Human red blood cell recognition enhancement with three-dimensional morphological features obtained by digital holographic imaging

Differential diagnosis of hereditary anemias from a fraction of blood drop by digital holography and hierarchical machine learning

Anemia affects about the 25% of the global population and can provoke severe diseases, ranging from weakness and dizziness to pregnancy problems, arrhythmias and hearth failures. About 10% of the patients are affected by rare anemias of which 80% are hereditary. Early differential diagnosis of anemia enables prescribing patients a proper treatment and diet, which is effective to mitigate the associated symptoms. Nevertheless, the differential diagnosis of these conditions is often difficult due to shared and overlapping phenotypes. Indeed, the complete blood count and unaided peripheral blood smear observation cannot always provide a reliable differential diagnosis, so that biomedical assays and genetic tests are needed. These procedures are not error-free, require skilled personnel, and severely impact the financial resources of national health systems. Here we show a differential screening system for hereditary anemias that relies on holographic imaging and artificial intelligence. Label-free holographic imaging is aided by a hierarchical machine learning decider that works even in the presence of a very limited dataset but is enough accurate for discerning between different anemia classes with minimal morphological dissimilarities. It is worth to notice that only a few tens of cells from each patient are sufficient to obtain a correct diagnosis, with the advantage of significantly limiting the volume of blood drawn. This work paves the way to a wider use of home screening systems for point of care blood testing and telemedicine with lab-on-chip platforms.

Deep Learning-based Phenotypic Assessment of Red Cell Storage Lesions for Safe Transfusions

This study presents a novel approach to automatically perform instant phenotypic assessment of red blood cell (RBC) storage lesion in phase images obtained by digital holographic microscopy. The proposed model combines a generative adversarial network (GAN) with marker-controlled watershed segmentation scheme. The GAN model performed RBC segmentations and classifications to develop ageing markers, and the watershed segmentation was used to completely separate overlapping RBCs. Our approach achieved good segmentation and classification accuracy with a Dices coefficient of 0.94 at a high throughput rate of about 152 cells per second. These results were compared with other deep neural network architectures. Moreover, our image-based deep learning models recognized the morphological changes that occur in RBCs during storage. Our deep learning-based classification results were in good agreement with previous findings on the changes in RBC markers (dominant shapes) affected by storage duration. We believe that our image-based deep learning models can be useful for automated assessment of RBC quality, storage lesions for safe transfusions, and diagnosis of RBC-related diseases.

Quantitative phase imaging of erythrocyte in epilepsy patients

The present study focuses on the quantitative phase imaging of erythrocytes with the aim to compare the morphological differences between epilepsy patients under antiepileptic treatment, who have no other disease which may affect the erythrocyte morphology, and the healthy control group. The white light diffraction phase microscopy (WDPM) has been used to obtain the interferogram of the erythrocyte surfaces. The continuous wavelet transform with Paul wavelet has been chosen to calculate the surface profiles from this interferogram image. For the determination of alteration in morphology, besides WDPM, erythrocyte surfaces have been investigated by light microscope and scanning electron microscope. In this way, it has been possible to see the difference in terms of precision and implementation between the most commonly used methods with regard to the quantitative phase imaging. Erythrocytes from all the samples have been examined and displayed in both two- and three-dimensional way. We have observed that erythrocytes of patients with effective antiepileptic blood levels were more affected in morphology than healthy subjects. When we compared the erythrocyte morphological changes of patients who received monotherapy or polytherapy, no difference was observed. In conclusion, antiepileptic drugs (AEDs) cause red blood cell (RBC) morphological changes and a combined usage of WDPM with Paul wavelet and light microscopy methods are very convenient for studying the erythrocyte morphologies on multiple patients.

Automatic detection and characterization of quantitative phase images of thalassemic red blood cells using a mask region-based convolutional neural network

SIGNIFICANCE

Label-free quantitative phase imaging is a promising technique for the automatic detection of abnormal red blood cells (RBCs) in real time. Although deep-learning techniques can accurately detect abnormal RBCs from quantitative phase images efficiently, their applications in diagnostic testing are limited by the lack of transparency. More interpretable results such as morphological and biochemical characteristics of individual RBCs are highly desirable.

AIM

An end-to-end deep-learning model was developed to efficiently discriminate thalassemic RBCs (tRBCs) from healthy RBCs (hRBCs) in quantitative phase images and segment RBCs for single-cell characterization.

APPROACH

Two-dimensional quantitative phase images of hRBCs and tRBCs were acquired using digital holographic microscopy. A mask region-based convolutional neural network (Mask R-CNN) model was trained to discriminate tRBCs and segment individual RBCs. Characterization of tRBCs was achieved utilizing SHapley Additive exPlanation analysis and canonical correlation analysis on automatically segmented RBC phase images.

RESULTS

The implemented model achieved 97.8% accuracy in detecting tRBCs. Phase-shift statistics showed the highest influence on the correct classification of tRBCs. Associations between the phase-shift features and three-dimensional morphological features were revealed.

CONCLUSIONS

The implemented Mask R-CNN model accurately identified tRBCs and segmented RBCs to provide single-RBC characterization, which has the potential to aid clinical decision-making.

Quantitative analysis of three-dimensional morphology and membrane dynamics of red blood cells during temperature elevation

The optimal functionality of red blood cells is closely associated with the surrounding environment. This study was undertaken to analyze the changes in membrane profile, mean corpuscular hemoglobin (MCH), and cell membrane fluctuations (CMF) of healthy red blood cells (RBC) at varying temperatures. The temperature was elevated from 17 °C to 41 °C within a duration of less than one hour, and the holograms were recorded by an off-axis configuration. After hologram reconstruction, we extracted single RBCs and evaluated their morphologically related features (projected surface area and sphericity coefficient), MCH, and CMF. We observed that elevating the temperature results in changes in the three-dimensional (3D) profile. Since CMF amplitude is highly correlated to the bending curvature of RBC membrane, temperature-induced shape changes can alter CMF’s map and amplitude; mainly larger fluctuations appear on dimple area at a higher temperature. Regardless of the shape changes, no alterations in MCH were seen with temperature variation.

Automated quantitative analysis of multiple cardiomyocytes at the single-cell level with three-dimensional holographic imaging informatics

Cardiomyocytes derived from human pluripotent stem cells are a promising tool for disease modeling, drug compound testing, and cardiac toxicity screening. Bio-image segmentation is a prerequisite step in cardiomyocyte image analysis by digital holography (DH) in microscopic configuration and has provided satisfactory results. In this study, we quantified multiple cardiac cells from segmented 3-dimensional DH images at the single-cell level and measured multiple parameters describing the beating profile of each individual cell. The beating profile is extracted by monitoring dry-mass distribution during the mechanical contraction-relaxation activity caused by cardiac action potential. We present a robust two-step segmentation method for cardiomyocyte low-contrast image segmentation based on region and edge information. The segmented single-cell contains mostly the nucleus of the cell since it is the best part of the cardiac cell, which can be perfectly segmented. Clustering accuracy was assessed by a silhouette index evaluation for k-means clustering and the Dice similarity coefficient (DSC) of the final segmented image. 3D representation of single of cardiomyocytes. The cell contains mostly the nucleus section and a small area of cytoplasm.

Quantification of stored red blood cell fluctuations by time-lapse holographic cell imaging

We propose methods to quantitatively calculate the fluctuation rate of red blood cells with nanometric axial and millisecond temporal sensitivity at the single-cell level by using time-lapse holographic cell imaging. For this quantitative analysis, cell membrane fluctuations (CMFs) were measured for RBCs stored at different storage times. Measurements were taken over the whole membrane for both the ring and dimple sections separately. The measurements show that healthy RBCs that maintain their discocyte shape become stiffer with storage time. The correlation analysis demonstrates a significant negative correlation between CMFs and the sphericity coefficient, which characterizes the morphological type of erythrocyte. In addition, we show the correlation results between CMFs and other morphological properties such as projected surface area, surface area, mean corpuscular volume, and mean corpuscular hemoglobin.

Automated three-dimensional morphology-based clustering of human erythrocytes with regular shapes: stomatocytes, discocytes, and echinocytes

We present unsupervised clustering methods for automatic grouping of human red blood cells (RBCs) extracted from RBC quantitative phase images obtained by digital holographic microscopy into three RBC clusters with regular shapes, including biconcave, stomatocyte, and sphero-echinocyte. We select some good features related to the RBC profile and morphology, such as RBC average thickness, sphericity coefficient, and mean corpuscular volume, and clustering methods, including density-based spatial clustering applications with noise, k-medoids, and k-means, are applied to the set of morphological features. The clustering results of RBCs using a set of three-dimensional features are compared against a set of two-dimensional features. Our experimental results indicate that by utilizing the introduced set of features, two groups of biconcave RBCs and old RBCs (suffering from the sphero-echinocyte process) can be perfectly clustered. In addition, by increasing the number of clusters, the three RBC types can be effectively clustered in an automated unsupervised manner with high accuracy. The performance evaluation of the clustering techniques reveals that they can assist hematologists in further diagnosis.