NUST MISIS Leads Collaborative Effort to Automate Blood Sample Handling with Machine Vision
NUST MISIS researchers, working with colleagues from BSTU named after Shukhov and the NICEM center, are developing a robotic system that uses machine vision to speed up laboratory work and cut down on human error. The project is being reported by socialbites.ca from NUST MISIS.
In medical research, blood tests are among the most common procedures. During the initial material preparation, samples are divided after centrifugation by transferring the liquid from the primary tube into several smaller tubes. In this stage, the pipette used for analysis must be inserted to varying depths depending on the boundaries between the liquid layers. Errors frequently arise here due to human factors, which can compromise the quality of the results and slow down the workflow.
A large portion of the mistakes that degrade laboratory quality occur precisely at the preliminary material preparation stage. A subpar lab test may mean a second blood draw at best and a misdiagnosis at worst. Sergey Khalapyan, an employee in the Automation and Information Control Systems Department at STI NUST MISIS, told Ru that the robotic system for plasma aliquoting will remove the human factor from this critical step.
The system being developed automatically identifies the boundary level between serum fractions in a test tube using machine vision powered by a neural network. It can recognize segments in the tube with 98 percent accuracy, after which the algorithm computes the exact depth for the pipette to reach the serum with an error of less than 0.5 millimeters. This capability significantly elevates the reliability of laboratory research.
“The neural network provides precise image segmentation, and the algorithm built on it calculates the insertion depth needed to collect a representative portion. It also accounts for the nature of fractional limits, ensuring the maximum number of aliquots can be obtained while preserving the high quality of diagnostic work”, Khalapyan stated with confidence.
Looking ahead, the researchers plan to enhance the system by adding two more robotic units. The first will retrieve a tube containing biomaterial from a rack, move it to the work area, wait for the plasma sampling to complete, and then place the tube on another rack. The second machine will handle the collection of plasma and its distribution into the prepared tubes, further streamlining the process and reducing manual handling time.