About PathoIntern

Problem Statement

Why blood smear triage is broken, what is at stake, and how PathoIntern responds.

The Problem

Manual blood smear review is one of the most cognitively demanding tasks in clinical pathology. A trained pathologist must visually identify subtle morphological abnormalities across thousands of cells per slide — a process that cannot be meaningfully abbreviated without risking missed findings.

5–10%
Annual workload growth
Slide volumes are rising each year without proportional growth in pathologist capacity.
No tools
Existing AI triage for blood smears
Unlike radiology, where AI worklist tools exist, blood smear triage has no comparable solution.
High stakes
Delay risk for critical findings
When urgent slides are mixed with routine cases, critical findings can be delayed by hours.

What Pathologists Must Identify

Blood smear review requires expert recognition of morphological findings across multiple clinical contexts:

Red Cell Morphology

  • Schistocytes (microangiopathic hemolytic anemia)
  • Echinocytes / burr cells
  • Target cells, spherocytes, elliptocytes
  • Sickle cells, bite cells

Hemoparasites

  • Plasmodium spp. (malaria)
  • Babesia spp.
  • Trypanosoma spp.
  • Ehrlichia/Anaplasma morulae

White Cell Abnormalities

  • Blast cells (acute leukemia)
  • Hypersegmented neutrophils
  • Reactive lymphocytes
  • Leukocytosis / leukopenia patterns

Platelet Abnormalities

  • Giant platelets
  • Platelet clumping
  • Thrombocytopenia patterns
  • Platelet satellitism

The Impact

Critical

Delayed Diagnosis for Critical Conditions

Conditions such as TTP (thrombotic thrombocytopenic purpura), acute leukemia, and severe malaria require treatment within hours. When critical slides wait in a routine queue, time-to-treatment increases.

High

Pathologist Burnout and Error Risk

High-volume review of predominantly normal slides creates cognitive fatigue. Research shows burnout directly correlates with increased diagnostic error rates (Khatab et al., 2024).

Systemic

No Systematic Triage Infrastructure

Unlike radiology — where AI worklist tools have been shown to cut critical-case turnaround by 56% (Baltruschat et al., 2021) — hematopathology has no equivalent assistive system.

PathoIntern's Response

PathoIntern provides a safe, assistive AI triage layer that prioritizes blood smear patches by anomaly signal — without ever making a clinical diagnosis. The pathologist receives a prioritized worklist with pattern-similarity context, not a diagnostic report.

Prioritized Worklist

Critical patches surface first. Pathologist time is applied where the anomaly signal is highest.

Pattern Similarity Context

Top-k similar previously reviewed patches are displayed alongside each result, providing morphological reference.

Verdict Capture

Pathologist agrees, disagrees, or modifies the AI tier. Every verdict is audited and attributed.

Explicit Non-Goals

These boundaries are architectural constraints — not limitations — and are enforced at every layer of the system.

No Automated Diagnosis

PathoIntern never assigns, suggests, or implies a clinical diagnosis. Output is triage priority only.

No CBC Data

The system operates on digitized image patches only. No lab values, clinical history, or CBC results are ingested.

No Unsupervised Decisions

Every scored patch requires pathologist review. The system cannot close or dismiss a case autonomously.

No Diagnostic Probability

Scores are labelled as pattern similarity to reviewed cases — never as probability of disease or condition.