Fast Convolutional Neural Network Training Using Selective Data Sampling: Application to Hemorrhage Detection in Color Fundus Images

Automatic segmentation of nine retinal layer boundaries in OCT images of non-exudative AMD patients using deep learning and graph search

We present a novel framework combining convolutional neural networks (CNN) and graph search methods (termed as CNN-GS) for the automatic segmentation of nine layer boundaries on retinal optical coherence tomography (OCT) images. CNN-GS first utilizes a CNN to extract features of specific retinal layer boundaries and train a corresponding classifier to delineate a pilot estimate of the eight layers. Next, a graph search method uses the probability maps created from the CNN to find the final boundaries. We validated our proposed method on 60 volumes (2915 B-scans) from 20 human eyes with non-exudative age-related macular degeneration (AMD), which attested to effectiveness of our proposed technique.

Deep image mining for diabetic retinopathy screening

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.

"How much realism is needed?" - the wrong question in silico imagers have been asking

PURPOSE

To discuss the use of realism as a first approximation for assessing computational imaging methods.

METHODS

Although in silico methods are increasingly becoming promising surrogates to physical experimentation for various stages of device development, their acceptance remains challenging. Realism is often considered as a first approximation for assessing computational imaging methods. However, realism is subjective and does not always ensure that key features of the methodologies reflect relevant aspects of devices of interest to imaging scientists, regulators, and medical practitioners. Moreover, in some cases (e.g., in computerized image analysis applications where human interpretation is not needed) how realistic in silico images are is irrelevant and perhaps misleading.

RESULTS

I emphasize a divergence from this methodology by providing a rationale for evaluating in silico imaging methods and tools in an objective and measurable manner.

CONCLUSIONS

Improved approaches for in silico imaging will lead to the rapid advancement and acceptance of computational techniques in medical imaging primarily but not limited to the regulatory evaluation of new imaging products.

Detecting retinal microaneurysms and hemorrhages with robustness to the presence of blood vessels

BACKGROUND AND OBJECTIVES

Diabetic Retinopathy is the leading cause of blindness in developed countries in the age group 20-74 years. It is characterized by lesions on the retina and this paper focuses on detecting two of these lesions, Microaneurysms and Hemorrhages, which are also known as red lesions. This paper attempts to deal with two problems in detecting red lesions from retinal fundus images: (1) false detections on blood vessels; and (2) different size of red lesions.

METHODS

To deal with false detections on blood vessels, novel filters have been proposed which can distinguish between red lesions and blood vessels. This distinction is based on the fact that vessels are elongated while red lesions are usually circular blob-like structures. The second problem of the different size of lesions is dealt with by applying the proposed filters on patches of different sizes instead of filtering the full image. These patches are obtained by dividing the original image using a grid whose size determines the patch size. Different grid sizes were used and lesion detection results for these grid sizes were combined using Multiple Kernel Learning.

RESULTS

Experiments on a dataset of 143 images showed that proposed filters detected Microaneurysms and Hemorrhages successfully even when these lesions were close to blood vessels. In addition, using Multiple Kernel Learning improved the results when compared to using a grid of one size only. The areas under receiver operating characteristic curve were found to be 0.97 and 0.92 for Microaneurysms and Hemorrhages respectively which are better than the existing related works.

CONCLUSIONS

Proposed filters are robust to the presence of blood vessels and surpass related works in detecting red lesions from retinal fundus images. Improved lesion detection using the proposed approach can help in automatic detection of Diabetic Retinopathy.

Deep Learning for Health Informatics

With a massive influx of multimodality data, the role of data analytics in health informatics has grown rapidly in the last decade. This has also prompted increasing interests in the generation of analytical, data driven models based on machine learning in health informatics. Deep learning, a technique with its foundation in artificial neural networks, is emerging in recent years as a powerful tool for machine learning, promising to reshape the future of artificial intelligence. Rapid improvements in computational power, fast data storage, and parallelization have also contributed to the rapid uptake of the technology in addition to its predictive power and ability to generate automatically optimized high-level features and semantic interpretation from the input data. This article presents a comprehensive up-to-date review of research employing deep learning in health informatics, providing a critical analysis of the relative merit, and potential pitfalls of the technique as well as its future outlook. The paper mainly focuses on key applications of deep learning in the fields of translational bioinformatics, medical imaging, pervasive sensing, medical informatics, and public health.