Analysis of Vision-based Abnormal Red Blood Cell Classification

Wong, Annika; Anantrasirichai, Nantheera; Chalidabhongse, Thanarat H.; Palasuwan, Duangdao; Palasuwan, Attakorn; Bull, David

doi:10.48550/arxiv.2106.00389

Cited by 3 publications

(6 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A total of 1000 samples were tested in this experiment, that is, 100 test samples for A comparison of different strategies with the proposed HPKNN classifier is shown in Table 6. The proposed HPKNN (97.86%) outperforms SVM-SOMTE 1 + CSL (76.09%), 33 TabNet-SOMTE 1 + CSL(69.7%), 33 optimizable SVM (97.29%), and optimizable Tree (95.98%) in terms of sensitivity analysis. The proposed HPKNN (97.86%) outperforms SVM-SOMTE 1 + CSL (76.73%), TabNet-SOMTE 1 + CSL (69.56%), and Effi-cientNet (91.04%), 3 optimizable SVM (96.98%), and optimizable Tree (95.87%) in terms of F 1 -score analysis.…”

Section: Robustness Analysis Of the Experimental Datasetsmentioning

confidence: 97%

HPKNN: Hyper‐parameter optimized KNN classifier for classification of poikilocytosis

Dhar

Kothandapani

Satti

et al. 2023

Int J Imaging Syst Tech

View full text Add to dashboard Cite

Red blood cells (RBCs) are the more crucial components of a healthy organism.Poikilocytosis is a disorder in which blood cells lose their standard shapes. The human bodies are affected by disorders like anemia and thalassemia because the structure of RBCs shape changes. Hematologists take longer time to detect abnormalities in RBCs. The proposed hyper-parameter optimizable K nearest neighbor (HPKNN) algorithm detects and classifies numerous types of poikilocytosis quickly and accurately. On the Chula-PIC-Lab and Erythrocytes IDB datasets, the proposed HPKNN algorithm was tested. After the preprocessing stage, the Freeman chain code has been utilized to retrieve the exact boundary shape of RBC morphology. On datasets of Erythrocytes IDB 1, 2, 3, and Chula-PIC-Lab(Chula-RBC-12), the morphological tools and Freeman chain code based segmentation provide 97.44% and 94.45% accuracy, respectively, which is superior than other methods. The proposed HPKNN classifier has an overall sensitivity of 97.9%, precision of 97.9%, and an f 1 score of 97.9% on the Chula-PIC-Lab (Chula-RBC-12) dataset to classify 10 classes, namely acanthocyte, dacrocyte, echinocyte, elliptocyte, normal, overlapped, schistocyte, spherocyte, stomatocyte, and target cell, which is greater than any other current approach. The HPKNN classifier achieves 98.085% precision, 98.085% sensitivity or recall, and 98.085% f 1 -score values on the Erythrocytes IDB datasets to classify circular, elongated, and other RBC classes.anemia, Freeman chain code, KNN classifier, poikilocytosis | INTRODUCTIONPoikilocytosis means the RBC blood cells are abnormal due to the changes in their standard shape. 1 The normal RBC blood cells look disk-shaped, biconcave, and trim. The RBC blood smear cells have no nucleus, lysosome, mitochondria, and endoplasmic. The diameter of normal RBC cells is 6.2 to 8.2 μ. 2 Generally, it is observed that the diameter of normal RBC cells is 7.2 μ. 2 The RBC cells are responsible for carrying different nutrients and oxygen to our body's organs and tissues. When RBCs' shape varies from normal, they are not having enough nutrients and oxygen. We are suffering from lots of diseases because of a lack of oxygen and nutrients. Poikilocytosis affects the body due to trauma or abnormalities occurring in the membrane. The poikilocytosis abnormality that occurs due to trauma is blister and keratocytes, teardrop cells, schistocytes, pyro-poikilocytes, and semilunar bodies. Poikilocytosis due to abnormalities in the membrane are target cells, burr cells, mouth

show abstract

Section: Robustness Analysis Of the Experimental Datasetsmentioning

confidence: 97%

HPKNN: Hyper‐parameter optimized KNN classifier for classification of poikilocytosis

Dhar

Kothandapani

Satti

et al. 2023

Int J Imaging Syst Tech

View full text Add to dashboard Cite

show abstract

“…Fortunately, rapid advancements in the field of machine learning have yielded deep convolutional neural network architectures (CNNs) that surpassed all former approaches to image classification tasks [15][16][17][18][19] . Most scientific fields have by now felt the impact of deep learning (DL) enabled image analysis and there have already been several attempts to utilize CNNs for classifying morphology of RBCs in different contexts [20][21][22][23] . Although these previous studies showed varying degrees of success, each approach has been tailored to a specific imaging modality, and the common lack of dataset and code accessibility made it impossible to compare the approaches.…”

mentioning

confidence: 99%

Deep ensemble learning enables highly accurate classification of stored red blood cell morphology

Routt

Yang

Piety

et al. 2023

Sci Rep

View full text Add to dashboard Cite

Changes in red blood cell (RBC) morphology distribution have emerged as a quantitative biomarker for the degradation of RBC functional properties during hypothermic storage. Previously published automated methods for classifying the morphology of stored RBCs often had insufficient accuracy and relied on proprietary code and datasets, making them difficult to use in many research and clinical applications. Here we describe the development and validation of a highly accurate open-source RBC morphology classification pipeline based on ensemble deep learning (DL). The DL-enabled pipeline utilized adaptive thresholding or semantic segmentation for RBC identification, a deep ensemble of four convolutional neural networks (CNNs) to classify RBC morphology, and Kalman filtering with Hungarian assignment for tracking changes in the morphology of individual RBCs over time. The ensembled CNNs were trained and evaluated on thousands of individual RBCs from two open-access datasets previously collected to quantify the morphological heterogeneity and washing-induced shape recovery of stored RBCs. Confusion matrices and reliability diagrams demonstrated under-confidence of the constituent models and an accuracy of about 98% for the deep ensemble. Such a high accuracy allowed the CNN ensemble to uncover new insights over our previously published studies. Re-analysis of the datasets yielded much more accurate distributions of the effective diameters of stored RBCs at each stage of morphological degradation (discocyte: 7.821 ± 0.429 µm, echinocyte 1: 7.800 ± 0.581 µm, echinocyte 2: 7.304 ± 0.567 µm, echinocyte 3: 6.433 ± 0.490 µm, sphero-echinocyte: 5.963 ± 0.348 µm, spherocyte: 5.904 ± 0.292 µm, stomatocyte: 7.080 ± 0.522 µm). The effective diameter distributions were significantly different across all morphologies, with considerable effect sizes for non-neighboring classes. A combination of morphology classification with cell tracking enabled the discovery of a relatively rare and previously overlooked shape recovery of some sphero-echinocytes to early-stage echinocytes after washing with 1% human serum albumin solution. Finally, the datasets and code have been made freely available online to enable replication, further improvement, and adaptation of our work for other applications.

show abstract

“…The average precisions are 90.25%, 80.41%, and 98.92% for platelets, RBCs, and WBCs respectively. (Wong et al, 2021) proposed RBC classification with 3 methods. Our dataset also was used in this work.…”

Section: Previous Studiesmentioning

confidence: 99%

“…The number near each cell shows the number of predicted RBC types. The number is in (Wong et al, 2021) reported the results of using SVM and TabNet to classify the RBCs into 11 classes on the same dataset as used in this paper. They employed the SMOTE technique with cost-sensitive learning to handle the imbalanced dataset.…”

Section: Rbc Classificationmentioning

confidence: 99%

“…In addition, blood smear slides may not have the same environment, such as lighting, zoom scale, and camera. Recently, a deep convolutional neural network for object detection and semantic segmentation has begun to be used for RBC detection (Qiu et al, 2019;Shakarami, Menhaj, Mahdavi-Hormat, & Tarrah, 2021;Wong et al, 2021). The benefits of these deep end-to-end methods are that, first, they allow training a possibly complex learning system represented by a single model, bypassing the intermediate layers usually found in the traditional pipeline approach.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Red blood cell segmentation and classification from microscopic images using machine learning

Naruenatthanaset¹

View full text Add to dashboard Cite

Red blood cell morphology analysis plays an essential role in diagnosing many diseases caused by RBC disorders. This manual inspection is a long process and requires practice and experience. Since recent computer vision and image processing in the medical imaging area can provide efficient tools, it can help hematologists to automatically analyze images from a microscope in a reduced time and cost. This research presents a new method to segment and classify RBCs from blood smear images. The process started from data collection, which a new application was created for precisely labeling. The normalization was done to reduce the color space and allowed the trained model to not be biased on color. Then, overlapping cells were separated using a new method to find concave points and use direct ellipse fitting to estimate the shape of a single RBC. Lastly, classification using EfficientNet-B1 on 12 red blood cell classes was done. However, to classify multiple classes with deep learning, imbalance problems are common in medical imaging because number of normal samples is always higher than number of rare disease samples. The imbalanced handling techniques were analyzed to deal with this problem. Experimental results showed that the weight balancing technique with augmentation had the potential to deal with imbalance problems.

show abstract

Analysis of Vision-based Abnormal Red Blood Cell Classification

Cited by 3 publications

References 28 publications

HPKNN: Hyper‐parameter optimized KNN classifier for classification of poikilocytosis

HPKNN: Hyper‐parameter optimized KNN classifier for classification of poikilocytosis

Deep ensemble learning enables highly accurate classification of stored red blood cell morphology

Red blood cell segmentation and classification from microscopic images using machine learning

Contact Info

Product

Resources

About