Investigating Beta-Variational Convolutional Autoencoders for the Unsupervised Classification of Chest Pneumonia.

The world's population is increasing and so is the challenge on existing healthcare infrastructure to cope with the growing demand in medical diagnosis and evaluation. Although human experts are primarily tasked with the diagnosis of different medical conditions, artificial intelligence (AI)-assisted diagnoses have become considerably useful in recent times. One of the critical lung infections, which requires early diagnosis and subsequent treatment to reduce the mortality rate, is pneumonia. There are different methods for obtaining a pneumonia diagnosis; however, the adoption of chest X-rays is popular since it is non-invasive. The AI systems for a pneumonia diagnosis using chest X-rays are often built on supervised machine-learning (ML) models, which require labeled datasets for development. However, collecting labeled datasets is sometimes infeasible due to constraints such as human resources, cost, and time. As such, the problem that we address in this paper is the unsupervised classification of pneumonia using unsupervised ML models including the beta-variational convolutional autoencoder (ß-VCAE) and other variants, such as convolutional autoencoders (CAE), denoising convolutional autoencoders (DCAE), and sparse convolutional autoencoders (SCAE). Namely, the pneumonia classification problem is cast into an anomaly detection to develop the aforementioned ML models. The experimental results show that pneumonia can be diagnosed with high recall, precision, f1-score, and f2-score using the proposed unsupervised models. In addition, we observe that the proposed models are competitive with the state-of-the-art models, which are trained on a labeled dataset.

Selective Hole Filling of Red Blood Cells for Improved Marker-Controlled Watershed Segmentation.

Manual counting and evaluation of red blood cells with the presence of malaria parasites is a tiresome, time-consuming process that can be altered by environmental conditions and human error. Many algorithms were presented to segment red blood cells for subsequent parasitemia evaluation by machine learning algorithms. However, the segmentation of overlapping red blood cells always has been a challenge. Marker-controlled watershed segmentation is one of the methods that was implemented to separate overlapping red blood cells. However, a high number of overlapped red blood cells were still an issue. We propose a novel approach to improve the segmentation efficiency of marker-controlled watershed segmentation. Local minimum histogram background segmentation with a selective hole filling algorithm was introduced to improve segmentation efficiency of marker-controlled watershed segmentation on a high number of overlapping red blood cells. The local minimum was selected on the smoothed histogram for background segmentation. The combination of selective filling, convex hull, and Hough circle detection algorithms was utilized for the intact segmentation of red blood cells. The markers were computed from the resulted mask, and finally, marker-controlled watershed segmentation was applied to separate overlapping red blood cells. As a result, the proposed algorithm achieved higher background segmentation accuracy compared to popular background segmentation algorithms, and the inclusion of corner details improved watershed segmentation efficiency.