Counting White Blood Cells from a Blood Smear Using Fourier Ptychographic Microscopy

Abstract

White blood cell (WBC) count is a valuable metric for assisting with diagnosis or prognosis of various diseases such as coronary heart disease, type 2 diabetes, or infection. Counting WBCs can be done either manually or automatically. Automatic methods are capable of counting a large number of cells to give a statistically more accurate reading of the WBC count of a sample, but the specialized equipment tends to be expensive. Manual methods are inexpensive since they only involve a conventional light microscope setup. However, it is more laborious and error-prone because the small field-of-view (FOV) of the microscope necessitates mechanical scanning of a specimen for counting an adequate number of WBCs. Here, we investigate the use of Fourier ptychographic microscopy (FPM) to bypass these issues of the manual methods. With a 2x objective, FPM can provide a FOV of 120 mm2 with enhanced resolution comparable to that of a 20x objective, which is adequate for non-differentially counting WBCs in just one FOV. A specialist was able to count the WBCs in FPM images with 100% accuracy compared to the count as determined from conventional microscope images. An automatic counting algorithm was also developed to identify WBCs from FPM's captured images with 95% accuracy, paving the way for a cost-effective WBC counting setup with the advantages of both the automatic and manual counting methods.

Fig 1. FPM setup.

a) The schematics of FPM setup. It consists of a conventional microscope, involving an objective, tube lens, a camera, and an LED matrix replacing the condenser for specimen illumination. b) Fourier spectrum of the sample at the objective’s back-focal plane. The circular subregion corresponds to the objective’s aperture size. It shifts with the shifting illumination angle. c) Shifting illumination angle is provided by illuminating LEDs at different locations on the matrix.

Fig 2. The full FOV image of a blood smear slide with SBP of ~0.9 gigapixel by FPM using 2x 0.08 NA objective (center) and its magnified subregions (surrounding).

The red subregions are directly extracted from the full FOV image, and the corresponding blue subregions are taken by a 20x 0.5 NA objective. They both offer enough details from which we can discern the shapes and sizes of the WBCs and their nuclei. The blue circle in the middle of the center image represents the relative size of the FOV achieved by a 20x objective compared to that of FPM.

Fig 3. Several white blood cells from the imaged regions by a 20x conventional microscope (above) and FPM (below).

Different morphologies can be observed among these cells. The left-most cell has an eccentric, single-lobed nucleus, suggesting that it is a lymphocyte, while other cells display multi-lobed nuclei structure, suggesting that they are eosinophils, basophils, or neutrophils.

Fig 4. WBC detection by the automatic counting algorithm.

The algorithm analyzes a section of the specimen image obtained by FPM (left), and outputs the image overlaid with red markings on the detected WBCs.