An incremental learning approach to automatically recognize pulmonary diseases from the multi-vendor chest radiographs.

The human respiratory network is a vital system that provides oxygen supply and nourishment to the whole body. Pulmonary diseases can cause severe respiratory problems, leading to sudden death if not treated timely. Many researchers have utilized deep learning systems (in both transfer learning and fine-tuning modes) to diagnose pulmonary disorders using chest X-rays (CXRs). However, such systems require exhaustive training efforts on large-scale (and well-annotated) data to effectively diagnose chest abnormalities (at the inference stage). Furthermore, procuring such large-scale data (in a clinical setting) is often infeasible and impractical, especially for rare diseases. With the recent advances in incremental learning, researchers have periodically tuned deep neural networks to learn different classification tasks with few training examples. Although, such systems can resist catastrophic forgetting, they treat the knowledge representations (which the network learns periodically) independently of each other, and this limits their classification performance. Also, to the best of our knowledge, there is no incremental learning-driven image diagnostic framework (to date) that is specifically designed to screen pulmonary disorders from the CXRs. To address this, we present a novel framework that can learn to screen different chest abnormalities incrementally (via few-shot training). In addition to this, the proposed framework is penalized through an incremental learning loss function that infers Bayesian theory to recognize structural and semantic inter-dependencies between incrementally learned knowledge representations to diagnose the pulmonary diseases effectively (at the inference stage), regardless of the scanner specifications. We tested the proposed framework on five public CXR datasets containing different chest abnormalities, where it achieved an accuracy of 0.8405 and the F1 score of 0.8303, outperforming various state-of-the-art incremental learning schemes. It also achieved a highly competitive performance compared to the conventional fine-tuning (transfer learning) approaches while significantly reducing the training and computational requirements.

RIDB: A Dataset of fundus images for retina based person identification.

The paper describes a dataset, entitled Retina Identification Database (RIDB). The stated dataset contains Retinal fundus images acquired using Fundus imaging camera TOPCON-TRC 50 EX. The abovementioned dataset holds a significant position in retinal recognition and identification. Retinal recognition is considered as one of the reliable biometric recognition features. Biometric recognition has become an integral part of any organization's security department. Before biometrics, the information was secured through passwords, pin keys, etc. However, the fear of decryption and hacking retained. Biometric verification includes behavioural (voice, signature, gait), morphological (Fingerprint, face, palm print, retina) and biological (Odour, saliva, DNA) features [1]. Amongst all of them, retina based identification is considered as the spoof proof and most accurate identification system. Since the retina is embedded inside the eye thus is least affected by the outer environment and retain in its original state. Moreover, the vascular pattern in the retina is unique and remains unchanged during the entire life span. The data presented in the paper is composed of 100 retinal images of 20 individuals (5 images were captured from each patient). The dataset is supported by research work [2] and [7]. These research papers proposed retinal recognition algorithms for biometric verification and identification. The proposed method utilized both vascular and non-vascular features for identification and yields recognition rates of 100 % and 92.5% respectively.

Extraction of Retinal Layers Through Convolution Neural Network (CNN) in an OCT Image for Glaucoma Diagnosis.

Glaucoma is a progressive and deteriorating optic neuropathy that leads to visual field defects. The damage occurs as glaucoma is irreversible, so early and timely diagnosis is of significant importance. The proposed system employs the convolution neural network (CNN) for automatic segmentation of the retinal layers. The inner limiting membrane (ILM) and retinal pigmented epithelium (RPE) are used to calculate cup-to-disc ratio (CDR) for glaucoma diagnosis. The proposed system uses structure tensors to extract candidate layer pixels, and a patch across each candidate layer pixel is extracted, which is classified using CNN. The proposed framework is based upon VGG-16 architecture for feature extraction and classification of retinal layer pixels. The output feature map is merged into SoftMax layer for classification and produces probability map for central pixel of each patch and decides whether it is ILM, RPE, or background pixels. Graph search theory refines the extracted layers by interpolating the missing points, and these extracted ILM and RPE are finally used to compute CDR value and diagnose glaucoma. The proposed system is validated using a local dataset of optical coherence tomography images from 196 patients, including normal and glaucoma subjects. The dataset contains manually annotated ILM and RPE layers; manually extracted patches for ILM, RPE, and background pixels; CDR values; and eventually final finding related to glaucoma. The proposed system is able to extract ILM and RPE with a small absolute mean error of 6.03 and 5.56, respectively, and it finds CDR value within average range of ± 0.09 as compared with glaucoma expert. The proposed system achieves average sensitivity, specificity, and accuracies of 94.6, 94.07, and 94.68, respectively.

An Integrated Approach for Spatio-Temporal Cholera Disease Hotspot Relation Mining for Public Health Management in Punjab, Pakistan.

Public health management can generate actionable results when diseases are studied in context with other candidate factors contributing to disease dynamics. In order to fully understand the interdependent relationships of multiple geospatial features involved in disease dynamics, it is important to construct an effective representation model that is able to reveal the relationship patterns and trends. The purpose of this work is to combine disease incidence spatio-temporal data with other features of interest in a mutlivariate spatio-temporal model for investigating characteristic disease and feature patterns over identified hotspots. We present an integrated approach in the form of a disease management model for analyzing spatio-temporal dynamics of disease in connection with other determinants. Our approach aligns spatio-temporal profiles of disease with other driving factors in public health context to identify hotspots and patterns of disease and features of interest in the identified locations. We evaluate our model against cholera disease outbreaks from 2015-2019 in Punjab province of Pakistan. The experimental results showed that the presented model effectively address the complex dynamics of disease incidences in the presence of other features of interest over a geographic area representing populations and sub populations during a given time. The presented methodology provides an effective mechanism for identifying disease hotspots in multiple dimensions and relation between the hotspots for cost-effective and optimal resource allocation as well as a sound reference for further predictive and forecasting analysis.

A Nonparametric Approach for Mild Cognitive Impairment to AD Conversion Prediction: Results on Longitudinal Data.

The goal of this study is to introduce a nonparametric technique for predicting conversion from Mild Cognitive impairment (MCI)-to-Alzheimer's disease (AD). Progression of a slowly progressing disease such as AD benefits from the use of longitudinal data; however, research till now is limited due to the insufficient patient data and short follow-up time. A small dataset size invalidates the estimation of underlying disease progression model; hence, a supervised nonparametric method is proposed. While depicting a real-world setting, longitudinal data of three years are employed for training, whereas only the baseline visit's data is used for validation. The train set is preprocessed for extraction of two dense clusters representing the subjects who remain stable at MCI or progress to AD after three years of the baseline visit. Similarity between these clusters and the test point is calculated in Euclidean space. Multiple features from two modalities of biomarkers, i.e., neuropsychological measures (NM) and structural magnetic resonance imaging (MRI) morphometry are also analyzed. Due to the limited MCI dataset size (NM: 145, MRI: 52, NM+MRI: 29), leave-one-out cross validation setup is employed for performance evaluation. The algorithm performance is noted for both unimodal case and bimodal cases. Superior performance (accuracy: 89.66%, sensitivity: 87.50%, specificity: 92.31%, precision: 93.33%) is delivered by multivariate predictors. Three notable conclusions of this study are: 1) Longitudinal data are more powerful than the temporal data, 2) MRI is a better predictor of MCI-to-AD conversion than NM, and 3) multivariate predictors outperform single predictor models.