Automated in-vitro cell detection and counting have been a key theme for artificial and intelligent biological analysis such as biopsy, drug analysis and decease diagnosis. Along with the rapid development of microfluidics and lab-on-chip technologies, in-vitro live cell analysis has been one of the critical tasks for both research and industry communities. However, it is a great challenge to obtain and then predict the precise information of live cells from numerous microscopic videos and images. In this paper, we investigated in-vitro detection of white blood cells using deep neural networks, and discussed how state-of-the-art machine learning techniques could fulfil the needs of medical diagnosis. The approach we used in this study was based on Faster Region-based Convolutional Neural Networks (Faster RCNNs), and a transfer learning process was applied to apply this technique to the microscopic detection of blood cells. Our experimental results demonstrated that fast and efficient analysis of blood cells via automated microscopic imaging can achieve much better accuracy and faster speed than the conventionally applied methods, implying a promising future of this technology to be applied to the microfluidic point-of-care medical devices.
Deep Dive into Automated Blood Cell Detection and Counting via Deep Learning for Microfluidic Point-of-Care Medical Devices.
Automated in-vitro cell detection and counting have been a key theme for artificial and intelligent biological analysis such as biopsy, drug analysis and decease diagnosis. Along with the rapid development of microfluidics and lab-on-chip technologies, in-vitro live cell analysis has been one of the critical tasks for both research and industry communities. However, it is a great challenge to obtain and then predict the precise information of live cells from numerous microscopic videos and images. In this paper, we investigated in-vitro detection of white blood cells using deep neural networks, and discussed how state-of-the-art machine learning techniques could fulfil the needs of medical diagnosis. The approach we used in this study was based on Faster Region-based Convolutional Neural Networks (Faster RCNNs), and a transfer learning process was applied to apply this technique to the microscopic detection of blood cells. Our experimental results demonstrated that fast and efficient a
Microfluidic technologies [1,2] have recently found wide-range applications in biological and medical applications, such as lab-on-chip and point-of-care (POC) diagnostic devices, which revolutionised the personalized medicine and rapid disease diagnosis. Point-of-care testing (or bedside testing) is generally defined as medical diagnostic testing at or near the point of caor in other words, at the time and place of patient care. This contrasts with the conventional treatment, in which testing was wholly or mostly confined to the medical laboratory. In this case, the specimens are often taken away from the point of care and then hours even days will be waited for the results, during which the point of care is asked to wait before the critical information is obtained.
Such the POC diagnosis has facilitated a paradigm shift from therapeutic treatments to predictive, personalized and preventive ones [1,2], and from the conventional diagnostic tests performed inside the clinical labs settings to near-patient ones. This will enable doctors to have timely diagnostic information to make quick decisions regarding to further diagnosis or immediate treatments. At the same time, patients are hugely benefited from the ease of usages of the POC devices, which allow them to personally monitor their own health in reliable and quantified ways, simply being at home. There is no requirement for the tested samples to be delivered to a lab and no need for the results to be transmitted physically manually or over long distance and time. Doctors, nurses, or even patients could perform the tests and immediately receives the results on the spot, thereby, saving huge amount of time with the rapid diagnosis.
Among various lab-on-chip based diagnosis tasks [1,2], microfluidics based blood cell diagnostics as shown in Fig. 1 is one of critical ones. White blood cell (WBC) counting, a routine test to measure the number of white blood cells for a patient, is an important means for medical diagnosis. A low WBC count may be linked to a toxic reaction, a viral infection, or side-effect from chemotherapy, or a disease in the bone marrow which limits the body’s ability to produce the normal WBCs, whereas a high WBC count might be linked to an imperative sign of an infection or leukaemia.
In hospitals, the conventional diagnostic procedures for blood cell analysis involve microscopic inspection of peripheral blood samples. For decades, this operation is performed by experienced operators with two main analysis procedures: classification and counting of the cells. However, lacking of automation and intelligent procedures has become a critical barrier for the integration of microscopic image analysis into microfluidic POC diagnostic system [3,4]. So far, blood cell counting in the commonly used POC devices is mostly done using flow cytometry, which is based on monitoring changes of the fluorescence signals as a quantitative tool [4], however, this is far inferior and less accurate compared to cell-by-cell counting based on the microscopic images. As it is well-known, microscopic cell counting is non-invasive and free of fluorescence dyes, and it is based on the lab-onchip devices and hence could drastically reduce the cost of the POC diagnostics.
In this work, aiming at the integration of automated microscopic image analysis into microfluidic POC device for blood cell counting, we investigated the methods of live cell detection techniques based on a recently developed artificial intelligence (AI) approach. Live-cell imaging experiments [3][4][5][6] provide the possibilities of identification, tracking, and analysis of cells from the microscopic images using a computer vision-based AI method. These experiments have been applied for wide-range applications such as biomedicine, material science and organic chemistry. Automated tracking of cell populations in-vitro using time-lapse microscopy images helps high-throughput spatiotemporal measurements of a wide variety of cell behaviours, including migration (translocation), mitosis (division), quiescence (inactivity) and apoptosis (death) of cells in addition to the reconstruction of cell lineages (mother-daughter relations) [7]. These capabilities are of enormous values in many areas of biomedical engineering, such as stem cell research, oncological studies, tissue engineering, drug discovery, genomics, and proteomics [7,8].
Early live cell detection and tracking methods conventionally apply classic computer vision techniques such as segmentation algorithms, motion detection, level-set methods, image descriptors (such as SIFT/LBP/HOG). However, fully automatic cell tracking faces many challenges [6][7][8][9][10] including poor contrasts with high noise levels, irregular cell contours, difficulty for entry and exit of the cells, all of which make it difficult to deploy to commercial applications such as microfluidic POC devices. On the other hand, classic machine learning approaches such as subs
…(Full text truncated)…
This content is AI-processed based on ArXiv data.
