Other affiliations: University of Western Brittany
Bio: Béatrice Cochener is an academic researcher from French Institute of Health and Medical Research. The author has contributed to research in topics: Cataract surgery & Intraocular lens. The author has an hindex of 39, co-authored 173 publications receiving 6319 citations. Previous affiliations of Béatrice Cochener include University of Western Brittany.
Papers published on a yearly basis
TL;DR: The feedback on the Messidor database, after more than 6 years of diffusion, is analyzed and the real interest and benefit of the research community is not easy to quantify.
Abstract: The Messidor database, which contains hundreds of eye fundus images, has been publicly distributed since 2008. It was created by the Messidor project in order to evaluate automatic lesion segmentation and diabetic retinopathy grading methods. Designing, producing and maintaining such a database entails significant costs. By publicly sharing it, one hopes to bring a valuable resource to the public research community. However, the real interest and benefit of the research community is not easy to quantify. We analyse here the feedback on the Messidor database, after more than 6 years of diffusion. This analysis should apply to other similar research databases.
TL;DR: The overall results show that microaneurysm detection is a challenging task for both the automatic methods as well as the human expert, and there is room for improvement as the best performing system does not reach the performance of thehuman expert.
Abstract: The detection of microaneurysms in digital color fundus photographs is a critical first step in automated screening for diabetic retinopathy (DR), a common complication of diabetes. To accomplish this detection numerous methods have been published in the past but none of these was compared with each other on the same data. In this work we present the results of the first international microaneurysm detection competition, organized in the context of the Retinopathy Online Challenge (ROC), a multiyear online competition for various aspects of DR detection. For this competition, we compare the results of five different methods, produced by five different teams of researchers on the same set of data. The evaluation was performed in a uniform manner using an algorithm presented in this work. The set of data used for the competition consisted of 50 training images with available reference standard and 50 test images where the reference standard was withheld by the organizers (M. Niemeijer, B. van Ginneken, and M. D. AbrA?moff). The results obtained on the test data was submitted through a website after which standardized evaluation software was used to determine the performance of each of the methods. A human expert detected microaneurysms in the test set to allow comparison with the performance of the automatic methods. The overall results show that microaneurysm detection is a challenging task for both the automatic methods as well as the human expert. There is room for improvement as the best performing system does not reach the performance of the human expert. The data associated with the ROC microaneurysm detection competition will remain publicly available and the website will continue accepting submissions.
TL;DR: Intacs technology can reduce the corneal steepening and astigmatism associated with keratoconus and increase in topographical regularity and increased uncorrected visual acuity.
Abstract: Purpose To evaluate the potential of intrastromal corneal ring technology (Intacs™, KeraVision) to correct keratoconus without central corneal scarring. Setting Department of Ophthalmology, Brest University Hospital, Brest, France. Methods In this prospective, noncomparative, interventional case series, Intacs segments were implanted in 10 keratoconic eyes with clear central corneas and contact lens intolerance after corneal pachymetry was checked. Segment thicknesses varied based on corneal topography analysis. Results No intraoperative complications occurred. The mean follow-up was 10.6 months. Postoperative results revealed a reduction in astigmatism and spherical correction and an increase in topographical regularity and increased uncorrected visual acuity. Conclusion Intacs technology can reduce the corneal steepening and astigmatism associated with keratoconus.
TL;DR: An automatic method to detect microaneurysms in retina photographs by locally matching a lesion template in sub- bands of wavelet transformed images is proposed, based on a genetic algorithm followed by Powell's direction set descent.
Abstract: In this paper, we propose an automatic method to detect microaneurysms in retina photographs. Microaneurysms are the most frequent and usually the first lesions to appear as a consequence of diabetic retinopathy. So, their detection is necessary for both screening the pathology and follow up (progression measurement). Automating this task, which is currently performed manually, would bring more objectivity and reproducibility. We propose to detect them by locally matching a lesion template in sub- bands of wavelet transformed images. To improve the method performance, we have searched for the best adapted wavelet within the lifting scheme framework. The optimization process is based on a genetic algorithm followed by Powell's direction set descent. Results are evaluated on 120 retinal images analyzed by an expert and the optimal wavelet is compared to different conventional mother wavelets. These images are of three different modalities: there are color photographs, green filtered photographs, and angiographs. Depending on the imaging modality, microaneurysms were detected with a sensitivity of respectively 89.62%, 90.24%, and 93.74% and a positive predictive value of respectively 89.50%, 89.75%, and 91.67%, which is better than previously published methods.
TL;DR: In this article, a generalization of the backpropagation method is proposed in order to train ConvNets that produce high-quality heatmaps, showing which pixels in images play a role in the image-level predictions.
Abstract: Deep learning is quickly becoming the leading methodology for medical image analysis. Given a large medical archive, where each image is associated with a diagnosis, efficient pathology detectors or classifiers can be trained with virtually no expert knowledge about the target pathologies. However, deep learning algorithms, including the popular ConvNets, are black boxes: little is known about the local patterns analyzed by ConvNets to make a decision at the image level. A solution is proposed in this paper to create heatmaps showing which pixels in images play a role in the image-level predictions. In other words, a ConvNet trained for image-level classification can be used to detect lesions as well. A generalization of the backpropagation method is proposed in order to train ConvNets that produce high-quality heatmaps. The proposed solution is applied to diabetic retinopathy (DR) screening in a dataset of almost 90,000 fundus photographs from the 2015 Kaggle Diabetic Retinopathy competition and a private dataset of almost 110,000 photographs (e-ophtha). For the task of detecting referable DR, very good detection performance was achieved: A z = 0.954 in Kaggle’s dataset and A z = 0.949 in e-ophtha. Performance was also evaluated at the image level and at the lesion level in the DiaretDB1 dataset, where four types of lesions are manually segmented: microaneurysms, hemorrhages, exudates and cotton-wool spots. For the task of detecting images containing these four lesion types, the proposed detector, which was trained to detect referable DR, outperforms recent algorithms trained to detect those lesions specifically, with pixel-level supervision. At the lesion level, the proposed detector outperforms heatmap generation algorithms for ConvNets. This detector is part of the Messidor® system for mobile eye pathology screening. Because it does not rely on expert knowledge or manual segmentation for detecting relevant patterns, the proposed solution is a promising image mining tool, which has the potential to discover new biomarkers in images.
TL;DR: An algorithm based on deep machine learning had high sensitivity and specificity for detecting referable diabetic retinopathy and diabetic macular edema in retinal fundus photographs from adults with diabetes.
Abstract: Importance Deep learning is a family of computational methods that allow an algorithm to program itself by learning from a large set of examples that demonstrate the desired behavior, removing the need to specify rules explicitly. Application of these methods to medical imaging requires further assessment and validation. Objective To apply deep learning to create an algorithm for automated detection of diabetic retinopathy and diabetic macular edema in retinal fundus photographs. Design and Setting A specific type of neural network optimized for image classification called a deep convolutional neural network was trained using a retrospective development data set of 128 175 retinal images, which were graded 3 to 7 times for diabetic retinopathy, diabetic macular edema, and image gradability by a panel of 54 US licensed ophthalmologists and ophthalmology senior residents between May and December 2015. The resultant algorithm was validated in January and February 2016 using 2 separate data sets, both graded by at least 7 US board-certified ophthalmologists with high intragrader consistency. Exposure Deep learning–trained algorithm. Main Outcomes and Measures The sensitivity and specificity of the algorithm for detecting referable diabetic retinopathy (RDR), defined as moderate and worse diabetic retinopathy, referable diabetic macular edema, or both, were generated based on the reference standard of the majority decision of the ophthalmologist panel. The algorithm was evaluated at 2 operating points selected from the development set, one selected for high specificity and another for high sensitivity. Results The EyePACS-1 data set consisted of 9963 images from 4997 patients (mean age, 54.4 years; 62.2% women; prevalence of RDR, 683/8878 fully gradable images [7.8%]); the Messidor-2 data set had 1748 images from 874 patients (mean age, 57.6 years; 42.6% women; prevalence of RDR, 254/1745 fully gradable images [14.6%]). For detecting RDR, the algorithm had an area under the receiver operating curve of 0.991 (95% CI, 0.988-0.993) for EyePACS-1 and 0.990 (95% CI, 0.986-0.995) for Messidor-2. Using the first operating cut point with high specificity, for EyePACS-1, the sensitivity was 90.3% (95% CI, 87.5%-92.7%) and the specificity was 98.1% (95% CI, 97.8%-98.5%). For Messidor-2, the sensitivity was 87.0% (95% CI, 81.1%-91.0%) and the specificity was 98.5% (95% CI, 97.7%-99.1%). Using a second operating point with high sensitivity in the development set, for EyePACS-1 the sensitivity was 97.5% and specificity was 93.4% and for Messidor-2 the sensitivity was 96.1% and specificity was 93.9%. Conclusions and Relevance In this evaluation of retinal fundus photographs from adults with diabetes, an algorithm based on deep machine learning had high sensitivity and specificity for detecting referable diabetic retinopathy. Further research is necessary to determine the feasibility of applying this algorithm in the clinical setting and to determine whether use of the algorithm could lead to improved care and outcomes compared with current ophthalmologic assessment.
Technische Universität München1, ETH Zurich2, University of Bern3, Harvard University4, National Institutes of Health5, University of Debrecen6, University Hospital Heidelberg7, McGill University8, University of Pennsylvania9, French Institute for Research in Computer Science and Automation10, University at Buffalo11, Microsoft12, University of Cambridge13, Stanford University14, University of Virginia15, Imperial College London16, Massachusetts Institute of Technology17, Columbia University18, Sabancı University19, Old Dominion University20, RMIT University21, Purdue University22, General Electric23
TL;DR: The Multimodal Brain Tumor Image Segmentation Benchmark (BRATS) as mentioned in this paper was organized in conjunction with the MICCAI 2012 and 2013 conferences, and twenty state-of-the-art tumor segmentation algorithms were applied to a set of 65 multi-contrast MR scans of low and high grade glioma patients.
Abstract: In this paper we report the set-up and results of the Multimodal Brain Tumor Image Segmentation Benchmark (BRATS) organized in conjunction with the MICCAI 2012 and 2013 conferences Twenty state-of-the-art tumor segmentation algorithms were applied to a set of 65 multi-contrast MR scans of low- and high-grade glioma patients—manually annotated by up to four raters—and to 65 comparable scans generated using tumor image simulation software Quantitative evaluations revealed considerable disagreement between the human raters in segmenting various tumor sub-regions (Dice scores in the range 74%–85%), illustrating the difficulty of this task We found that different algorithms worked best for different sub-regions (reaching performance comparable to human inter-rater variability), but that no single algorithm ranked in the top for all sub-regions simultaneously Fusing several good algorithms using a hierarchical majority vote yielded segmentations that consistently ranked above all individual algorithms, indicating remaining opportunities for further methodological improvements The BRATS image data and manual annotations continue to be publicly available through an online evaluation system as an ongoing benchmarking resource
TL;DR: Collagen crosslinking may be a new way for stopping the progression of keratectasia in patients with keratoconus and the need for penetrating keratoplasty might then be significantly reduced in keratconus.
Abstract: ● PURPOSE: In animal eyes, a significant increase in corneal biomechanical stiffness has been found after collagen crosslinking by combined riboflavin/ultraviolet-A (UVA) treatment. The aim of the present study was to evaluate the clinical usefulness of riboflavin/ UVA-induced collagen crosslinking for bringing the progression of keratoconus to a halt. ● DESIGN: Prospective, nonrandomized clinical pilot study. ● METHODS: Twenty-three eyes of 22 patients with moderate or advanced progressive keratoconus (maximum K value, 48 ‐72 diopters) were included. After central corneal abrasion, photosensitizing riboflavin drops were applied and the eyes exposed to UVA (370 nm, 3 mW/cm 2 ) in a 1-cm distance for 30 minutes. Postoperative examinations were performed in 6-month intervals, including visual acuity testing, corneal topography, slit-lamp examination, measurement of endothelial cell density, and photographic documentation. The follow-up time was between 3 months and 4 years. ● RESULTS: In all treated eyes, the progression of keratoconus was at least stopped. In 16 eyes (70%) regression with a reduction of the maximal keratometry readings by 2.01 diopters and of the refractive error by 1.14 diopters was found. Corneal and lens transparency, endothelial cell density, and intraocular pressure remained unchanged. Visual acuity improved slightly in 15 eyes (65%). ● CONCLUSIONS: Collagen crosslinking may be a new way for stopping the progression of keratectasia in patients with keratoconus. The need for penetrating keratoplasty might then be significantly reduced in keratoconus. Given the simplicity and minimal costs of the treatment, it might also be well-suited for developing countries. Long-term results are necessary to evaluate the duration of the stiffening effect and to exclude long
TL;DR: The "normal" central corneal thickness (CCT) value in human corneas was determined based on reported literature values for within-study average CCT values, and the reported impact of physiological variables, contact lens wear, pharmaceuticals, ocular disease, and ophthalmic surgery on CCT was assessed.
Abstract: We determined the "normal" central corneal thickness (CCT) value in human corneas based on reported literature values for within-study average CCT values, and used this as a reference to assess the reported impact of physiological variables (especially age and diurnal effects), contact lens wear, pharmaceuticals, ocular disease, and ophthalmic surgery on CCT. With the expected CCT and its variance defined, it should be possible to determine the potential impact of differences in CCT in intraocular pressure (IOP) assessments, especially by applanation tonometry, using a meta-analysis approach. Some 600 sets of CCT data were identified from the worldwide literature over the period of 1968 through mid-1999, of which 134 included IOP measures as well. The within-study average CCT values and reported variance (SD) was noted along with the number of eyes and any special characteristics, including probable ethnic origin of the study subjects. Various sets of data were subjected to statistical analyses. From 300 data sets from eyes designated as normal, the group-averaged CCT was 0.534 mm. From 230 data sets where interindividual variance was reported, the group-averaged CCT was 0.536 mm (median 0.536 mm; average SD of 0. 031 mm, average coefficient of variation = 5.8%). Overall, studies using slit-lamp-based pachometry have reported marginally lower CCT values (average 0.530 mm, average SD 0.029 mm) compared to ultrasound-based studies (average 0.544, average SD 0.034 mm), which perhaps reflects the type of individual studied (non-surgical vs. pre-surgical patients) rather than the technique itself. A slight chronological increase in reported average CCT values (approximately 0.006 mm/decade) was evident, but a substantial chronological increase was evident for ultrasound pachometry studies (approximately 0.015 mm/decade). Within the meta-analysis-generated average and variance, age had no obvious impact on CCT measures for *whites, although an age-related decline in CCT is evident for non-whites. Any diurnal effects are likely concealed within the expected variance in CCT. Contact lens wear and pharmaceuticals generally produced changes in CCT that were well within the expected variance in CCT. Of the ocular diseases, only those associated with collagen disorders (including keratoconus) or endothelial-based corneal dystrophies (e.g., Fuchs) were likely to result in decreases or increases, respectively, of CCT beyond the normal variance. Routine contact lens wear and diseases such as diabetes seem unlikely to produce changes in CCT of a magnitude that would justify pachometry as a monitoring method beyond routine slit-lamp evaluation. Increases in CCT beyond the expected variance were reported after a range of intraocular surgeries (cataract operations, penetrating keratoplasty), whereas photorefractive surgery produces a measurable decrease in CCT. A meta-analysis of possible association between CCT and IOP measures of 133 data sets, regardless of the type of eyes assessed, revealed a statistically significant correlation; a 10% difference in CCT would result in a 3. 4 +/- 0.9 mm Hg difference in IOP (P = 0.001, r = 0.419). The observed phenomenon was much smaller for eyes designated as healthy (1.1 +/- 0.6 mm Hg for a 10% difference in CCT, P = 0.023, r = 0. 331). For eyes with chronic diseases, the change was 2.5 +/- 1.1 mm Hg for a 10% difference in CCT (P = 0.005, r = 0.450), whereas a substantial but highly variable association was seen for eyes with acute onset disease (approximately 10.0 +/- 3.1 mm Hg for a 10% difference in CCT, P = 0.004, r = 0.623). Based on the meta-analysis, normal CCT in white adults would be expected to be within +/-11.6% (+/-2 SD) of 0.535 mm, i.e., 0.473-0.597 mm (95% CI, 0.474-0.596). The impact of CCT on applanation tonometry of healthy eyes is unlikely to achieve clinical significance, but for corneas of eyes with chronic disease, pachometry should be performed if the tonometry reveals IOP readi
01 May 1995