skip to main content

Title: A Deep Learning Framework for Sickle Cell Disease Microfluidic Biomarker Assays
Introduction: Vaso-occlusive crises (VOCs) are a leading cause of morbidity and early mortality in individuals with sickle cell disease (SCD). These crises are triggered by sickle red blood cell (sRBC) aggregation in blood vessels and are influenced by factors such as enhanced sRBC and white blood cell (WBC) adhesion to inflamed endothelium. Advances in microfluidic biomarker assays (i.e., SCD Biochip systems) have led to clinical studies of blood cell adhesion onto endothelial proteins, including, fibronectin, laminin, P-selectin, ICAM-1, functionalized in microchannels. These microfluidic assays allow mimicking the physiological aspects of human microvasculature and help characterize biomechanical properties of adhered sRBCs under flow. However, analysis of the microfluidic biomarker assay data has so far relied on manual cell counting and exhaustive visual morphological characterization of cells by trained personnel. Integrating deep learning algorithms with microscopic imaging of adhesion protein functionalized microfluidic channels can accelerate and standardize accurate classification of blood cells in microfluidic biomarker assays. Here we present a deep learning approach into a general-purpose analytical tool covering a wide range of conditions: channels functionalized with different proteins (laminin or P-selectin), with varying degrees of adhesion by both sRBCs and WBCs, and in both normoxic and hypoxic environments. Methods: Our neural more » networks were trained on a repository of manually labeled SCD Biochip microfluidic biomarker assay whole channel images. Each channel contained adhered cells pertaining to clinical whole blood under constant shear stress of 0.1 Pa, mimicking physiological levels in post-capillary venules. The machine learning (ML) framework consists of two phases: Phase I segments pixels belonging to blood cells adhered to the microfluidic channel surface, while Phase II associates pixel clusters with specific cell types (sRBCs or WBCs). Phase I is implemented through an ensemble of seven generative fully convolutional neural networks, and Phase II is an ensemble of five neural networks based on a Resnet50 backbone. Each pixel cluster is given a probability of belonging to one of three classes: adhered sRBC, adhered WBC, or non-adhered / other. Results and Discussion: We applied our trained ML framework to 107 novel whole channel images not used during training and compared the results against counts from human experts. As seen in Fig. 1A, there was excellent agreement in counts across all protein and cell types investigated: sRBCs adhered to laminin, sRBCs adhered to P-selectin, and WBCs adhered to P-selectin. Not only was the approach able to handle surfaces functionalized with different proteins, but it also performed well for high cell density images (up to 5000 cells per image) in both normoxic and hypoxic conditions (Fig. 1B). The average uncertainty for the ML counts, obtained from accuracy metrics on the test dataset, was 3%. This uncertainty is a significant improvement on the 20% average uncertainty of the human counts, estimated from the variance in repeated manual analyses of the images. Moreover, manual classification of each image may take up to 2 hours, versus about 6 minutes per image for the ML analysis. Thus, ML provides greater consistency in the classification at a fraction of the processing time. To assess which features the network used to distinguish adhered cells, we generated class activation maps (Fig. 1C-E). These heat maps indicate the regions of focus for the algorithm in making each classification decision. Intriguingly, the highlighted features were similar to those used by human experts: the dimple in partially sickled RBCs, the sharp endpoints for highly sickled RBCs, and the uniform curvature of the WBCs. Overall the robust performance of the ML approach in our study sets the stage for generalizing it to other endothelial proteins and experimental conditions, a first step toward a universal microfluidic ML framework targeting blood disorders. Such a framework would not only be able to integrate advanced biophysical characterization into fast, point-of-care diagnostic devices, but also provide a standardized and reliable way of monitoring patients undergoing targeted therapies and curative interventions, including, stem cell and gene-based therapies for SCD. Disclosures Gurkan: Dx Now Inc.: Patents & Royalties; Xatek Inc.: Patents & Royalties; BioChip Labs: Patents & Royalties; Hemex Health, Inc.: Consultancy, Current Employment, Patents & Royalties, Research Funding. « less
; ; ; ; ; ; ; ;
Award ID(s):
Publication Date:
Journal Name:
Supplement 1
Page Range or eLocation-ID:
15 to 16
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Sickle cell disease (SCD) is a major public health priority throughout much of the world, affecting millions of people. In many regions, particularly those in resource-limited settings, SCD is not consistently diagnosed. In Africa, where the majority of SCD patients reside, more than 50% of the 0.2–0.3 million children born with SCD each year will die from it; many of these deaths are in fact preventable with correct diagnosis and treatment. Here, we present a deep learning framework which can perform automatic screening of sickle cells in blood smears using a smartphone microscope. This framework uses two distinct, complementarymore »deep neural networks. The first neural network enhances and standardizes the blood smear images captured by the smartphone microscope, spatially and spectrally matching the image quality of a laboratory-grade benchtop microscope. The second network acts on the output of the first image enhancement neural network and is used to perform the semantic segmentation between healthy and sickle cells within a blood smear. These segmented images are then used to rapidly determine the SCD diagnosis per patient. We blindly tested this mobile sickle cell detection method using blood smears from 96 unique patients (including 32 SCD patients) that were imaged by our smartphone microscope, and achieved ~98% accuracy, with an area-under-the-curve of 0.998. With its high accuracy, this mobile and cost-effective method has the potential to be used as a screening tool for SCD and other blood cell disorders in resource-limited settings.

    « less
  2. Sickle cell anemia (SCA) is a disease that affects red blood cells (RBCs). Healthy RBCs are highly deformable objects that under flow can penetrate blood capillaries smaller than their typical size. In SCA there is an impaired deformability of some cells, which are much stiffer and with a different shape than healthy cells, and thereby affect regular blood flow. It is known that blood from patients with SCA has a higher viscosity than normal blood. However, it is unclear how the rigidity of cells is related to the viscosity of blood, in part because SCA patients are often treated withmore »transfusions of variable amounts of normal RBCs and only a fraction of cells will be stiff. Here, we report systematic experimental measurements of the viscosity of a suspension varying the fraction of rigid particles within a suspension of healthy cells. We also perform systematic numerical simulations of a similar mixed suspension of soft RBCs, rigid particles, and their hydrodynamic interactions. Our results show that there is a rheological signature within blood viscosity to clearly identify the fraction of rigidified cells among healthy deformable cells down to a 5% volume fraction of rigidified cells. Although aggregation of RBCs is known to affect blood rheology at low shear rates, and our simulations mimic this effect via an adhesion potential, we show that such adhesion, or aggregation, is unlikely to provide a physical rationalization for the viscosity increase observed in the experiments at moderate shear rates due to rigidified cells. Through numerical simulations, we also highlight that most of the viscosity increase of the suspension is due to the rigidity of the particles rather than their sickled or spherical shape. Our results are relevant to better characterize SCA, provide useful insights relevant to rheological consequences of blood transfusions, and, more generally, extend to the rheology of mixed suspensions having particles with different rigidities, as well as offering possibilities for developments in the field of soft material composites.« less
  3. Red blood cells (RBCs) are subjected to recurrent changes in shear stress and oxygen tension during blood circulation. The cyclic shear stress has been identified as an important factor that alone can weaken cell mechanical deformability. The effects of cyclic hypoxia on cellular biomechanics have yet to be fully investigated. As the oxygen affinity of hemoglobin plays a key role in the biological function and mechanical performance of RBCs, the repeated transitions of hemoglobin between its R (high oxygen tension) and T (low oxygen tension) states may impact their mechanical behavior. The present study focuses on developing a novel microfluidics-basedmore »assay for characterization of the effect of cyclic hypoxia on cell biomechanics. The capability of this assay is demonstrated by a longitudinal study of individual RBCs in health and sickle cell disease subjected to cyclic hypoxia conditions of various durations and levels of low oxygen tension. Viscoelastic properties of cell membranes are extracted from tensile stretching and relaxation processes of RBCs induced by the electrodeformation technique. Results demonstrate that cyclic hypoxia alone can significantly reduce cell deformability, similar to the fatigue damage accumulated through cyclic mechanical loading. RBCs affected by sickle cell disease are less deformable (significantly higher membrane shear modulus and viscosity) than normal RBCs. The fatigue resistance of sickle RBCs to the cyclic hypoxia challenge is significantly inferior to normal RBCs, and this trend is more significant in mature erythrocytes of sickle cells. When oxygen affinity of sickle hemoglobin is enhanced by anti-sickling drug treatment of 5-hydroxymethyl-2-furfural (5-HMF), sickle RBCs show ameliorated resistance to fatigue damage induced by cyclic hypoxia. These results illustrate that an important biophysical mechanism underlying RBC senescence in which cyclic hypoxia challenge alone can lead to mechanical degradation of the RBC membrane. We envision the application of this assay can be further extended to RBCs in other blood diseases and other types of cells.« less
  4. Profiling circulating tumour cells (CTCs) in cancer patients' blood samples is critical to understand the complex and dynamic nature of metastasis. This task is challenged by the fact that CTCs are not only extremely rare in circulation but also highly heterogeneous in their molecular programs and cellular functions. Here we report a combinational approach for the simultaneous biochemical and functional phenotyping of patient-derived CTCs, using an integrated inertial ferrohydrodynamic cell separation (i 2 FCS) method and a single-cell microfluidic migration assay. This combinatorial approach offers unique capability to profile CTCs on the basis of their surface expression and migratory characteristics.more »We achieve this using the i 2 FCS method that successfully processes whole blood samples in a tumor cell marker and size agnostic manner. The i 2 FCS method enables an ultrahigh blood sample processing throughput of up to 2 × 10 5 cells s −1 with a blood sample flow rate of 60 mL h −1 . Its short processing time (10 minutes for a 10 mL sample), together with a close-to-complete CTC recovery (99.70% recovery rate) and a low WBC contamination (4.07-log depletion rate by removing 99.992% of leukocytes), results in adequate and functional CTCs for subsequent studies in the single-cell migration device. For the first time, we employ this new approach to query CTCs with single-cell resolution in accordance with their expression of phenotypic surface markers and migration properties, revealing the dynamic phenotypes and the existence of a high-motility subpopulation of CTCs in blood samples from metastatic lung cancer patients. This method could be adopted to study the biological and clinical value of invasive CTC phenotypes.« less
  5. Background and objectives Previous work has identified that gene expression differences in cell adhesion pathways exist between humans and chimpanzees. Here, we used a comparative cell biology approach to assay interspecies differences in cell adhesion phenotypes in order to better understand the basic biological differences between species’ epithelial cells that may underly the organism-level differences we see in wound healing and cancer. Methodology We used skin fibroblast cell lines from humans and chimpanzees to assay cell adhesion and migration. We then utilized published RNA-Seq data from the same cell lines exposed to a cancer / wound-healing mimic to determine whatmore »gene expression changes may be corresponding to altered cellular adhesion dynamics between species. Results The functional adhesion and migration assays revealed that chimpanzee fibroblasts adhered sooner and remained adherent for significantly longer and move into a “wound” at faster rate than human fibroblasts. The gene expression data suggest that the enhanced adhesive properties of chimpanzee fibroblasts may be due to chimpanzee fibroblasts exhibiting significantly higher expression of cell and focal adhesion molecule genes than human cells, both during a wound healing assay and at rest. Conclusions and implications Chimpanzee fibroblasts exhibit stronger adhesion and greater cell migration than human fibroblasts. This may be due to divergent gene expression of focal adhesion and cell adhesion molecules, such as integrins, laminins, and cadherins, as well as ECM proteins like collagens. This is one of few studies demonstrating that these divergences in gene expression between closely related species can manifest in fundamental differences in cell biology. Our results provide better insight into species-specific cell biology phenotypes and how they may influence more complex traits, such as cancer metastasis and wound healing.« less