Advanced RBC Application

The CellaVision DM96 and DM1200 with the Advanced RBC Application are automated in-vitro diagnostic systems designed to improve the efficiency and detail of red blood cell morphology analysis from peripheral blood smears. Utilizing image analysis with artificial neural networks, these devices pre-characterize RBCs into 21 morphological categories, including various forms of poikilocytosis and inclusions, presenting results with both overview and individual cell images for operator verification and classification.

The CellaVision DM96 and DM1200 systems, equipped with the Advanced RBC Application, are automated in-vitro diagnostic tools designed for detailed analysis of red blood cell (RBC) morphology in peripheral blood smears. Their primary purpose is to automatically locate and present images of blood cells from samples flagged as abnormal, aiding skilled operators in cell identification and classification. The Advanced RBC Application employs a sophisticated methodology that pre-characterizes RBCs based on abnormalities in color, size, shape (Poikilocytosis), specific types of Poikilocytosis, and the presence of cellular inclusions. This analysis is powered by segmentation, feature calculation, and deterministic artificial neural networks (ANNs) trained to distinguish among 21 distinct RBC morphological characteristics, providing users with an overview of the RBC monolayer, percentage of abnormal cells with a corresponding grade, and critically, individual cell images grouped by morphology for detailed review and potential re-classification. This Advanced RBC Application represents a significant upgrade from previous CellaVision DM system functionalities, which identified only 6 RBC morphological groups using a decision-tree algorithm and offered only an overview image. The enhanced system's capacity to differentiate 21 characteristics, including specific poikilocytes (e.g., schistocytes, sickle cells) and inclusions (e.g., Howell-Jolly bodies), alongside the ability to view individual cell images, facilitates a more comprehensive and interactive diagnostic process. Extensive performance evaluations, including reproducibility and repeatability studies compliant with CLSI EP05-A3, confirmed the system's reliability, meeting predefined acceptance criteria for grading and proportional cell counts. Furthermore, clinical studies adhering to CLSI H20-A2, comparing the application against manual microscopy and automated cell counters for size parameters, demonstrated acceptable overall agreement, sensitivity, and specificity across various RBC morphology groups and individual characteristics, affirming its effectiveness for RBC characterization.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the performance of the deterministic artificial neural networks (ANNs) in the Advanced RBC Application compare to convolutional neural networks (CNNs) or other deep learning models for RBC classification, particularly for subtle or rare morphologies?; What is the quantitative impact on laboratory workflow efficiency and turnaround time when implementing the CellaVision DM systems with the Advanced RBC Application compared to traditional manual microscopy or systems with less detailed RBC characterization?; Are there specific RBC morphological characteristics among the 21 analyzed by the Advanced RBC Application for which the system shows significantly higher or lower concordance with expert manual review, and what are the underlying reasons for any discrepancies?; How does the system handle samples with extremely high or low cell counts, or those with significant artifacts (e.g., stain precipitate, platelet clumps) that might interfere with RBC image segmentation and feature extraction?; What were the key parameters of the training datasets and validation methods used for the ANNs, and how does the system ensure robustness across diverse patient populations and potential staining variations?