Sub-category classifiers for multiple-instance learning and its application to retinal nerve fiber layer visibility classification

Siyamalan Manivannan (Lead / Corresponding author), Caroline Cobb, Stephen Burgess, Emanuele Trucco

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Citations (Scopus)
14 Downloads (Pure)

Abstract

We propose a novel multiple instance learning method to assess the visibility (visible/not visible) of the retinal nerve fiber layer (RNFL) in fundus camera images. Using only image-level labels,our approach learns to classify the images as well as to localize the RNFL visible regions. We transform the original feature space to a discriminative subspace,and learn a region-level classifier in that subspace. We propose a margin-based loss function to jointly learn this subspace and the region-level classifier. Experiments with a RNFL dataset containing 576 images annotated by two experienced ophthalmologists give an agreement (kappa values) of 0.65 and 0.58 respectively,with an inter-annotator agreement of 0.62. Note that our system gives higher agreements with the more experienced annotator. Comparative tests with three public datasets (MESSIDOR and DR for diabetic retinopathy,UCSB for breast cancer) show improved performance over the state-of-the-art.

Original languageEnglish
Title of host publicationMedical Image Computing and Computer-Assisted Intervention - MICCAI 2016 - 19th International Conference, Proceedings
PublisherSpringer Verlag
Pages308-316
Number of pages9
Volume9901 LNCS
ISBN (Print)9783319467221
DOIs
Publication statusPublished - 17 Oct 2016

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9901 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Fingerprint

Nerve
Visibility
Classifiers
Classifier
Fiber
Fibers
Subspace
Labels
Loss Function
Feature Space
Cameras
Breast Cancer
Margin
Camera
Classify
Transform
Learning
Experiments
Experiment

Cite this

Manivannan, S., Cobb, C., Burgess, S., & Trucco, E. (2016). Sub-category classifiers for multiple-instance learning and its application to retinal nerve fiber layer visibility classification. In Medical Image Computing and Computer-Assisted Intervention - MICCAI 2016 - 19th International Conference, Proceedings (Vol. 9901 LNCS, pp. 308-316). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9901 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-46723-8_36
Manivannan, Siyamalan ; Cobb, Caroline ; Burgess, Stephen ; Trucco, Emanuele. / Sub-category classifiers for multiple-instance learning and its application to retinal nerve fiber layer visibility classification. Medical Image Computing and Computer-Assisted Intervention - MICCAI 2016 - 19th International Conference, Proceedings. Vol. 9901 LNCS Springer Verlag, 2016. pp. 308-316 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{83c9161de2a249f29ed8714521a296e0,
title = "Sub-category classifiers for multiple-instance learning and its application to retinal nerve fiber layer visibility classification",
abstract = "We propose a novel multiple instance learning method to assess the visibility (visible/not visible) of the retinal nerve fiber layer (RNFL) in fundus camera images. Using only image-level labels,our approach learns to classify the images as well as to localize the RNFL visible regions. We transform the original feature space to a discriminative subspace,and learn a region-level classifier in that subspace. We propose a margin-based loss function to jointly learn this subspace and the region-level classifier. Experiments with a RNFL dataset containing 576 images annotated by two experienced ophthalmologists give an agreement (kappa values) of 0.65 and 0.58 respectively,with an inter-annotator agreement of 0.62. Note that our system gives higher agreements with the more experienced annotator. Comparative tests with three public datasets (MESSIDOR and DR for diabetic retinopathy,UCSB for breast cancer) show improved performance over the state-of-the-art.",
author = "Siyamalan Manivannan and Caroline Cobb and Stephen Burgess and Emanuele Trucco",
note = "EPSRC grant EP/M005976/1.",
year = "2016",
month = "10",
day = "17",
doi = "10.1007/978-3-319-46723-8_36",
language = "English",
isbn = "9783319467221",
volume = "9901 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "308--316",
booktitle = "Medical Image Computing and Computer-Assisted Intervention - MICCAI 2016 - 19th International Conference, Proceedings",
address = "Germany",

}

Manivannan, S, Cobb, C, Burgess, S & Trucco, E 2016, Sub-category classifiers for multiple-instance learning and its application to retinal nerve fiber layer visibility classification. in Medical Image Computing and Computer-Assisted Intervention - MICCAI 2016 - 19th International Conference, Proceedings. vol. 9901 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9901 LNCS, Springer Verlag, pp. 308-316. https://doi.org/10.1007/978-3-319-46723-8_36

Sub-category classifiers for multiple-instance learning and its application to retinal nerve fiber layer visibility classification. / Manivannan, Siyamalan (Lead / Corresponding author); Cobb, Caroline; Burgess, Stephen; Trucco, Emanuele.

Medical Image Computing and Computer-Assisted Intervention - MICCAI 2016 - 19th International Conference, Proceedings. Vol. 9901 LNCS Springer Verlag, 2016. p. 308-316 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9901 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Sub-category classifiers for multiple-instance learning and its application to retinal nerve fiber layer visibility classification

AU - Manivannan, Siyamalan

AU - Cobb, Caroline

AU - Burgess, Stephen

AU - Trucco, Emanuele

N1 - EPSRC grant EP/M005976/1.

PY - 2016/10/17

Y1 - 2016/10/17

N2 - We propose a novel multiple instance learning method to assess the visibility (visible/not visible) of the retinal nerve fiber layer (RNFL) in fundus camera images. Using only image-level labels,our approach learns to classify the images as well as to localize the RNFL visible regions. We transform the original feature space to a discriminative subspace,and learn a region-level classifier in that subspace. We propose a margin-based loss function to jointly learn this subspace and the region-level classifier. Experiments with a RNFL dataset containing 576 images annotated by two experienced ophthalmologists give an agreement (kappa values) of 0.65 and 0.58 respectively,with an inter-annotator agreement of 0.62. Note that our system gives higher agreements with the more experienced annotator. Comparative tests with three public datasets (MESSIDOR and DR for diabetic retinopathy,UCSB for breast cancer) show improved performance over the state-of-the-art.

AB - We propose a novel multiple instance learning method to assess the visibility (visible/not visible) of the retinal nerve fiber layer (RNFL) in fundus camera images. Using only image-level labels,our approach learns to classify the images as well as to localize the RNFL visible regions. We transform the original feature space to a discriminative subspace,and learn a region-level classifier in that subspace. We propose a margin-based loss function to jointly learn this subspace and the region-level classifier. Experiments with a RNFL dataset containing 576 images annotated by two experienced ophthalmologists give an agreement (kappa values) of 0.65 and 0.58 respectively,with an inter-annotator agreement of 0.62. Note that our system gives higher agreements with the more experienced annotator. Comparative tests with three public datasets (MESSIDOR and DR for diabetic retinopathy,UCSB for breast cancer) show improved performance over the state-of-the-art.

UR - http://www.scopus.com/inward/record.url?scp=84996490513&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-46723-8_36

DO - 10.1007/978-3-319-46723-8_36

M3 - Conference contribution

AN - SCOPUS:84996490513

SN - 9783319467221

VL - 9901 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 308

EP - 316

BT - Medical Image Computing and Computer-Assisted Intervention - MICCAI 2016 - 19th International Conference, Proceedings

PB - Springer Verlag

ER -

Manivannan S, Cobb C, Burgess S, Trucco E. Sub-category classifiers for multiple-instance learning and its application to retinal nerve fiber layer visibility classification. In Medical Image Computing and Computer-Assisted Intervention - MICCAI 2016 - 19th International Conference, Proceedings. Vol. 9901 LNCS. Springer Verlag. 2016. p. 308-316. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-46723-8_36