In vitro diagnostics is at an inflection point. Traditional hematology analyzers, built for centralized laboratories, struggle to keep pace with a world that increasingly demands rapid, high-resolution insights at the point of care. On one side are numerical analyzers that deliver counts and indices but lack morphological context; on the other are manual smears and microscopy that provide rich detail but remain slow, operator-dependent, and hard to scale.

Ozelle’s AI × CBM (Complete Blood Morphology) architecture is a response to this structural tension. By fusing cell morphology imaging, multi-class AI recognition, and integrated multi-assay testing into a scenario-ready system, Ozelle aims to deliver lab-grade depth from a single drop of blood—whether the device is in a tertiary hospital, a primary clinic, a pharmacy, or a mobile setting.

This article unpacks the technical foundations and clinical implications of AI × CBM, and explores how it may reshape the future of blood diagnostics.

Modern Healthcare Faces Three Converging Pressures in Diagnostics:

1. Rising test volume and complexity Aging populations, chronic disease, and multi-morbidity drive more testing per patient, while the range of clinically relevant biomarkers continues to expand. Central labs face mounting workloads without commensurate increases in skilled staff.
2. Decentralization of care. Care is moving closer to patients — into primary care clinics, pharmacies, community centers, and even mobile and home-based environments. These sites demand compact, low-maintenance devices that can deliver actionable insights quickly, often with non-laboratory personnel.
3. Limitations of conventional hematology platforms Conventional 3-part or 5-part differential analyzers excel at numerical parameters but are blind to subtle cell morphology and abnormal forms. When deeper insight is needed, labs must fall back to manual smear review, which is:
   1. Time-consuming and heavily operator-dependent
   2. Difficult to standardize and scale
   3. Poorly suited to decentralized or high-throughput environments

At the same time, clinicians increasingly expect more than raw numbers. They need synthesized insights: which patterns are abnormal, what they might mean, and which additional tests are most relevant — ideally within a single, streamlined workflow.

This is the context in which AI × CBM emerges: not just as another analyzer, but as a new diagnostic architecture that integrates morphology, multi-modal assays, and AI-native software.

Technical Core: AI × CBM Algorithms and Data

From CBC Counts To Complete Blood Morphology

Traditional hematology analyzers measure electrical impedance or optical scatter to derive cell counts and indices. Ozelle’s CBM approach keeps the quantitative backbone but adds a full imaging and AI layer on top.

At the hardware level, CBM combines:

• High-resolution microscopy and optics A customized optical system captures microscopic images at oil-immersion–level resolution in real time, enabling visualization of cell size, shape, nuclear segmentation, and cytoplasmic features.
• Liquid-phase staining and Z-stack imaging Wet staining (e.g., Wright–Giemsa–like protocols) enriches color contrast, while Z-stack imaging reconstructs 3D-like cell representations. This produces multi-dimensional views that are highly informative for AI models.
• High-speed full-field scanning Automated scanning covers the entire field, capturing thousands of cells per slide-equivalent without manual navigation.

These images are then fed into the AI engine that powers CBM.

Multi-Class Cell Recognition: Beyond “5-Part Diff”

Where a traditional differential might separate neutrophils, lymphocytes, monocytes, eosinophils, and basophils, CBM extends classification to a broader set of clinically important morphologies, including:

• Neutrophil subtypes
  • NST: Neutrophilic stab granulocytes (band forms / earlier precursors), reflecting left shift and bone marrow stress
  • NSG: Neutrophilic segmented granulocytes (mature neutrophils, first line of defense)
  • NSH: Neutrophilic hypersegmented granulocytes, often linked to dysregulated maturation or megaloblastic processes
• Abnormal lymphoid and atypical cells
  • ALY: Atypical lymphocytes, which may suggest viral infections or reactive lymphocytosis
• Erythroid-related parameters
  • RET: Reticulocytes, indicating marrow response to anemia or hemolysis
• Platelet and other formed elements
  • PAg and detailed platelet parameters, with the ability to visualize platelets and aggregate patterns
  • Morphologically abnormal RBCs: schistocytes, echinocytes, teardrop cells, etc.

This multi-class recognition is not an add-on rule set; it is driven by deep learning models trained end-to-end on real-world image data.

Algorithmic Engine: Deep Learning at Scale

Ozelle’s AI stack is built around convolutional neural networks (CNNs) trained on one of the largest known datasets of real blood cell images in routine practice.

Key characteristics include:

• Massive training corpus from real-world devices With more than 50,000 analyzers installed worldwide and tens of millions of cell images generated every day, Ozelle’s cumulative database exceeds 100 billion data points. This allows the system to capture:
  • Inter-patient variability
  • Ethnic and regional diversity
  • Instrument and reagent drift over time
  • Rare morphologies that are difficult to collect in controlled studies
• Multi-dimensional inputs and advanced enhancement The imaging pipeline provides multi-spectral, multi-angle views, with CNN-based image enhancement and super-resolution techniques to push beyond the limits of raw optics. This improves edge definition, nuclear segmentation, and granule visibility — all crucial for robust morphology classification.
• Auto-ML and continual learning The model set is not static. Algorithm performance is iteratively refined using automated machine-learning pipelines and feedback from an industry-level quality control system. This enables:
  • Continuous calibration across global deployments
  • Reduction of false positives/negatives in rare cell classes
  • Progressive alignment with expert pathologist-level performance

The result is an AI recognition engine that can classify a broad set of white blood cell subtypes, abnormal cells, and formed elements with high precision, while maintaining the speed and reproducibility expected from an automated analyzer.

Integrated Architecture: Hematology, Biochemistry, Immunoassay In One Workflow

Scenario-Driven Panels Built On a Single Platform

The AI × CBM concept extends beyond morphology. Ozelle’s platform is designed as an integrated mini-lab that unifies:

• Hematology / CBM: Complete blood count plus morphology and extended parameters
• Biochemistry (dry chemistry): e.g., GLU, TG, TC, UA, renal and liver function markers
• Immunoassay (fluorescence immunochromatography): infection and inflammation markers, hormones, cardiac markers, etc.

All tests are run through a maintenance-free, cartridge-based system, which:

• Uses single-use integrated test kits (hematology, biochemistry, immunoassay cards)
• Eliminates traditional liquid systems, tubing, and frequent maintenance
• Minimizes risk of cross-contamination and simplifies operation for non-lab users

From this unified platform, clinicians can configure scenario-specific panels, such as:

• Infection typing: CBC + CRP + SAA
• Diabetes management: CBC + HbA1c
• Cardiac screening: CBC + NT-proBNP, with optional troponin and CK-MB
• Thyroid, kidney, or bone metabolism panels, and more as the menu expands

The system is thus not just a hematology analyzer with AI, but a multi-modal diagnostic engine that can tailor its test menu to the clinical question at hand.

Single-Drop, High-Efficiency Workflow

In practical terms, AI × CBM translates to:

• Sample requirement as low as ~30 µL of capillary blood for hematology in many configurations — suitable for fingerstick collection in outpatient and pharmacy settings
• Throughput around 10 samples per hour, balancing point-of-care needs with small-lab workflows
• One-click operation with automated sample pretreatment, staining, imaging, analysis, and reporting

This design addresses the common pain points of decentralized sites: limited staff, limited time, and limited tolerance for complex maintenance.

Intelligent AI Workbench (Open Dx): From Numbers To Guidance

Integrated Digital Workbench Inside The Analyzer

A critical layer of AI × CBM is the Intelligent AI Workbench (Open Dx), which moves diagnostic intelligence into the core of the analyzer user interface.

Open Dx integrates:

• Test ordering and panel selection
• Real-time result visualization, with access to:
  • Raw numerical parameters
  • Cell histograms and scatter plots
  • High-resolution cell images and morphology tiles
• AI-assisted guidance and interpretation

This transforms the analyzer from a passive data source into an interactive diagnostic console.

From Static Reports To Interactive Diagnostic Insights

Traditional analyzers output static printed reports. In contrast, Open Dx provides:

• Automated abnormality detection The system highlights deviations from reference ranges, abnormal distributions, and suspicious morphological patterns.
• Risk flagging and pattern recognition Using pre-trained diagnostic patterns, the workbench can suggest possible clinical scenarios — for example:
  • Early bacterial infection with left shift and inflammation markers
  • Viral infection with lymphocyte pattern changes and SAA dynamics
  • Potential hematological abnormalities warranting follow-up
• Structured overview interpretations Instead of leaving clinicians to manually synthesize dozens of parameters, the AI workbench surfaces a concise interpretation section, offering:
  • Summary of key abnormalities
  • Possible pathophysiological mechanisms
  • Suggestions for further tests or clinical correlation
• Conversational AI for report consultation A dialogue-like interface allows clinicians to query the system directly: “Why is NST elevated?” “What could this combination of low LYM and high MON mean?” The system responds by contextualizing results with relevant literature-based and rule-based knowledge, helping clinicians interpret complex patterns more quickly and confidently.

For veterinary applications, the same framework extends to species-specific guidance and medication references, highlighting the extensibility of the AI workbench concept.

Clinical and Operational Impact: Redefining Efficiency At The Front Line

For Clinicians: Deeper Insight, Earlier Detection

By combining AI × CBM with scenario-based panels, clinicians gain:

• Earlier detection of subtle abnormalities Multi-class neutrophil analysis (NST/NSG/NSH), abnormal lymphocytes, reticulocytes, and RBC shape abnormalities can reveal early infection, marrow stress, or hematologic disorders that might be overlooked with simple counts.
• Richer context from a single encounter Instead of sending multiple samples to different analyzers or external labs, clinicians can access CBC, morphology, inflammatory markers, cardiac markers, and metabolic indicators from a single device and visit.
• Decision support in time-constrained environments AI-generated interpretations and risk flags help non-specialists — such as primary care physicians or pharmacists — interpret complex reports rapidly and determine whether to:
  • Keep management at primary level
  • Escalate to specialist referral
  • Order more specific follow-up tests

For Primary Care, Pharmacies, and Decentralized Sites

In decentralized environments, AI × CBM targets the core operational barriers that previously limited advanced diagnostics:

• Minimal maintenance and training Cartridge-based design, absence of fluidic pipelines, and guided workflows reduce the dependency on highly trained laboratory technicians.
• Compact footprint and multi-functionality A single device replaces separate hematology, biochemistry, and immunoassay analyzers, saving space and simplifying procurement and service.
• Improved economics Consolidating tests into all-in-one cartridges and minimizing instrument maintenance lowers total cost of ownership, making advanced diagnostics viable for smaller sites.

The cumulative effect is a redistribution of diagnostic capability: more of what was once confined to hospital labs becomes accessible closer to patients, without compromising analytical depth.

Future Outlook: Where AI × CBM Can Go Next

AI × CBM is not a static product; it is a platform architecture with significant headroom for innovation.

Several directions stand out:

1. Expanded biomarker menus via software updates With modular assay cards and OTA (over-the-air) upgrade capabilities, new parameters — from novel cardiac markers to emerging inflammatory or oncology signatures — can be added without redesigning the core hardware.
2. More granular and rare-cell detection As the global data corpus grows, AI models can be trained to recognize increasingly rare morphologies (e.g., specific blast types, dysplastic forms) and integrate them into screening algorithms for early hematologic disease detection.
3. Personalized and longitudinal analytics Leveraging the broader Ozelle IoT platform, analyzers can be connected to cloud-based systems that track patient trends over time, enabling:
   1. Personalized baselines and dynamic reference ranges
   2. Alerting on subtle deviations before overt disease manifests
4. Integrated care pathways and telemedicine AI-enriched reports can be securely shared with remote specialists, forming the backbone of tele-hematology and collaborative care models between primary providers and tertiary centers.
5. Cross-species and cross-domain expansion The veterinary implementations of AI × CBM illustrate how the same morphological and AI framework can be adapted to different biological contexts. Future extensions may include niche clinical domains or specific research applications.

Conclusão

“AI × CBM: The Next Generation of Complete Blood Morphology” is more than a slogan. It encapsulates a structural rethinking of how hematology and related diagnostics should function in a healthcare system that is becoming more distributed, data-rich, and outcome-driven.

By combining:

• High-resolution, AI-enabled morphology
• Multi-class cell recognition beyond traditional differentials
• Integrated hematology, biochemistry, and immunoassay testing in one maintenance-free device
• And an Intelligent AI Workbench that transforms raw data into interactive diagnostic guidance

Ozelle’s AI × CBM platform offers a template for scenario-ready diagnostics: deep enough for specialists, yet simple and robust enough for frontline use.

For laboratory leaders, medical technologists, and health-tech decision-makers, the key takeaway is this: AI in diagnostics delivers the greatest value not when it is bolted onto existing analyzers, but when it is designed into the entire diagnostic stack — from optics and assays to algorithms, workflows, and clinical decision support. AI × CBM represents one of the most mature examples of that end-to-end integration now entering routine practice.

FAQs

Q1: Is the system maintenance-free? A1: Yes. Single-use cartridges and no internal liquid lines eliminate routine cleaning.

Q2: Is additional reagent purchase required? A2: No. All necessary reagents are integrated into the test kits.

Q3: How difficult is operation? A3: Operation is fully automated after loading the cartridge, suitable for non-lab staff.

Q4: Do reagents need cold-chain transport? A4: No. They remain stable at typical room temperatures.

Q5: Can AI × CBM connect to LIS/HIS?

A5: Yes. The platform supports standard LIS/HIS and network connectivity for data integration.