Machine Learning Augmented Red Blood Cell Imaging and Analytics
Red Blood Cells (RBCs) tend to exhibit morphological abnormalities associated with pathological diseases and as a result to a chemically-altered environment. We propose a combinatorial imaging platform, guided by machine learning, for the identification and classification of RBCs with a direct impact in monitoring blood-related disorders.
Description
RBCs are the most abundant cell type in blood, with their primary function being the transport of oxygen throughout the body. In view of their circulation path through arteries, veins and small capillaries, RBCs are highly deformable and are generally characterized by a biconcave disciform shape. RBCs extracted from donors diagnosed with pathological diseases tend to exhibit morphological abnormalities, which can be measured and characterized. Red cell morphology constitutes a key tool in laboratory hematology, where dysmorphic RBCs can be detected and the presence of an underlying condition assessed.
Our unique approach relies upon the combination of high-throughput label-free live cell imaging with high resolution scanning electron microscopy and scanning probe microscopy methodologies. The data obtained from these state-of-the-art techniques is collected and analyzed using machine learning. ML-based data analytics provides an automated and more standardized approach for RBC classification in health and blood related diseases.
Goal
The learning objectives of the project include:
- Learn and understand several concepts related to the topic. These include blood and blood-related diseases, blood imaging, microscopy and machine learning
- Gain experience with different microscopy techniques, including holo-tomography, atomic force microscopy, scanning electron microscopy
- Be familiar with the principles of image analysis using different software, including FIJI, Imaris, Tomviz, Gwyddion
- Gain experience with the application of machine learning for cell classification
