When Two Eyes Don’t Suffice—Learning Difficult Hyperfluorescence Segmentations in Retinal Fundus Autofluorescence Images via Ensemble Learning

Santarossa, Monty; Beyer, Tebbo Tassilo; Scharf, Amelie Bernadette Antonia; Tatli, Ayse; von der Burchard, Claus; Nazarenus, Jakob; Roider, Johann Baptist; Koch, Reinhard

doi:10.3390/jimaging10050116

Open AccessArticle

When Two Eyes Don’t Suffice—Learning Difficult Hyperfluorescence Segmentations in Retinal Fundus Autofluorescence Images via Ensemble Learning

by

Monty Santarossa

^1,*

,

Tebbo Tassilo Beyer

¹,

Amelie Bernadette Antonia Scharf

²,

Ayse Tatli

²,

Claus von der Burchard

²

,

Jakob Nazarenus

¹

,

Johann Baptist Roider

² and

Reinhard Koch

¹

Department of Computer Science, Kiel University, 24118 Kiel, Germany

²

Department of Ophthalmology, Kiel University, 24118 Kiel, Germany

^*

Author to whom correspondence should be addressed.

J. Imaging 2024, 10(5), 116; https://doi.org/10.3390/jimaging10050116

Submission received: 16 April 2024 / Revised: 3 May 2024 / Accepted: 6 May 2024 / Published: 9 May 2024

(This article belongs to the Special Issue Medical Image Classification and Segmentation: Progress and Challenges)

Download

Browse Figures

Versions Notes

Abstract

Hyperfluorescence (HF) and reduced autofluorescence (RA) are important biomarkers in fundus autofluorescence images (FAF) for the assessment of health of the retinal pigment epithelium (RPE), an important indicator of disease progression in geographic atrophy (GA) or central serous chorioretinopathy (CSCR). Autofluorescence images have been annotated by human raters, but distinguishing biomarkers (whether signals are increased or decreased) from the normal background proves challenging, with borders being particularly open to interpretation. Consequently, significant variations emerge among different graders, and even within the same grader during repeated annotations. Tests on in-house FAF data show that even highly skilled medical experts, despite previously discussing and settling on precise annotation guidelines, reach a pair-wise agreement measured in a Dice score of no more than 63–80% for HF segmentations and only 14–52% for RA. The data further show that the agreement of our primary annotation expert with herself is a 72% Dice score for HF and 51% for RA. Given these numbers, the task of automated HF and RA segmentation cannot simply be refined to the improvement in a segmentation score. Instead, we propose the use of a segmentation ensemble. Learning from images with a single annotation, the ensemble reaches expert-like performance with an agreement of a 64–81% Dice score for HF and 21–41% for RA with all our experts. In addition, utilizing the mean predictions of the ensemble networks and their variance, we devise ternary segmentations where FAF image areas are labeled either as confident background, confident HF, or potential HF, ensuring that predictions are reliable where they are confident (97% Precision), while detecting all instances of HF (99% Recall) annotated by all experts.

Keywords: CSCR; central serous chorioretinopathy; fundus autofluorescence; hyperfluorescence; reduced autofluorescence; inter-observer variability; intra-observer variability; ternary; segmentation; ensemble; deep learning; U-Net; image analysis; retinal; ambiguous; annotation

Share and Cite

MDPI and ACS Style

Santarossa, M.; Beyer, T.T.; Scharf, A.B.A.; Tatli, A.; von der Burchard, C.; Nazarenus, J.; Roider, J.B.; Koch, R. When Two Eyes Don’t Suffice—Learning Difficult Hyperfluorescence Segmentations in Retinal Fundus Autofluorescence Images via Ensemble Learning. J. Imaging 2024, 10, 116. https://doi.org/10.3390/jimaging10050116

AMA Style

Santarossa M, Beyer TT, Scharf ABA, Tatli A, von der Burchard C, Nazarenus J, Roider JB, Koch R. When Two Eyes Don’t Suffice—Learning Difficult Hyperfluorescence Segmentations in Retinal Fundus Autofluorescence Images via Ensemble Learning. Journal of Imaging. 2024; 10(5):116. https://doi.org/10.3390/jimaging10050116

Chicago/Turabian Style

Santarossa, Monty, Tebbo Tassilo Beyer, Amelie Bernadette Antonia Scharf, Ayse Tatli, Claus von der Burchard, Jakob Nazarenus, Johann Baptist Roider, and Reinhard Koch. 2024. "When Two Eyes Don’t Suffice—Learning Difficult Hyperfluorescence Segmentations in Retinal Fundus Autofluorescence Images via Ensemble Learning" Journal of Imaging 10, no. 5: 116. https://doi.org/10.3390/jimaging10050116

APA Style

Santarossa, M., Beyer, T. T., Scharf, A. B. A., Tatli, A., von der Burchard, C., Nazarenus, J., Roider, J. B., & Koch, R. (2024). When Two Eyes Don’t Suffice—Learning Difficult Hyperfluorescence Segmentations in Retinal Fundus Autofluorescence Images via Ensemble Learning. Journal of Imaging, 10(5), 116. https://doi.org/10.3390/jimaging10050116

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu

When Two Eyes Don’t Suffice—Learning Difficult Hyperfluorescence Segmentations in Retinal Fundus Autofluorescence Images via Ensemble Learning

Abstract

Share and Cite

Article Metrics

Article Access Statistics

Further Information

Guidelines

MDPI Initiatives

Follow MDPI