Article Preview
Top1. Introduction
Malaria is a benign blood disease caused by parasite Plasmodium. Malaria is one of the classic globally widespread but neglected infectious diseases. According to a 2017 World Malaria Report, data of about 216 million cases of malaria have been collected from 91 countries. The provisional epidemiological report states that as compared to the previous year, there is an increase of about 5 million infected positive cases with 44,5000 deaths over 2015. According to the National Strategic Plan (NSP) for Malaria Elimination (2017-22) reports, there is over 10 lakh infected positive cases in India’s 36 states and UTs, resulting in more than 331 deaths (World Health Organization, 2017).
Malaria is a complex rapidly growing disease and requires a number of steps in manual assessment leading to late diagnosis. The increase in a number of malaria distribution cases are unstable due to various factors such as the expansion of drug resistance; inter- and intra-observer variance; lack of qualitative assessment; reproducible measures to assess patient’s biopsies; global warming; and population mobility of different kinds (World Health Organization, 2017). It is difficult to visually inspect malaria and it has become a challenge to differentiate the infected vs non-infected erythrocytes from the enormous density of microscopic blood smear images. The visual inspection assessment of automated identification of malaria is subjective, time-consuming, prone to human error and requires experienced diagnostic hands. The experimental malaria diagnosis procedure generally includes histopathology of blood smear slide samples, where the big data deluge is reported (Jan and Zahoor, 2017). A computational microscopic evaluation of blood sample image without human intervention provides the best objective and efficient quality for malaria diagnosis (Das, Mukherjee, and Chakraborty, 2015).
During the last decades, based on detecting a polymorphism of Plasmodium parasites, their growth stages, and diagnosis via blood smear microscopic images, several computational techniques have emerged. Malaria-infected erythrocytes image analysis has a significant contribution to the technological development. Under these challenging contexts, the twelve (12) most popular morphological computational techniques have paved the way for better qualitative assessment towards personalized health care in pathology. This computational strategy builds a wholesome bond between the clinical and research experts, lessening down the burden between the two camps. Outcomes of these 12 morphological computational vision techniques have been highly valid and of excellent quality. Comparing among morphological techniques can provide greater effectiveness and efficiency for interpreting different malaria datasets and handling the challenge of intensity variations among different cell size and textures. Hence, the idea of automated computational assessment comes to light with intelligent supervisions of morphological techniques that can help to detect infected malaria erythrocytes in an image no matter where they are located.
This paper contributes to four major areas. The primary aim is to evolve an end-to-end framework that helps to learn feature representation from the data and to identify infected erythrocytes with time analysis. First, the study focuses on (i) detection and segmentation of blood smear image into infected vs. non-infected erythrocytes; and (ii) quality qualification performed via a random forest for classification in blood smear images. Second, a larger number of features or nodules are involved in this comparative study. Third, the effort is made here to achieve a solution to these challenges of subjectivity induced bias in erythrocytes representation by training an individual of computational morphological techniques on blood smear images using two different datasets. Finally, a comprehensive experiment with the comparative analysis is performed via two different films of datasets. Figure 1 illustrates a model of exhaustive malaria-infected erythrocyte detection using computational techniques with the random forest as a classifier.