SISE-PC: Semi-supervised Image Subsampling for Explainable Pathology Classification.

Although automated pathology classification using deep learning (DL) has proved to be predictively efficient, DL methods are found to be data and compute cost intensive. In this work, we aim to reduce DL training costs by pre-training a ResNet feature extractor using SimCLR contrastive loss for latent encoding of OCT images. We propose a novel active learning framework that identifies a minimal sub-sampled dataset containing the most uncertain OCT image samples using label propagation on the SimCLR latent encodings. The pre-trained ResNet model is then fine-tuned with the labelled minimal sub-sampled data and the underlying pathological sites are visually explained. Our framework identifies upto 2% of OCT images to be most uncertain that need prioritized specialist attention and that can fine-tune a ResNet model to achieve upto 97% classification accuracy. The proposed method can be extended to other medical images to minimize prediction costs.

Deep-ReAP: Deep Representations And Partial label learning for Multi-pathology Classification.

Automated detection of pathology in images with multiple pathologies is one of the most challenging problems in medical diagnostics. The primary hurdles for automated systems include data imbalance across pathology categories and structural variations in pathological manifestations across patients. In this work, we present a novel method to detect a minimal dataset to train deep learning models that classify and explain multiple pathologies through the deep representations. We implement partial label learning with 1% false labels to identify the under-fit pathological categories that need further training followed by fine-tuning the deep representations. The proposed method identifies 54% of available training images as optimal for explainable classification of upto 7 pathological categories that can co-exist in 36 various combinations in retinal images, with overall precision/recall/Fß scores of 57%/87%/80%. Thus, the proposed method can lead to explainable inferencing for multi-label medical image data sets.

Depthwise Separable Convolutional Neural Network Model for Intra-Retinal Cyst Segmentation.

Intra-retinal cysts (IRCs) are significant in detecting several ocular and retinal pathologies. Segmentation and quantification of IRCs from optical coherence tomography (OCT) scans is a challenging task due to present of speckle noise and scan intensity variations across the vendors. This work proposes a convolutional neural network (CNN) model with an encoder-decoder pair architecture for IRC segmentation across different cross-vendor OCT scans. Since deep CNN models have high computational complexity due to a large number of parameters, the proposed method of depthwise separable convolutional filters aids model generalizability and prevents model over-fitting. Also, the swish activation function is employed to prevent the vanishing gradient problem. The optima cyst segmentation challenge (OCSC) dataset with four different vendor OCT device scans is used to evaluate the proposed model. Our model achieves a mean Dice score of 0.74 and mean recall/precision rate of 0.72/0.82 across different imaging vendors and it outperforms existing algorithms on the OCSC dataset.

A benchmark study of automated intra-retinal cyst segmentation algorithms using optical coherence tomography B-scans.

(BACKGROUND AND OBJECTIVES): Retinal cysts are formed by accumulation of fluid in the retina caused by leakages from inflammation or vitreous fractures. Analysis of the retinal cystic spaces holds significance in detection and treatment of several ocular diseases like age-related macular degeneration, diabetic macular edema etc. Thus, segmentation of intra-retinal cysts and quantification of cystic spaces are vital for retinal pathology and severity detection. In the recent years, automated segmentation of intra-retinal cysts using optical coherence tomography B-scans has gained significant importance in the field of retinal image analysis. The objective of this paper is to compare different intra-retinal cyst segmentation algorithms for comparative analysis and benchmarking purposes. (METHODS): In this work, we employ a modular approach for standardizing the different segmentation algorithms. Further, we analyze the variations in automated cyst segmentation performances and method scalability across image acquisition systems by using the publicly available cyst segmentation challenge dataset (OPTIMA cyst segmentation challenge). (RESULTS): Several key automated methods are comparatively analyzed using quantitative and qualitative experiments. Our analysis demonstrates the significance of variations in signal-to-noise ratio (SNR), retinal layer morphology and post-processing steps on the automated cyst segmentation processes. (CONCLUSION): This benchmarking study provides insights towards the scalability of automated processes across vendor-specific imaging modalities to provide guidance for retinal pathology diagnostics and treatment processes.

Optic Disc Boundary and Vessel Origin Segmentation of Fundus Images.

This paper presents a novel classification-based optic disc (OD) segmentation algorithm that detects the OD boundary and the location of vessel origin (VO) pixel. First, the green plane of each fundus image is resized and morphologically reconstructed using a circular structuring element. Bright regions are then extracted from the morphologically reconstructed image that lie in close vicinity of the major blood vessels. Next, the bright regions are classified as bright probable OD regions and non-OD regions using six region-based features and a Gaussian mixture model classifier. The classified bright probable OD region with maximum Vessel-Sum and Solidity is detected as the best candidate region for the OD. Other bright probable OD regions within 1-disc diameter from the centroid of the best candidate OD region are then detected as remaining candidate regions for the OD. A convex hull containing all the candidate OD regions is then estimated, and a best-fit ellipse across the convex hull becomes the segmented OD boundary. Finally, the centroid of major blood vessels within the segmented OD boundary is detected as the VO pixel location. The proposed algorithm has low computation time complexity and it is robust to variations in image illumination, imaging angles, and retinal abnormalities. This algorithm achieves 98.8%-100% OD segmentation success and OD segmentation overlap score in the range of 72%-84% on images from the six public datasets of DRIVE, DIARETDB1, DIARETDB0, CHASE_DB1, MESSIDOR, and STARE in less than 2.14 s per image. Thus, the proposed algorithm can be used for automated detection of retinal pathologies, such as glaucoma, diabetic retinopathy, and maculopathy.

Automated detection of neovascularization for proliferative diabetic retinopathy screening.

Neovascularization is the primary manifestation of proliferative diabetic retinopathy (PDR) that can lead to acquired blindness. This paper presents a novel method that classifies neovascularizations in the 1-optic disc (OD) diameter region (NVD) and elsewhere (NVE) separately to achieve low false positive rates of neovascularization classification. First, the OD region and blood vessels are extracted. Next, the major blood vessel segments in the 1-OD diameter region are classified for NVD, and minor blood vessel segments elsewhere are classified for NVE. For NVD and NVE classifications, optimal region-based feature sets of 10 and 6 features, respectively, are used. The proposed method achieves classification sensitivity, specificity and accuracy for NVD and NVE of 74%, 98.2%, 87.6%, and 61%, 97.5%, 92.1%, respectively. Also, the proposed method achieves 86.4% sensitivity and 76% specificity for screening images with PDR from public and local data sets. Thus, the proposed NVD and NVE detection methods can play a key role in automated screening and prioritization of patients with diabetic retinopathy.

Artery/vein classification of retinal blood vessels using feature selection.

Automated classification of retinal vessels in fundus images is the first step towards measurement of retinal characteristics that can be used to screen and diagnose vessel abnormalities for cardiovascular and retinal disorders. This paper presents a novel approach to vessel classification to compute the artery/vein ratio (AVR) for all blood vessel segments in the fundus image. The features extracted are then subjected to a selection procedure using Random Forests (RF) where the features that contribute most to classification accuracy are chosen as input to a polynomial kernel Support Vector Machine (SVM) classifier. The most dominant feature was found to be the vessel information obtained from the Light plane of the LAB color space. The SVM is then subjected to one time training using 10-fold cross validation on images randomly selected from the VICAVR dataset before testing on an independent test dataset, derived from the same database. An Area Under the ROC Curve (AUC) of 97.2% was obtained on an average of 100 runs of the algorithm. The proposed algorithm is robust due to the feature selection procedure, and it is possible to get similar accuracies across many datasets.

Classification of large-scale fundus image data sets: a cloud-computing framework.

Large medical image data sets with high dimensionality require substantial amount of computation time for data creation and data processing. This paper presents a novel generalized method that finds optimal image-based feature sets that reduce computational time complexity while maximizing overall classification accuracy for detection of diabetic retinopathy (DR). First, region-based and pixel-based features are extracted from fundus images for classification of DR lesions and vessel-like structures. Next, feature ranking strategies are used to distinguish the optimal classification feature sets. DR lesion and vessel classification accuracies are computed using the boosted decision tree and decision forest classifiers in the Microsoft Azure Machine Learning Studio platform, respectively. For images from the DIARETDB1 data set, 40 of its highest-ranked features are used to classify four DR lesion types with an average classification accuracy of 90.1% in 792 seconds. Also, for classification of red lesion regions and hemorrhages from microaneurysms, accuracies of 85% and 72% are observed, respectively. For images from STARE data set, 40 high-ranked features can classify minor blood vessels with an accuracy of 83.5% in 326 seconds. Such cloud-based fundus image analysis systems can significantly enhance the borderline classification performances in automated screening systems.

Iterative Vessel Segmentation of Fundus Images.

This paper presents a novel unsupervised iterative blood vessel segmentation algorithm using fundus images. First, a vessel enhanced image is generated by tophat reconstruction of the negative green plane image. An initial estimate of the segmented vasculature is extracted by global thresholding the vessel enhanced image. Next, new vessel pixels are identified iteratively by adaptive thresholding of the residual image generated by masking out the existing segmented vessel estimate from the vessel enhanced image. The new vessel pixels are, then, region grown into the existing vessel, thereby resulting in an iterative enhancement of the segmented vessel structure. As the iterations progress, the number of false edge pixels identified as new vessel pixels increases compared to the number of actual vessel pixels. A key contribution of this paper is a novel stopping criterion that terminates the iterative process leading to higher vessel segmentation accuracy. This iterative algorithm is robust to the rate of new vessel pixel addition since it achieves 93.2-95.35% vessel segmentation accuracy with 0.9577-0.9638 area under ROC curve (AUC) on abnormal retinal images from the STARE dataset. The proposed algorithm is computationally efficient and consistent in vessel segmentation performance for retinal images with variations due to pathology, uneven illumination, pigmentation, and fields of view since it achieves a vessel segmentation accuracy of about 95% in an average time of 2.45, 3.95, and 8 s on images from three public datasets DRIVE, STARE, and CHASE_DB1, respectively. Additionally, the proposed algorithm has more than 90% segmentation accuracy for segmenting peripapillary blood vessels in the images from the DRIVE and CHASE_DB1 datasets.

3-D localization of Diabetic Macular Edema using OCT thickness maps.

This paper presents a novel automated system that segments 3-D thickness maps of sub-retinal layers from healthy and abnormal OCT image stacks with Diabetic Macular Edema (DME). These automated thickness maps are well correlated (r > 0.7) with the manually segmented thickness maps. The thickness maps demonstrate highly irregular regions in the inner and outer nuclear layers for patients with DME when compared to the healthy images. The combined area of irregularity in the inner and outer nuclear layers can thereby be extracted as a novel metric with correlation r = 0.99 to track the severity of DME. No other existing automated algorithm has extracted inner sub-retinal layer thickness maps from OCT image stacks of DME patients. The proposed system is fast and robust in locating the sub-retinal changes caused by DME in the 3-D sub-retinal micro-structure.

Blood Vessel Segmentation of Fundus Images by Major Vessel Extraction and Subimage Classification.

This paper presents a novel three-stage blood vessel segmentation algorithm using fundus photographs. In the first stage, the green plane of a fundus image is preprocessed to extract a binary image after high-pass filtering, and another binary image from the morphologically reconstructed enhanced image for the vessel regions. Next, the regions common to both the binary images are extracted as the major vessels. In the second stage, all remaining pixels in the two binary images are classified using a Gaussian mixture model (GMM) classifier using a set of eight features that are extracted based on pixel neighborhood and first and second-order gradient images. In the third postprocessing stage, the major portions of the blood vessels are combined with the classified vessel pixels. The proposed algorithm is less dependent on training data, requires less segmentation time and achieves consistent vessel segmentation accuracy on normal images as well as images with pathology when compared to existing supervised segmentation methods. The proposed algorithm achieves a vessel segmentation accuracy of 95.2%, 95.15%, and 95.3% in an average of 3.1, 6.7, and 11.7 s on three public datasets DRIVE, STARE, and CHASE_DB1, respectively.

DREAM: diabetic retinopathy analysis using machine learning.

This paper presents a computer-aided screening system (DREAM) that analyzes fundus images with varying illumination and fields of view, and generates a severity grade for diabetic retinopathy (DR) using machine learning. Classifiers such as the Gaussian Mixture model (GMM), k-nearest neighbor (kNN), support vector machine (SVM), and AdaBoost are analyzed for classifying retinopathy lesions from nonlesions. GMM and kNN classifiers are found to be the best classifiers for bright and red lesion classification, respectively. A main contribution of this paper is the reduction in the number of features used for lesion classification by feature ranking using Adaboost where 30 top features are selected out of 78. A novel two-step hierarchical classification approach is proposed where the nonlesions or false positives are rejected in the first step. In the second step, the bright lesions are classified as hard exudates and cotton wool spots, and the red lesions are classified as hemorrhages and micro-aneurysms. This lesion classification problem deals with unbalanced datasets and SVM or combination classifiers derived from SVM using the Dempster-Shafer theory are found to incur more classification error than the GMM and kNN classifiers due to the data imbalance. The DR severity grading system is tested on 1200 images from the publicly available MESSIDOR dataset. The DREAM system achieves 100% sensitivity, 53.16% specificity, and 0.904 AUC, compared to the best reported 96% sensitivity, 51% specificity, and 0.875 AUC, for classifying images as with or without DR. The feature reduction further reduces the average computation time for DR severity per image from 59.54 to 3.46 s.

Automated localization of cysts in diabetic macular edema using optical coherence tomography images.

This paper presents a novel automated system that localizes cysts in optical coherence tomography (OCT) images of patients with diabetic macular edema (DME). First, in each image, six sub-retinal layers are detected using an iterative high-pass filtering approach. Next, significantly dark regions within the retinal micro-structure are detected as candidate cystoid regions. Each candidate cystoid region is then further analyzed using solidity, mean and maximum pixel value of the negative OCT image as decisive features for estimating the area of cystoid regions. The proposed system achieves 90% correlation between the estimated cystoid area and the manually marked area, and a mean error of 4.6%. Finally the proposed algorithm locates the cysts in the inner plexiform region, inner nuclear region and outer nuclear region with an accuracy of 88%, 86% and 80%, respectively.

Promoting healthy spacing between pregnancies in India: need for differential education campaigns.

OBJECTIVE: Birth spacing intervals are relatively short in India. Healthy spacing of 3-5 years between births is an effective way to prevent maternal and child mortality and morbidities. Socio-cultural and structural barriers, including limited awareness, socio-cultural norms, and misconceptions need to be addressed for behavior change. Hence the objective was to understand these barriers and accordingly develop separate messages for young women, her husband and her mother-in-law. METHODS: Data were collected from young women, husbands and mothers-in-law using qualitative methods. Altogether 16 Focus Group Discussions and 30 in-depth interviews were conducted. Beliefs related to need of spacing, disadvantages of closely spaced pregnancies and messages considered suitable for different stakeholders were investigated. Messages were identified for women, husband and mother-in-law; communication aids prepared and community workers trained to appropriately communicate the messages to stakeholders. Quantitative data were collected to measure the effect of the intervention. RESULTS: Educational campaign resulted in higher use of contraceptives for spacing among registered pregnant women from experimental area compared to control area. CONCLUSION: Differential audience specific educational campaign is feasible and effective. PRACTICE IMPLICATIONS: For an effective communication in the community, workers should know how exactly to convey the different health messages to different target population.