研究目的
To develop automated computer vision algorithms for quantifying immunomagnetic beads and leukemia cells from bright-field optical microscope images, addressing the gap in automated quantification in the presence of immunomagnetic beads.
研究成果
The proposed algorithms successfully quantify immunomagnetic beads and leukemia cells with high precision and low error rates, demonstrating potential as a signal readout mechanism for biochips in cancer monitoring. Future work should focus on improving image quality and exploring feature reduction techniques.
研究不足
Image quality issues such as blurry edges and uneven illumination affect algorithm performance, particularly in 20× images. The method requires standardization of imaging processes to improve accuracy. Feature extraction and machine learning parameters may need optimization for better results.
1：Experimental Design and Method Selection:
The study uses image processing and computer vision algorithms, including color-based thresholding, Hough transform for circular object detection, and machine learning methods like SVM, Random Forest, and Neural Networks for cell detection.
2：Sample Selection and Data Sources:
B lymphoblast cells (CCRF-SB) and immunomagnetic beads conjugated with anti-human CD19 antibodies are used. Images are acquired using a bright-field microscope with 20× and 40× objectives.
3：List of Experimental Equipment and Materials:
Immunomagnetic beads (
4：5 μm diameter, ThermoFisher), B lymphoblast cells (ATCC), microscope with DS-Ri1 CCD camera (Nikon), MATLAB software for algorithm development. Experimental Procedures and Operational Workflow:
Cells are cultured and bound to immunomagnetic beads using a magnetic separator. Images are captured, converted to YUV color space, thresholded, and processed with morphological operations and machine learning for detection and counting.
5：Data Analysis Methods:
Performance is evaluated using precision, recall, specificity, F-measure, false negative rate, and detection error rate, with manual counting by an expert as ground truth.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
