Label-Free White Blood Cell Classification Using Refractive Index Tomography and Deep Learning

Abstract

Objective and Impact Statement. We propose a rapid and accurate blood cell identification method exploiting deep learning and label-free refractive index (RI) tomography. Our computational approach that fully utilizes tomographic information of bone marrow (BM) white blood cell (WBC) enables us to not only classify the blood cells with deep learning but also quantitatively study their morphological and biochemical properties for hematology research. Introduction. Conventional methods for examining blood cells, such as blood smear analysis by medical professionals and fluorescence-activated cell sorting, require significant time, costs, and domain knowledge that could affect test results. While label-free imaging techniques that use a specimen’s intrinsic contrast (e.g., multiphoton and Raman microscopy) have been used to characterize blood cells, their imaging procedures and instrumentations are relatively time-consuming and complex. Methods. The RI tomograms of the BM WBCs are acquired via Mach-Zehnder interferometer-based tomographic microscope and classified by a 3D convolutional neural network. We test our deep learning classifier for the four types of bone marrow WBC collected from healthy donors (n=10): monocyte, myelocyte, B lymphocyte, and T lymphocyte. The quantitative parameters of WBC are directly obtained from the tomograms. Results. Our results show >99% accuracy for the binary classification of myeloids and lymphoids and >96% accuracy for the four-type classification of B and T lymphocytes, monocyte, and myelocytes. The feature learning capability of our approach is visualized via an unsupervised dimension reduction technique. Conclusion. We envision that the proposed cell classification framework can be easily integrated into existing blood cell investigation workflows, providing cost-effective and rapid diagnosis for hematologic malignancy.

Virtual Staining, Segmentation, and Classification of Blood Smears for Label-Free Hematology Analysis

“…Label-free techniques for quantitative analysis can address many of the limitations of conventional methods by eliminating the need for staining or exogenous contrast agents, thereby simplifying and speeding up the workflow. Several such techniques have been explored, including hyperspectral imaging [10], Raman microscopy [11], fluorescence lifetime imaging microscopy [12], and quantitative phase imaging [13][14][15][16]. While each method has its own unique advantages and disadvantages, there is a trade-off between the information provided by each method and its cost, complexity, and speed.…”

“…In this work, we take advantage of the capabilities of deep learning for segmentation [25,26], classification [15,27,28], and image-to-image translation [29][30][31][32] of label-free microscopy images, to develop an automated hematology analysis framework that operates on single-channel UV images acquired at 260 nm (having inherent nuclear contrast due to the absorption peak of nucleic acids), enabling simpler instrumentation and a factor of three improvement in imaging speed without sacrificing accuracy. Our virtual staining scheme accurately mimics the colors produced by the goldstandard Giemsa staining using only a single-channel image (single-wavelength imaging instead of multispectral imaging), unlike the pseudocolorization scheme introduced previously [21].…”

“…While several methods for segmentation and classification of WBCs have been proposed, most of them rely on feature extraction or training DNNs using stained images [35][36][37][38][39] or fail to provide an accurate five-part white blood cell differential [15,16,27,28,40]. Here, we present a segmentation method that uses only grayscale images (and is independent of the virtual staining branch of the pipeline) and has very high accuracy, with an average dice score of 0.9899 for cellular segmentation and 0.9718 for nuclear segmentation on an unseen test dataset.…”

Immunomagnetic leukocyte differential in whole blood on an electronic microdevice

“…Recently, alternative methods have been developed to address the limitations of the existing instrumentation by using machine learning for classification from cell images, 22 using refractive index tomography, 23 Raman spectroscopy, 24 miniaturized laser setups with fluorescent dyes [25][26][27][28] or lowcost systems such as smartphone-coupled paper-based assays 29,30 and microfluidic platforms with impedance measurements [31][32][33] and capture arrays. 34,35 However, these works were often unable to process the samples with erythrocytes present due to the overwhelming interference they would cause.…”

Automated Classification of Breast Cancer Cells Using High-Throughput Holographic Cytometry

“…QPI-based profiling has already demonstrated high-throughput capabilities and high accuracy levels in identifying cell-cycle phases, in ultrafast all-optical laser scanning approaches [10,11]. In other QPI approaches, deep learning has proven to be a powerful tool used by many for QPI classification [10][11][12][13], inference [14][15][16][17], and reconstruction [18][19][20]. We now seek to use ultra-high throughput QPI to build phenotypic profiles of single cells subjected to carcinogens, which can be, in combination with deep learning, used to develop prognostic biomarkers of precancerous cells.…”