Research Article: An on-chip instrument for white blood cells classification based on a lens-less shadow imaging technique

Date Published: March 28, 2017

Publisher: Public Library of Science

Author(s): Yuan Fang, Ningmei Yu, Runlong Wang, Dong Su, Giuseppe Chirico.


Routine blood tests provide important basic information for disease diagnoses. The proportions of three subtypes of white blood cells (WBCs), which are neutrophils, monocytes, lymphocytes, is key information for disease diagnosis. However, current instruments for routine blood tests, such as blood cell analyzers, flow cytometers, and optical microscopes, are cumbersome, time consuming and expensive. To make a smaller, automatic low-cost blood cell analyzer, much research has focused on a technique called lens-less shadow imaging, which can obtain microscopic images of cells in a lens-less system. Nevertheless, the efficiency of this imaging system is not satisfactory because of two problems: low resolution and imaging diffraction phenomena. In this paper, a novel method of classifying cells with the shadow imaging technique was proposed. It could be used for the classification of the three subtypes of WBCs, and the correlation of the results of classification between the proposed system and the reference system (BC-5180, Mindray) was 0.93. However, the instrument was only 10 × 10 × 10 cm, and the cost was less than $100. Depending on the lens-free shadow imaging technology, the main hardware could be integrated on a chip scale and could be called an on-chip instrument.

Partial Text

In medical fields, classification of white blood cells (WBCs) is a basic and important method to diagnose diseases [1–4]. There are three major types of WBCs (neutrophils, monocytes, lymphocytes) in human whole blood, and the proportions of these different types is basic information for medical diagnosis [5–7]. To determine the proportions of the different types of blood cells, classification is performed by microscopy, electrical impedance or laser light scattering. The traditional method uses a blood smear and Wright’s staining, and the different type of white cells are counted under the objective lens of a microscope [8]. In this method, microscopic images of stained cells are classified, and the staining and counting are time consuming. Recently, flow cytometry and hematology analyzers have been used instead of the manual method. These instruments use flowing liquid to focus on the cells that go through a laser beam and then capture spectral information, such as the forward and side scattering [9], or they drive cells through the micro-orifices to obtain the impedance change, which is called the Coulter effect [10]. The light scattering and implement changes provide information about the cells, such as size, internal complexity and granularity of cells, which enables the sorting of different cell types.

The feature vector of the image of a monocyte obtained with a 10× objective lens and the amplification effect of the shadow image of the lens-free shadow imaging system were similar to those obtained with a 4× objective lens. However, the low amplification of the image indicates low resolution, and the lower the resolution of the image, the less information provided. Fortunately, experiments demonstrated that the PPED was still useful with the low-resolution image. We used the 10× and the 4× objective lenses to capture two images of the same cell and extract the PPED vector (Fig 5B and 5D).

In summary, a WBC classification method based on shadow imaging of a CMOS sensor was proposed. The system contained inexpensive parts, such as the CMOS sensor, a white LED and an aluminum alloy shell, and the total cost of this system was below $100. A white LED was used in the proposed system, which means that the LED could be replaced by sunlight. The system power consumption would be significantly reduced and more conducive to POCT with sunlight. Due to the algorithms proposed in this paper, the system was not only cheaper but also automated. For the blood tests from whole blood samples from six outpatients (proportion of WBCs), the result obtained was a mean correlation index of 0.96. The mean errors of percentage of neutrophils (3.45%), monocytes (6.04%) and lymphocytes (6.7%) were lower than the method proposed by Mohendra Roy [14]. Although the performance was still less than traditional methods, it was sufficient to demonstrate that the proposed instrument and classification method were effective. The instrument was 10 × 10 × 10 cm in volume, but the aluminum alloy shell could be further reduced to chip size. The widespread application of the instruments would provide a more convenient method of routine blood examination.




0 0 vote
Article Rating
Notify of
Inline Feedbacks
View all comments