Computational Hematology
Part of the Clarion Sentinel Detection Suite

Every drop of blood tells a story

Leukemia hiding in a routine CBC. Sickle cell crises predicted before they strike. Infections typed before cultures return. Sentinel Hema reads blood the way it was meant to be read — every cell, every parameter, every pattern.

500K+
Cells per blood smear — far more than any human microscopist can examine. Our AI analyzes every single one in under 90 seconds.
15-20%
Discordance rate between expert microscopists on identical smears
37+
CBC parameters analyzed simultaneously
95%+
AI cell classification AUROC
90sec
Full smear analysis time
The Hidden Diagnostic

The complete blood count is the most ordered laboratory test in the world — and the most underread. Clinicians glance at hemoglobin, white count, and platelets. If they're "normal," the report is filed. But buried within those 37+ parameters are patterns that reveal leukemia months before blasts appear, infections hours before cultures turn positive, iron deficiency before anemia develops, and bone marrow stress before the patient decompensates. The CBC is not a screening test. It is a diagnostic oracle — if you have the intelligence to read it.

Sentinel Hema transforms hematological analysis from a counting exercise into an AI-powered diagnostic intelligence platform. It analyzes every CBC parameter, every cell on the peripheral smear, every trend across serial draws — and correlates them with clinical context, medication profiles, and disease trajectories to surface the diagnoses that routine interpretation misses.

Analysis Engines

Eight engines. Every cell interrogated.

From automated differential to peripheral smear morphology to longitudinal trend analysis — Sentinel Hema reads blood the way the world's best hematologist would if they had unlimited time and perfect memory.

Engine 01
CBC Pattern Intelligence
AI analysis of 37+ CBC parameters simultaneously — finding disease signatures that single-parameter review misses.
Analyzes cross-parameter correlations invisible to standard interpretation

Clinicians read CBCs parameter by parameter — hemoglobin, WBC, platelets — comparing each to a reference range. But disease hides in the relationships between parameters, not in any single value. A WBC of 9.2 is "normal." A platelet count of 155 is "normal." A hemoglobin of 11.8 is "borderline." But the combination — with an elevated RDW, low reticulocyte production index, and falling MCV trend — is the signature of myelodysplastic syndrome months before blasts appear. Sentinel Hema evaluates all 37+ parameters as a constellation, not a checklist, and cross-references the pattern against a library of 200+ hematological disease signatures.

Performance
37+
CBC parameters analyzed simultaneously as a pattern
200+
Hematological disease signatures in the reference library
34%
More diagnoses surfaced vs. standard reference-range interpretation
Key Parameters
WBC + DiffHgb/HctMCV/MCH/MCHCRDWPlatelet + MPVRetic CountIG%NRBCIPF
Engine 02
Peripheral Smear AI Morphology
Deep learning analysis of every cell on the smear — classifying, counting, and flagging abnormalities in 90 seconds.
Trained on 500,000+ blood smear images — the largest dataset ever assembled

A standard blood smear contains thousands of individual cells — far more than any human microscopist can examine. Studies show 15-20% discordance rates between experienced microscopists examining identical smears. Sentinel Hema's AI morphology engine scans every cell on the smear at oil-immersion resolution, classifying each into 30+ cell types, flagging morphological abnormalities (blasts, atypical lymphocytes, dysplastic forms, schistocytes, sickle cells, parasites), and generating a quantified morphology report that triages routine cases and highlights anything unusual for human review.

Performance
95%+
AUROC for cell type classification across 30+ categories
90sec
Full smear analysis — every cell examined
500K+
Training images from the world's largest annotated smear dataset
30+
Distinct cell types classified including rare morphologies
Morphologies Detected
BlastsAtypical LymphsSchistocytesSpherocytesSickle CellsTarget CellsHowell-JollyToxic GranulationDöhle BodiesHypersegmented NeutParasitesNRBC
Engine 03
Leukemia & Lymphoma Early Detection
Pattern recognition across CBC trends and smear morphology to detect hematologic malignancy at its earliest stage.
Detects circulating blast cells and MDS signatures months before clinical diagnosis

Blood cancers often announce themselves in routine blood work months before clinical diagnosis — through subtle cytopenias, gradual lymphocytosis, circulating immature cells, or morphological dysplasia visible only on careful smear review. Yet these early signals are frequently missed because automated analyzers lack the morphological discrimination and clinical context to flag them. Sentinel Hema's malignancy detection engine monitors CBC trends longitudinally, analyzes smear morphology for blast cells and dysplastic forms, correlates findings with patient demographics and clinical context, and generates a hematologic malignancy risk score that triggers appropriate workup — flow cytometry, bone marrow biopsy, or hematology referral — before the disease declares itself clinically.

Performance
2-6mo
Earlier detection of hematologic malignancy vs. standard care
92%
Sensitivity for circulating blast detection on peripheral smear
88%
Accuracy classifying malignancy subtype pre-flow cytometry
Engine 04
Anemia Classification & Etiology
Automated anemia workup using RBC indices, reticulocyte parameters, and iron studies correlation.
Classifies anemia etiology with 91% accuracy from CBC alone

Anemia affects 1.6 billion people worldwide, but "anemia" is not a diagnosis — it is a symptom with dozens of etiologies, each requiring different treatment. Iron deficiency. B12 deficiency. Folate deficiency. Chronic disease. Hemolysis. Thalassemia. Sideroblastic anemia. Aplastic anemia. Each leaves a distinctive fingerprint in the CBC — MCV, MCH, MCHC, RDW, reticulocyte count, reticulocyte hemoglobin content (CHr), and immature reticulocyte fraction (IRF) — that Sentinel Hema reads as a pattern to classify etiology with 91% accuracy from the CBC alone, often eliminating the need for expensive and time-consuming iron studies, B12/folate levels, and hemoglobin electrophoresis as first-line tests.

Performance
91%
Anemia etiology classification accuracy from CBC parameters alone
4mo
Earlier detection of iron deficiency before clinical anemia develops
42%
Reduction in unnecessary follow-up laboratory orders
Engine 05
Coagulation & Platelet Intelligence
Integrated analysis of platelet count, MPV, IPF, PT/INR, aPTT, fibrinogen, and D-dimer for coagulopathy detection.
Detects DIC, TTP/HUS, and ITP patterns before clinical crisis

The coagulation system is invisible on a standard CBC — until it fails catastrophically. But the early signals are present: falling platelet trends, rising mean platelet volume (immature platelets entering circulation), elevated immature platelet fraction, widening PT/INR, and rising D-dimer. Sentinel Hema correlates these with peripheral smear findings (schistocytes for microangiopathic hemolysis, platelet clumping artifacts) and clinical context to detect DIC, thrombotic thrombocytopenic purpura (TTP), hemolytic uremic syndrome (HUS), heparin-induced thrombocytopenia (HIT), and immune thrombocytopenia (ITP) — conditions where delayed diagnosis is measured in lives, not days.

Performance
8hr
Earlier detection of DIC progression (integrated with Sentinel Sepsis)
94%
Schistocyte detection accuracy for TTP/HUS screening
96%
HIT probability scoring accuracy (4T score + AI enhancement)
Engine 06
Infection Typing & Severity
Differentiates bacterial, viral, fungal, and parasitic infections from CBC morphology patterns — hours before cultures.
Bacterial vs. viral differentiation 24-48 hours before culture results

Infections leave distinctive fingerprints in the blood that are visible long before culture results return. Bacterial infections produce left shift (elevated bands, metamyelocytes, immature granulocytes), toxic granulation, and Döhle bodies. Viral infections produce atypical lymphocytes and relative lymphocytosis. Parasitic infections elevate eosinophils. Fungal infections in immunocompromised patients produce subtle monocyte and neutrophil changes. Sentinel Hema reads these morphological and quantitative patterns to type the infection and assess severity — enabling targeted empiric therapy 24-48 hours before culture results confirm the pathogen.

Performance
88%
Accuracy differentiating bacterial from viral infection
24-48hr
Earlier infection typing vs. waiting for culture results
92%
Sensitivity for left shift and immature granulocyte detection
Engine 07
Bone Marrow Stress Indicators
Detects signs of marrow failure, infiltration, and stress from peripheral blood patterns — guiding biopsy decisions.
Triggers appropriate BM biopsy 3 weeks earlier than standard evaluation

The bone marrow communicates its status through the blood. Pancytopenia patterns, leukoerythroblastic pictures (nucleated red blood cells and immature white cells in circulation), teardrop cells suggesting myelofibrosis, circulating plasma cells suggesting myeloma, and reticulocyte patterns revealing production failure — all visible on CBC and peripheral smear, all interpretable by AI, and all requiring different and often urgent clinical responses. Sentinel Hema monitors these marrow stress indicators to identify patients who need bone marrow biopsy, hematology referral, or urgent evaluation for marrow infiltration by metastatic cancer.

Performance
3wk
Earlier BM biopsy trigger vs. standard clinical evaluation
90%
Accuracy predicting abnormal marrow from peripheral findings
Engine 08
Longitudinal Trend Intelligence
Serial CBC analysis across months and years — detecting the slow drifts that signal developing disease.
The most powerful engine — because disease reveals itself over time

A single CBC is a snapshot. A series of CBCs over months and years is a movie — and in that movie, the trajectory of disease becomes visible long before any single value crosses a threshold. A hemoglobin that has drifted from 14.2 to 12.8 over two years is "normal" on every individual draw but represents a trajectory that demands investigation. A platelet count falling from 240 to 160 over six months — always "normal" — is the signature of an occult myeloproliferative disorder or early ITP. Sentinel Hema analyzes every historical CBC, calculates trajectories for all 37+ parameters, identifies statistically significant trends, and alerts clinicians to the slow-motion deterioration that point-in-time interpretation will always miss.

Performance
Analyzes every historical CBC in the patient's record
6mo
Average earlier detection of chronic hematological disorders
28%
More early-stage diagnoses vs. standard CBC interpretation
Proven Impact

Diagnoses found. Diseases caught. Lives changed.

Results from our deployed health systems.

Academic Medical Center — Hematology Division

AI-assisted peripheral smear analysis detecting 23 early leukemias in year one

The Outcome

Sentinel Hema's peripheral smear AI and longitudinal trend engine were deployed across all CBC orders in a 900-bed academic medical center. In the first year, the system flagged 23 cases of suspected hematologic malignancy from routine blood work — 19 of which were confirmed on subsequent flow cytometry and bone marrow biopsy. Of these 19 confirmed cases, 14 were identified at a stage earlier than would have been detected through standard clinical pathways. The hematology department estimated that earlier detection advanced treatment initiation by an average of 3.4 months across these patients.

23
Suspected malignancies flagged
19
Confirmed on workup
14
Detected at earlier stage
3.4mo
Earlier treatment initiation
Community Hospital Network — 6 Facilities

Infection typing from CBC morphology reducing unnecessary antibiotic use

The Outcome

Deployed across 6 community hospitals, Sentinel Hema's infection typing engine analyzed CBC morphology patterns for every patient with a suspected infection. By differentiating bacterial from viral patterns 24-48 hours before culture results, the system enabled clinicians to narrow antibiotic therapy earlier in 1,840 cases. Unnecessary broad-spectrum antibiotic days dropped 38%. For emergency departments, the system's rapid infection severity assessment reduced time to appropriate therapy by 3.1 hours in bacterial cases — a metric with direct mortality implications.

1,840
Cases with earlier typing
38%
Fewer broad-spectrum days
3.1hr
Faster targeted therapy
88%
Typing accuracy
Primary Care Network — 48 Clinics

Longitudinal trend analysis uncovering hidden anemia and early MDS in outpatients

The Outcome

Deployed across a 48-clinic primary care network serving 380,000 patients, the longitudinal trend engine analyzed historical CBC data to identify patients with slowly developing hematological abnormalities. In the first year, it flagged 2,400 patients with trending parameters warranting evaluation — of whom 860 had actionable findings (iron deficiency pre-anemia, B12 deficiency, early MDS, monoclonal B-cell lymphocytosis). Eight patients were found to have early-stage MDS that would not have been detected for months to years under standard care. Primary care physicians reported that the system transformed their approach to "normal" CBC results.

380K
Patients monitored
2,400
Trending abnormalities flagged
860
Actionable findings confirmed
8
Early MDS cases discovered
Clinician Voices

I've been a hematopathologist for twenty-six years. I've looked at more blood smears than I can count. And I will tell you plainly: this AI sees things I miss. Not because I'm not good at what I do — but because a human cannot examine 500,000 cells on a smear and maintain the same attention on cell 499,000 as on cell one. The machine can. And it does.

Director of Hematopathology
Fellowship-Trained, 26 Years Practice
Academic Medical Center

The longitudinal trend engine changed how I practice primary care medicine. I used to file a CBC with "normal" hemoglobin and move on. Now I see that my patient's hemoglobin has dropped from 14.1 to 12.4 over three years — still "normal" every time, but the trajectory screams iron depletion. I caught a colon cancer from a CBC trend. A colon cancer. From a routine blood test. That is what this system does.

Primary Care Physician
Internal Medicine, 14 Years Practice
Community Health Network

For my sickle cell patients, the crisis prediction capability is life-changing — literally. We can see the reticulocyte count climbing, the LDH rising, the haptoglobin falling — and we intervene with hydration and pain management before the crisis hits. My patients are spending fewer nights in the emergency department. That's not a metric. That's a life.

Hematologist-Oncologist
Sickle Cell Disease Program Director
Children's Hospital
500K+
Cells analyzed per smear
95%+
Cell classification AUROC
340+
Hospitals deployed
34%
More diagnoses surfaced
Read the Blood

Every cell has a diagnosis to tell

Schedule a clinical demonstration of Sentinel Hema — configured for your laboratory volume, your analyzer integration, and your patient population.

Or contact our clinical team at hema@brindwell.com