RBC Total Count Estimation via Image Segmentation

The patent segments the blood smear analysis into multiple independent Field of Views (FoVs) that are systematically scanned. Instead of relying on a single automated measurement point, the system divides the sample into multiple regions, each contributing to the final count. This segmentation allows the system to capture abnormal cells that might be missed in random sampling while maintaining automated operation.

The patent performs preliminary scanning and identification of cell types across multiple FoVs before finalizing the total count. The system pre-processes each FoV to identify RBCs, WBCs, and platelets, and pre-calculates their contributions to the total count. This preliminary action ensures that abnormal cells are detected and accounted for before the final measurement is made, improving identification accuracy while maintaining automation.

The patent replaces the chemical reagent-based measurement system with an optical/image-based system. Instead of using reagents to chemically interact with blood cells for measurement, the system uses digital imaging and image processing algorithms to identify and count cells. This substitution eliminates reagent consumption entirely while maintaining the ability to measure blood cell parameters, thereby reducing costs without sacrificing measurement capability.

The patent creates digital copies (images) of the blood smear sample and performs all measurements on these copies. Instead of consuming reagents to physically interact with the original sample, the system makes optical copies through imaging and analyzes these copies computationally. This approach allows repeated measurements without additional reagent consumption, as the same digital image can be processed multiple times.

The patent systematically segments the blood smear into multiple predetermined Field of Views that cover the entire smear area. Instead of randomly selecting FoVs, the system divides the sample into a grid or systematic pattern, ensuring comprehensive coverage. This segmentation approach maintains relatively simple analysis procedures while eliminating the bias inherent in random sampling, as every region of the smear has an equal and known probability of being included in the final count.

The patent changes the sampling parameter from random selection to systematic coverage. Instead of using random coordinates to select FoVs, the system uses a predetermined systematic pattern (such as a grid pattern or spiral pattern) to select FoVs that collectively represent the entire smear. This parameter change maintains the simplicity of automated image acquisition while significantly improving estimation accuracy by ensuring all regions are represented proportionally.

The patent changes the approach from requiring consistent physical stain quality to accepting variable stain characteristics as input parameters. Instead of using expensive devices to ensure uniform staining, the system captures images with standard devices and uses image processing algorithms that are robust to staining variations. The system adapts to different stain qualities by analyzing the actual pixel values and cell morphology in each image, rather than requiring predetermined uniform staining conditions.

The patent replaces expensive quality-control devices with computational image processing methods. Instead of using high-cost instruments to ensure and verify stain quality, the system uses software-based image analysis that can handle variations in staining. The image processing algorithms automatically adjust to different stain intensities and qualities, extracting meaningful cellular information without requiring expensive equipment to maintain perfect staining conditions.