabv3+-R50-Repeat Factor-Lovasz Mean IoU (test) 75.17 # 4
Compare
Mean IoU (val) 79.14 # 5
Compare
2D Semantic Segmentation task 1 (8 classes) CaDIS DeepLabv3+-R50-Repeat Factor-Lovasz Mean IoU (test) 85.65 # 3
Compare
Mean IoU (val) 88.48 # 3
Compare
2D Semantic Segmentation task 3 (25 classes) CaDIS OCR-R50-Repeat Factor-Lovasz Mean IoU (test) 71.94 # 1
Compare
Mean IoU (val) 79.4 # 1
Compare
2D Semantic Segmentation task 2 (17 classes) CaDIS OCR-R50-Repeat Factor-Lovasz Mean IoU (test) 79.09 # 1
Compare
Mean IoU (val) 83.25 # 1
Compare

Methods
Add Remove

No methods listed for this paper. Add relevant methods here
Contact us on: hello@paperswithcode.com . Papers With Code is a free resource with all data licensed under CC-BY-SA.
Terms Data policy Cookies policy