Blood Sample Quality Management
Cognex Deep Learning analyzes centrifuged blood for proper separation
Blood testing analyzers rely on accurately prepared samples and test setups. Centrifuged blood samples are ranked according to different indices (e.g. hemoglobin, bilirubin, and intralipid index levels) and receive quality scores based on turbidity and plasma color. All of these indices can vary in appearance based on how samples are loaded and oriented in the rack. Blood separation and the presence/absence of labels and caps are important factors in quality assessment, which is crucial for robust workflows in highly automated labs. Because there are so many judgment-based factors, this inspection often falls to humans.
Deep learning-based image analysis can ascertain whether centrifuged blood has been effectively separated into distinct phases (plasma, buffy coat, and red blood cells) and classify samples by the criteria used in processing. The Cognex Deep Learning classification tool trains on annotated images of different classes until it successfully conceptualizes and generalizes the normal appearance of different phases. During runtime, Cognex Deep Learning sorts multiple classes within a single vial, extracting blood quality factors like plasma color and turbidity, buffy coat volume, and centrifugation status into distinct classes while ignoring irrelevant qualities like cap status and label presence. Based on the classes, it separates passing samples from failing samples. This information can also provide useful process control information about the samples being drawn and re-centrifuged.