Classification of Red Blood Cells using Principal Component Analysis Technique

Abstract

Principal component analysis (PCA) is based feature reduction that reduces the correlation of features. In this research, a novel approach is proposed by applying the PCA technique on various morphologies of red blood cells (RBCs). According to hematologists, this method successfully classified 40 different types of abnormal RBCs. The classification of RBCs into various distinct subtypes using three machine learning algorithms is important in clinical and laboratory tests for detecting blood diseases. The most common abnormal RBCs are considered as anemic. The RBC features are sufficient to identify the type of anemia and the disease that caused it. Therefore, we found that several features extracted from RBCs in the blood smear images are not significant for classification when observed independently but are significant when combined with other features. The number of feature vectors is reduced from 271 to 8 as time resuming in training and accuracy percentage increased to 98%.

AI based image analysis of red blood cells in oscillating microchannels

“…16,36 For example, Kihm et al divided the cells in two groups called “Slippers” and “Croissants”, 14 implemented a CNN (Convolutional Neural Network) to train 4000 images and classify the RBCs uniquely based on their shape characteristics. Considering the increasing demand for advancements and the potential for significant impact and popularity in this field, Recktenwald et al 37 and its follow-up study 38 adopted the approach, proposed by Kihm et al and Alkrimi et al , 14,15 to benchmark different AI techniques classifying RBCs and similarly to Kihm et al , 14 the classification was based fully on morphology. Lee et al 39 uses not just shape but also texture features to classify normal and abnormal RBCs and similarly to Das et al , 12 they classify cells in more than two categories.…”

“…Alkrimi et al 15 have classified RBCs using machine learning principal component analysis to reduce the correlation of features. Yet, the morphological analysis was done on a blood smear and not involving microfluidic channels.…”

“…Because there is a strong interest in the viscoelastic properties of the red blood cells oen only their shape has been analysed but not the healthiness of the cells. 14,15 The use of a channel to probe the mechanical properties of the red blood cells in conjunction with AI to detect a disease has, to our knowledge, only attempted once. 16 Here, we demonstrate an AI based image analysis using Ten-sorFlow that can decide between native, untreated red blood cells and red blood cells with chemically altered mechanical properties to mimic disease.…”

“…Because there is a strong interest in the viscoelastic properties of the red blood cells often only their shape has been analysed but not the healthiness of the cells. 14,15 The use of a channel to probe the mechanical properties of the red blood cells in conjunction with AI to detect a disease has, to our knowledge, only attempted once. 16…”