Imperial College London
Portrait Picture

Dr Ben Glocker

Faculty of Engineering, Department of Computing

Professor in Machine Learning for Imaging
 
 
 
//

Contact

 

+44 (0)20 7594 8334b.glocker Website CV

 
 
//

Location

 

377Huxley BuildingSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@article{Islam:2023:10.1007/s11548-023-02847-9,
author = {Islam, M and Seenivasan, L and Sharan, SP and Viekash, VK and Gupta, B and Glocker, B and Ren, H},
doi = {10.1007/s11548-023-02847-9},
journal = {International Journal of Computer Assisted Radiology and Surgery},
pages = {1875--1883},
title = {Paced-curriculum distillation with prediction and label uncertainty for image segmentation},
url = {http://dx.doi.org/10.1007/s11548-023-02847-9},
volume = {18},
year = {2023}
}
Download

RIS format (EndNote, RefMan)

TY  - JOUR
AB  - PURPOSE: In curriculum learning, the idea is to train on easier samples first and gradually increase the difficulty, while in self-paced learning, a pacing function defines the speed to adapt the training progress. While both methods heavily rely on the ability to score the difficulty of data samples, an optimal scoring function is still under exploration. METHODOLOGY: Distillation is a knowledge transfer approach where a teacher network guides a student network by feeding a sequence of random samples. We argue that guiding student networks with an efficient curriculum strategy can improve model generalization and robustness. For this purpose, we design an uncertainty-based paced curriculum learning in self-distillation for medical image segmentation. We fuse the prediction uncertainty and annotation boundary uncertainty to develop a novel paced-curriculum distillation (P-CD). We utilize the teacher model to obtain prediction uncertainty and spatially varying label smoothing with Gaussian kernel to generate segmentation boundary uncertainty from the annotation. We also investigate the robustness of our method by applying various types and severity of image perturbation and corruption. RESULTS: The proposed technique is validated on two medical datasets of breast ultrasound image segmentation and robot-assisted surgical scene segmentation and achieved significantly better performance in terms of segmentation and robustness. CONCLUSION: P-CD improves the performance and obtains better generalization and robustness over the dataset shift. While curriculum learning requires extensive tuning of hyper-parameters for pacing function, the level of performance improvement suppresses this limitation.
AU  - Islam,M
AU  - Seenivasan,L
AU  - Sharan,SP
AU  - Viekash,VK
AU  - Gupta,B
AU  - Glocker,B
AU  - Ren,H
DO  - 10.1007/s11548-023-02847-9
EP  - 1883
PY  - 2023///
SN  - 1861-6410
SP  - 1875
TI  - Paced-curriculum distillation with prediction and label uncertainty for image segmentation
T2  - International Journal of Computer Assisted Radiology and Surgery
UR  - http://dx.doi.org/10.1007/s11548-023-02847-9
UR  - https://www.ncbi.nlm.nih.gov/pubmed/36862365
UR  - https://link.springer.com/article/10.1007/s11548-023-02847-9
UR  - http://hdl.handle.net/10044/1/102896
VL  - 18
ER  -
Download