Chapter 6 - Deep learning methodology proposal for the classification of erythrocytes and leukocytes

Abstract

Improving diagnostics has been one of the most important challenges facing health systems in recent years, due to innovation and scientific research, image classification techniques, medical information management systems brought about by health technology, and developments never seen before that enable major breakthroughs in the areas of prevention, early disease detection, and disease control. Blood is investigated through blood count, which is subordinate to manual and/or automated procedures. This reliance on medical fields for new innovations directs this chapter to the development of a deep learning framework for classifying white cell subtypes in digital images that fits the criteria for reliability and efficiency of blood cell detection, making the methodology more accessible to diverse populations. Because laboratory medical examinations are often costly and inaccessible to populations of underdeveloped and developing countries, it is of great importance to create tools that facilitate the obtaining of medical reports at low cost and with high reliability. For this, Python using a Jupyter notebook was developed and experiments were conducted using a dataset containing 12,500 digital images of human blood smear fields comprising nonpathological leukocytes. As a result, the accuracy of the approach was 86.17%, demonstrating the high reliability of the methodology. Therefore the algorithm is considered a reliable, accurate, and inexpensive method and can be employed as the third most practicable procedure for blood counts for often underprivileged people of developing and underdeveloped countries.