A Deep Learning Approach to Digitally Stain Optical Coherence Tomography Images of the Optic Nerve Head

Devalla, Sripad Krishna; Chin, Khai Sing; Mari, Jean-Martial; Tun, Tin A.; Strouthidis, Nicholas G.; Aung, Tin; Thiéry, Alexandre H.; Girard, M

doi:10.1167/iovs.17-22617

Cited by 90 publications

(50 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…We observed that DRUNET offered no significant differences in the performance of digital staining when tested upon compensated (blood vessel shadows removed), or uncompensated images, as opposed to our previous patch-based method [23], that performed better on compensated images. This may be attributed to the extensive online data augmentation we used herein that also included occluding patches to mimic the presence of blood vessel shadows.…”

Section: Discussioncontrasting

confidence: 64%

See 1 more Smart Citation

DRUNET: a dilated-residual U-Net deep learning network to segment optic nerve head tissues in optical coherence tomography images

Devalla

Renukanand

Sreedhar

et al. 2018

Biomed. Opt. Express

Self Cite

170

110

View full text Add to dashboard Cite

Given that the neural and connective tissues of the optic nerve head (ONH) exhibit complex morphological changes with the development and progression of glaucoma, their simultaneous isolation from optical coherence tomography (OCT) images may be of great interest for the clinical diagnosis and management of this pathology. A deep learning algorithm (custom U-NET) was designed and trained to segment 6 ONH tissue layers by capturing both the local (tissue texture) and contextual information (spatial arrangement of tissues). The overall Dice coefficient (mean of all tissues) was 0.91 ± 0.05 when assessed against manual segmentations performed by an expert observer. Further, we automatically extracted six clinically relevant neural and connective tissue structural parameters from the segmented tissues. We offer here a robust segmentation framework that could also be extended to the 3D segmentation of the ONH tissues.

show abstract

Section: Discussioncontrasting

confidence: 64%

“…In our previous study, we developed a histogram-based [22] approach that was able to digitally stain (isolate) the connective and neural tissues of the ONH. Following this, we proposed a more accurate deep-learning (patch-based) [23] approach to isolate these tissues.…”

Section: Introductionmentioning

confidence: 99%

DRUNET: a dilated-residual U-Net deep learning network to segment optic nerve head tissues in optical coherence tomography images

Devalla

Renukanand

Sreedhar

et al. 2018

Biomed. Opt. Express

Self Cite

170

110

View full text Add to dashboard Cite

show abstract

“…Previous applications of deep learning in glaucoma have been limited to classification rather than forecasting and included analysis of disc photos, 17,19,28 OCT images 18,29 and VFs. 15 Recently, Li et al used a deep learning system to classify over 48,000 optic disc photos for referable glaucomatous optic neuropathy and reported an area under the receiver operating characteristic curve (AUROC) of 0.986 with a 95.6% sensitivity and 92.0% specificity.…”

Section: Discussionmentioning

confidence: 99%

Forecasting Future Humphrey Visual Fields Using Deep Learning

Wen

Lee

Keane

et al. 2018

Preprint

View full text Add to dashboard Cite

Purpose: To determine if deep learning networks could be trained to forecast a future 24-2 Humphrey Visual Field (HVF).Design: Retrospective database study.Participants: All patients who obtained a HVF 24-2 at the University of Washington.Methods: All datapoints from consecutive 24-2 HVFs from 1998 to 2018 were extracted from a University of Washington database. Ten-fold cross validation with a held out test set was used to develop the three main phases of model development: model architecture selection, dataset combination selection, and time-interval model training with transfer learning, to train a deep learning artificial neural network capable of generating a point-wise visual field prediction. Main outcome measures: Mean absolute error (MAE) and difference in Mean Deviation (MD) between predicted and actual future HVF.Results: More than 1.7 million perimetry points were extracted to the hundredth decibel from 32,443 24-2 HVFs. The best performing model with 20 million trainable parameters, CascadeNet-5, was selected. The overall MAE for the test set was 2.47 dB (95% CI: 2.45 dB to 2.48 dB). The 100 fully trained models were able to successfully predict progressive field loss in glaucomatous eyes up to 5.5 years in the future with a correlation of 0.92 between the MD of predicted and actual future HVF (p < 2.2 x 10 -16 ) and an average difference of 0.41 dB.Conclusions: Using unfiltered real-world datasets, deep learning networks show an impressive ability to not only learn spatio-temporal HVF changes but also to generate predictions for future HVFs up to 5.5 years, given only a single HVF.

show abstract

“…Some recent reports have attempted to evaluate the structures in the optic nerve head using machine learning or artificial intelligence methods, in which various test set to validation set approaches have identified some aspects of structure with minimal human intervention. 23 As yet, these methods do not identify the position of the FIGURE 8. Bland-Altman Plot comparing observer and algorithm in lamina position.…”

Section: Discussionmentioning

confidence: 99%

Evaluation of Automated Segmentation Algorithms for Optic Nerve Head Structures in Optical Coherence Tomography Images

Duan

Jefferys

Quigley

2018

Invest. Ophthalmol. Vis. Sci.

View full text Add to dashboard Cite

PURPOSE. To compare the identification of optic nerve head (ONH) structures in optical coherence tomography images by observers and automated algorithms.METHODS. ONH images in 24 radial scan sets by optical coherence tomography were obtained in 51 eyes of 29 glaucoma patients and suspects. Masked intraobserver and interobserver comparisons were made of marked endpoints of Bruch's membrane opening (BMO) and the anterior lamina cribrosa (LC). BMO and LC positional markings were compared between observer and automated algorithm. Repeated analysis on 20 eyes by the algorithm was compared. Regional ONH data were derived from the algorithms. RESULTS.Intraobserver difference in BMO width was not significantly different from zero (P ‡ 0.32) and the difference in LC position was less than 1% different (P ¼ 0.04). Interobserver were slightly larger than intraobserver differences, but interobserver BMO width difference was 0.36% (P ¼ 0.63). Mean interobserver difference in LC position was 14.74 lm (P ¼ 0.004), 3% of the typical anterior lamina depth (ALD). Between observer and algorithm, BMO width differed by 1.85% (P ¼ 0.23) and mean LC position was not significantly different (3.77 lm, P ¼ 0.77). Repeat algorithmic analysis had a mean difference in BMO area of 0.38% (P ¼ 0.47) and mean ALD difference of 0.54 6 0.72%. Regional ALD had greater variability in the horizontal ONH regions. Some individual outlier images were not validly marked by either observers or algorithm.CONCLUSIONS. Automated identification of ONH structures is comparable to observer markings for BMO and anterior LC position, making BMO a practical reference plane for algorithmic analysis.

show abstract

A Deep Learning Approach to Digitally Stain Optical Coherence Tomography Images of the Optic Nerve Head

Abstract: Our deep learning algorithm can simultaneously stain the neural and connective tissues of the ONH, offering a framework to automatically measure multiple key structural parameters of the ONH that may be critical to improve glaucoma management.

Cited by 90 publications

References 54 publications

DRUNET: a dilated-residual U-Net deep learning network to segment optic nerve head tissues in optical coherence tomography images

DRUNET: a dilated-residual U-Net deep learning network to segment optic nerve head tissues in optical coherence tomography images

Forecasting Future Humphrey Visual Fields Using Deep Learning

Evaluation of Automated Segmentation Algorithms for Optic Nerve Head Structures in Optical Coherence Tomography Images

Contact Info

Product

Resources

About