Skip to main content

Hematology

Digital Imaging

The integration of Digital Imaging (DI) into the hematology laboratory represents the most significant shift in morphological assessment since the invention of the automated cell counter. These systems, often exemplified by platforms like CellaVision, automate the traditional manual differential. Instead of a laboratory scientist physically looking through microscope oculars, an automated system scans the slide, captures high-resolution images of cells, and uses Artificial Neural Networks (ANN) to pre-classify them. This technology standardizes the differential, reduces ergonomic strain, and facilitates remote consultation (telepathology)

Principles of Operation

Digital imaging systems function as a bridge between the slide maker/stainer and the Laboratory Information System (LIS). The process involves robotic manipulation of the slide combined with advanced image processing algorithms

The Scanning Process

  • Monolayer Detection: Once a stained slide is loaded (either manually or via a track system), the instrument applies immersion oil automatically. It first performs a low-power scan (typically equivalent to 10x) to identify the “Feather Edge” or monolayer - the optimal zone where RBCs are evenly spaced and not overlapping
  • High-Resolution Capture: The system locates white blood cells (WBCs) using coordinates derived from the low-power scan. It then switches to high-power magnification (equivalent to 100x) to capture digital images of individual WBCs. Most systems are programmed to capture a set number of cells (e.g., 100 or 200 WBCs) based on the order type
  • RBC and Platelet Views: In addition to single WBC images, the system captures broader Fields of View (FOV) to allow the laboratory scientist to assess Red Blood Cell morphology (anisocytosis, poikilocytosis) and estimate platelet numbers

Artificial Neural Networks (ANN)

The “brain” of the digital imaging system is the Artificial Neural Network. This is a form of Machine Learning trained on a massive database of hundreds of thousands of cell images verified by hematology experts

  • Segmentation: The software analyzes the image by separating the cell from the background. It identifies the nucleus (by color and density) and the cytoplasm
  • Feature Extraction: The system measures specific features of the cell, including:
    • Nuclear-to-Cytoplasmic (N:C) ratio
    • Nuclear shape (round, lobed, indented)
    • Chromatin texture (clumped, smooth, lacy)
    • Cytoplasmic color (basophilic, eosinophilic)
    • Granularity (presence and type of granules)
  • Classification: Based on these features, the algorithm assigns the cell to a specific category (e.g., “Segmented Neutrophil,” “Lymphocyte,” “Blast,” “Variant Lymph”). It assigns a probability score to this classification. If the probability is low, the cell may be placed in an “Unidentified” or “Artifact” category for human review

The Operational Workflow

The role of the laboratory scientist shifts from finding the cells to verifying the computer’s decisions. This is known as verification rather than counting

  • The Cell Gallery: The user interface presents the cells sorted by class. For example, all cells the computer thinks are Neutrophils are displayed in one row, and all Lymphocytes in another. This allows for rapid review
    • Drag and Drop: If a cell is misclassified (e.g., a Band is placed in the Segmented Neutrophil row), the laboratory scientist simply clicks the image and drags it to the correct category. The count updates automatically
  • RBC Morphology Grading: The system presents several high-quality images of the RBC field. The software often pre-grades morphology (e.g., “Polychromasia: 2+,” “Poikilocytosis: 1+”). The laboratory scientist reviews these images and accepts or modifies the grading
  • Platelet Estimation: The system offers a magnified view of the slide background to allow the laboratory scientist to verify that the platelet count matches the automated analyzer result and to check for clumps

Operational Advantages

Digital imaging offers distinct advantages over manual microscopy in terms of workflow efficiency and quality assurance

  • Standardization: In a manual differential, the quality of results depends heavily on the individual tech’s skill and fatigue level. Digital systems ensure every slide is reviewed under identical lighting and magnification conditions. The “Gallery View” allows techs to compare cells side-by-side, improving consistency in distinguishing bands from segs
  • Ergonomics: Traditional microscopy is a leading cause of repetitive strain injury (neck and back strain) in laboratory professionals. Digital imaging allows the laboratory scientist to sit upright and view images on a computer monitor, significantly reducing physical stress
  • Remote Consultation and Education
    • Telepathology: Difficult cases (e.g., suspected leukemia) can be reviewed instantly by a pathologist located in a different office or even a different hospital within the network, without physically transporting the glass slide
    • Archiving: Interesting cases can be tagged and saved indefinitely. These libraries are invaluable for training students and for competency assessment of staff

Limitations & Troubleshooting

Despite the sophistication of AI, the digital imaging system is not a replacement for clinical judgment. The laboratory scientist remains the final authority

  • Recognition Limitations
    • Smudge Cells: The segmentation software often struggles with Smudge Cells (Albumin helps) because they lack a defined border. These are frequently dumped into the “Artifact” or “Unclassified” bin
    • Pathological Cells: While excellent at identifying normal cells, the system often struggles to differentiate specific pathological types (e.g., distinguishing a Myeloblast from a Monoblast, or a Sezary cell from a normal Lymphocyte). These are usually flagged as “Blasts” or “Variant Lymphs” for human classification
    • RBC Inclusions: Small inclusions like Howell-Jolly bodies or Basophilic Stippling may be missed by the software if the focus is not perfect
  • Sample Quality Issues
    • Stain Quality: The ANN is color-dependent. If the stainer is drifting (too blue or too red), the computer may misclassify cells (e.g., calling Eosinophils “Neutrophils” if the granules don’t stain bright orange). This makes the DI system an excellent secondary QC monitor for the stainer
    • Slide Quality: A slide that is too thick (no monolayer) or too thin (ragged edge) will cause the system to reject the slide or produce “Out of Focus” errors. The system requires a high-quality wedge smear
  • The “Megascan”: For body fluids or low-cellularity samples, the system can stitch together hundreds of images to create a digital map of the entire slide. This allows the laboratory scientist to “zoom in” on any area, mimicking the action of moving the stage on a manual microscope