Similarity regularized sparse group lasso for cup to disc ratio computation

Machine Learning and Deep Learning Techniques for Optic Disc and Cup Segmentation - A Review

Background

Globally, glaucoma is the second leading cause of blindness. Detecting glaucoma in the early stages is essential to avoid disease complications, which lead to blindness. Thus, computer-aided diagnosis systems are powerful tools to overcome the shortage of glaucoma screening programs.

Methods

A systematic search of public databases, including PubMed, Google Scholar, and other sources, was performed to identify relevant studies to overview the publicly available fundus image datasets used to train, validate, and test machine learning and deep learning methods. Additionally, existing machine learning and deep learning methods for optic cup and disc segmentation were surveyed and critically reviewed.

Results

Eight fundus images datasets were publicly available with 15,445 images labeled with glaucoma or non-glaucoma, and manually annotated optic disc and cup boundaries were found. Five metrics were identified for evaluating the developed models. Finally, three main deep learning architectural designs were commonly used for optic disc and optic cup segmentation.

Conclusion

We provided future research directions to formulate robust optic cup and disc segmentation systems. Deep learning can be utilized in clinical settings for this task. However, many challenges need to be addressed before using this strategy in clinical trials. Finally, two deep learning architectural designs have been widely adopted, such as U-net and its variants.

Conditional Adversarial Transfer for Glaucoma Diagnosis

Deep learning has achieved great success in image classification task when given sufficient labeled training images. However, in fundus image based glaucoma diagnosis, we often have very limited training data due to expensive cost in data labeling. Moreover, when facing a new application environment, it is difficult to train a network with limited labeled training images. In this case, some images from some auxiliary domains (i.e., source domain) could be exploited to improve the performance. Unfortunately, direct using the source domain data may not achieve promising performance for the domain of interest (i.e., target domain) due to reasons like distribution discrepancy between two domains. In this paper, focusing on glaucoma diagnosis, we propose a deep adversarial transfer learning method conditioned on label information to match the distributions of source and target domains, so that the labeled source images can be leveraged to improve the classification performance in the target domain. Different from the most existing adversarial transfer learning methods which consider marginal distribution matching only, we seek to match the label conditional distributions by handling images with different labels separately. We conduct experiments on three glaucoma datasets and adopt multiple evaluation metrics to verify the effectiveness of our proposed method.

Clinical validation of RIA-G, an automated optic nerve head analysis software

Purpose

To clinically validate a new automated glaucoma diagnosis software RIA-G.

Methods

A double-blinded study was conducted where 229 valid random fundus images were evaluated independently by RIA-G and three expert ophthalmologists. Optic nerve head parameters [vertical and horizontal cup-disc ratio (CDR) and neuroretinal rim (NRR) changes] were quantified. Disc damage likelihood scale (DDLS) staging and presence of glaucoma were noted. The software output was compared with consensus values of ophthalmologists.

Results

Mean difference between the vertical CDR output by RIA-G and the ophthalmologists was - 0.004 ± 0.1. Good agreement and strong correlation existed between the two [interclass correlation coefficient (ICC) 0.79; r = 0.77, P < 0.005]. Mean difference for horizontal CDR was - 0.07 ± 0.13 with a moderate to strong agreement and correlation (ICC 0.48; r = 0.61, P < 0.05). Experts and RIA-G found a violation of the inferior-superior NRR in 47 and 54 images, respectively (Cohen's kappa = 0.56 ± 0.07). RIA-G accurately detected DDLS in 66.2% cases, while in 93.8% cases, output was within ± 1 stage (ICC 0.51). Sensitivity and specificity of RIA-G to diagnose glaucomatous neuropathy were 82.3% and 91.8%, respectively. Overall agreement between RIA-G and experts for glaucoma diagnosis was good (Cohen's kappa = 0.62 ± 0.07). Overall accuracy of RIA-G to detect glaucomatous neuropathy was 90.3%. A detection error rate of 5% was noted.

Conclusion

RIA-G showed good agreement with the experts and proved to be a reliable software for detecting glaucomatous optic neuropathy. The ability to quantify optic nerve head parameters from simple fundus photographs will prove particularly useful in glaucoma screening, where no direct patient-doctor contact is established.

Accurate, fast, data efficient and interpretable glaucoma diagnosis with automated spatial analysis of the whole cup to disc profile

Background

Glaucoma is the leading cause of irreversible blindness worldwide. It is a heterogeneous group of conditions with a common optic neuropathy and associated loss of peripheral vision. Both over and under-diagnosis carry high costs in terms of healthcare spending and preventable blindness. The characteristic clinical feature of glaucoma is asymmetrical optic nerve rim narrowing, which is difficult for humans to quantify reliably. Strategies to improve and automate optic disc assessment are therefore needed to prevent sight loss.

Methods

We developed a novel glaucoma detection algorithm that segments and analyses colour photographs to quantify optic nerve rim consistency around the whole disc at 15-degree intervals. This provides a profile of the cup/disc ratio, in contrast to the vertical cup/disc ratio in common use. We introduce a spatial probabilistic model, to account for the optic nerve shape, we then use this model to derive a disc deformation index and a decision rule for glaucoma. We tested our algorithm on two separate image datasets (ORIGA and RIM-ONE).

Results

The spatial algorithm accurately distinguished glaucomatous and healthy discs on internal and external validation (AUROC 99.6% and 91.0% respectively). It achieves this using a dataset 100-times smaller than that required for deep learning algorithms, is flexible to the type of cup and disc segmentation (automated or semi-automated), utilises images with missing data, and is correlated with the disc size (p = 0.02) and the rim-to-disc at the narrowest rim (p<0.001, in external validation).

Discussion

The spatial probabilistic algorithm is highly accurate, highly data efficient and it extends to any imaging hardware in which the boundaries of cup and disc can be segmented, thus making the algorithm particularly applicable to research into disease mechanisms, and also glaucoma screening in low resource settings.

Sparse Range-Constrained Learning and Its Application for Medical Image Grading

Sparse learning has been shown to be effective in solving many real-world problems. Finding sparse representations is a fundamentally important topic in many fields of science including signal processing, computer vision, genome study, and medical imaging. One important issue in applying sparse representation is to find the basis to represent the data, especially in computer vision and medical imaging where the data are not necessary incoherent. In medical imaging, clinicians often grade the severity or measure the risk score of a disease based on images. This process is referred to as medical image grading. Manual grading of the disease severity or risk score is often used. However, it is tedious, subjective, and expensive. Sparse learning has been used for automatic grading of medical images for different diseases. In the grading, we usually begin with one step to find a sparse representation of the testing image using a set of reference images or atoms from the dictionary. Then in the second step, the selected atoms are used as references to compute the grades of the testing images. Since the two steps are conducted sequentially, the objective function in the first step is not necessarily optimized for the second step. In this paper, we propose a novel sparse range-constrained learning (SRCL) algorithm for medical image grading. Different from most of existing sparse learning algorithms, SRCL integrates the objective of finding a sparse representation and that of grading the image into one function. It aims to find a sparse representation of the testing image based on atoms that are most similar in both the data or feature representation and the medical grading scores. We apply the new proposed SRCL to two different applications, namely, cup-to-disc ratio (CDR) computation from retinal fundus images and cataract grading from slit-lamp lens images. Experimental results show that the proposed method is able to improve the accuracy in CDR computation and cataract grading.

Structure-Preserving Guided Retinal Image Filtering and Its Application for Optic Disk Analysis

Retinal fundus photographs have been used in the diagnosis of many ocular diseases such as glaucoma, pathological myopia, age-related macular degeneration, and diabetic retinopathy. With the development of computer science, computer aided diagnosis has been developed to process and analyze the retinal images automatically. One of the challenges in the analysis is that the quality of the retinal image is often degraded. For example, a cataract in human lens will attenuate the retinal image, just as a cloudy camera lens which reduces the quality of a photograph. It often obscures the details in the retinal images and posts challenges in retinal image processing and analyzing tasks. In this paper, we approximate the degradation of the retinal images as a combination of human-lens attenuation and scattering. A novel structure-preserving guided retinal image filtering (SGRIF) is then proposed to restore images based on the attenuation and scattering model. The proposed SGRIF consists of a step of global structure transferring and a step of global edge-preserving smoothing. Our results show that the proposed SGRIF method is able to improve the contrast of retinal images, measured by histogram flatness measure, histogram spread, and variability of local luminosity. In addition, we further explored the benefits of SGRIF for subsequent retinal image processing and analyzing tasks. In the two applications of deep learning-based optic cup segmentation and sparse learning-based cup-to-disk ratio (CDR) computation, our results show that we are able to achieve more accurate optic cup segmentation and CDR measurements from images processed by SGRIF.

Joint Optic Disc and Cup Segmentation Based on Multi-Label Deep Network and Polar Transformation

Glaucoma is a chronic eye disease that leads to irreversible vision loss. The cup to disc ratio (CDR) plays an important role in the screening and diagnosis of glaucoma. Thus, the accurate and automatic segmentation of optic disc (OD) and optic cup (OC) from fundus images is a fundamental task. Most existing methods segment them separately, and rely on hand-crafted visual feature from fundus images. In this paper, we propose a deep learning architecture, named M-Net, which solves the OD and OC segmentation jointly in a one-stage multi-label system. The proposed M-Net mainly consists of multi-scale input layer, U-shape convolutional network, side-output layer, and multi-label loss function. The multi-scale input layer constructs an image pyramid to achieve multiple level receptive field sizes. The U-shape convolutional network is employed as the main body network structure to learn the rich hierarchical representation, while the side-output layer acts as an early classifier that produces a companion local prediction map for different scale layers. Finally, a multi-label loss function is proposed to generate the final segmentation map. For improving the segmentation performance further, we also introduce the polar transformation, which provides the representation of the original image in the polar coordinate system. The experiments show that our M-Net system achieves state-of-the-art OD and OC segmentation result on ORIGA data set. Simultaneously, the proposed method also obtains the satisfactory glaucoma screening performances with calculated CDR value on both ORIGA and SCES datasets.