Clinical decision support that reasons like a pathologist: read a surgical report clinically, give a frozen-section answer within its limits, build the IHC algorithm for the unknown primary, and order the molecular test that changes therapy — with the decision rule and exactly what would change it. Built for verified physicians.
Two real pathology questions, each reasoned the way HeyHippocrates answers: the skills fire, the answer is cited, the decision rule is explicit, and what's missing is flagged. Scroll to each.
The surgeon asks for a frozen on a lung mass and its margin mid-resection — what can I tell them, and what are the limits?
A frozen section answers the questions that change the operation in progress: is the lesion benign or malignant, is the margin involved, and is the tissue lesional and adequate for permanent analysis. Give a rapid provisional answer with your confidence, but respect the limits — freezing artifact degrades fine cytologic and architectural detail, some entities cannot be resolved intraoperatively, and lymphoma, some spindle lesions and definitive grading are best deferred. Do not over-call: report what is certain, defer what is not, and correlate with the permanent sections.
Decision ruleFrozen answers benign-vs-malignant / margin / adequacy to guide the operation; freezing artifact limits fine detail; defer some diagnoses (lymphoma, definitive grading, some spindle lesions); give a provisional answer + confidence, do not over-call, correlate with permanent.
Not established from this alone — the tissue quality, the entity and the surgical question change what a frozen can answer.
A metastatic adenocarcinoma with no obvious primary — how do I work up the origin?
Work it up with a stepwise immunohistochemistry algorithm rather than a shotgun panel. Start broad with the CK7 and CK20 pattern to narrow the field, then add lineage-specific markers — TTF-1 for lung and thyroid, CDX2 for a gastrointestinal origin, GATA3 for breast and urothelial, PAX8 for renal, gynecologic and thyroid — and correlate with the clinical picture and imaging. Add molecular profiling where it changes management. Report the most likely origin with an honest level of certainty, not a false precision.
Decision ruleMetastatic adenocarcinoma, unknown primary → stepwise IHC: CK7/CK20 pattern → lineage markers (TTF-1 lung/thyroid, CDX2 GI, GATA3 breast/urothelial, PAX8 renal/gyn/thyroid) + clinical/imaging correlation; molecular where it changes management; report the likely origin with a stated certainty.
Not established from this alone — the morphology, the clinical context and the marker pattern change the panel and the conclusion.
Illustrative simulations. The physician always decides.
The exact questions your field searches for — answered in the format the knowledge base speaks: the answer, the decision rule, what's missing, and the source.
Read the report as a structured argument, not a single line. The diagnosis line is the conclusion; the gross and microscopic descriptions, the tumor type and grade, the margins, the lymphovascular invasion, the node status and any synoptic staging data are the evidence. Weigh the pathologist certainty language — definitive versus suggestive versus deferred — and note pending stains or molecular tests. When the report does not fit the clinical or radiologic picture, call the pathologist; correlation resolves more discrepancies than a repeat biopsy.reporting standards
Decision ruleRead the whole report: diagnosis (conclusion) + type/grade/margins/LVI/nodes/synoptic (evidence) + certainty language + pending tests; when it does not fit the clinical or imaging picture, correlate with the pathologist before re-biopsy.
Missing data: the clinical context, the specimen and pending studies change how you read the report.
A frozen section is for the decision that cannot wait: is it benign or malignant, is the margin clear, is the tissue diagnostic and adequate, is a sentinel node involved. It trades resolution for speed — ice-crystal artifact obscures fine cytology and architecture, fat and calcified tissue freeze poorly, and definitive subtyping, grading and lymphoma diagnosis usually wait for permanent sections and stains. Give a clear provisional read with the confidence, and reconcile every frozen with its permanent.intraoperative consultation
Decision ruleFrozen = rapid intraoperative answer (benign/malignant, margin, adequacy, sentinel node); limited resolution (freezing artifact, fat/calcium poor); defer definitive subtyping/grading/lymphoma to permanent; state confidence; always reconcile frozen with permanent.
Missing data: the tissue type, the clinical question and the artifact change what the frozen can deliver.
Use IHC as a decision tree, not a scattergun. First classify the broad lineage — carcinoma, melanoma, lymphoma or sarcoma — with a small screening panel. For a carcinoma, the CK7 and CK20 combination narrows the likely origin, and lineage markers then refine it: TTF-1 and napsin A for lung, CDX2 and SATB2 for lower gastrointestinal, GATA3 for breast and urothelial, PAX8 for renal, gynecologic and thyroid. Always interpret the panel against morphology and the clinical picture, and add or subtract markers as the pattern evolves.IHC algorithm
Decision ruleIHC as a decision tree: screen lineage (carcinoma/melanoma/lymphoma/sarcoma) → for carcinoma, CK7/CK20 → lineage markers (TTF-1/napsin lung, CDX2/SATB2 lower GI, GATA3 breast/urothelial, PAX8 renal/gyn/thyroid); interpret against morphology + clinical context; adapt the panel iteratively.
Missing data: the morphology, the first-round pattern and the clinical context change the next markers.
Order the molecular test that carries a therapeutic or prognostic consequence, not everything possible. In lung adenocarcinoma, EGFR, ALK, ROS1 and others open targeted therapy; in colorectal cancer, RAS and BRAF and mismatch-repair or microsatellite status guide biologics and immunotherapy; PD-L1 and tumor mutational burden inform immunotherapy across tumors; and BRCA and other markers guide specific agents. Ensure the specimen is adequate and preserved for the assay, integrate the result with morphology and stage, and report it so it drives the treatment decision.molecular pathology
Decision ruleTest for actionable/prognostic markers, not everything: lung (EGFR/ALK/ROS1), colorectal (RAS/BRAF/MMR-MSI), pan-tumor (PD-L1, tumor mutational burden, mismatch repair), BRCA and others; ensure specimen adequacy; integrate with morphology + stage; report to drive therapy.
Missing data: the tumor type, the specimen adequacy and the clinical goal change the test panel.
Your whole practice, one specialist brain — grounded in champion-authored, cited knowledge.
A generalist pathology colleague across the surgical report, the frozen section, immunohistochemistry, molecular testing, cytopathology and the critical result — cited, non-directive, always with the decision rule.
Deep specialist agents for surgical pathology, cytopathology, hematopathology and molecular pathology — each its own soul and knowledge base — plus your own uploaded materials.
From the frozen-section call to the actionable mutation the margin is thin. Only one of these reasons like a pathology colleague whose every claim you can trace.
The waitlist is a founding cohort. Members shape the pathology build, get in before anyone else, and lock the founding price for good.
Leave your name and work email. We'll reach out the moment pathology opens — with your Founding Pathologist offer.
One email at launch. No spam, ever.
It reasons in the frameworks you use — how to read the surgical report, the frozen-section limits, the immunohistochemistry algorithm and actionable molecular testing — and always shows the reasoning, the decision rule and what would change it. The decision always stays your call.
Both. Residents use it to pressure-test a plan and learn the decision rule; attendings use it as a fast, cited second opinion at the scope and the frozen bench. Access is gated to verified physicians.
Surgical pathology, cytopathology, hematopathology and molecular pathology each have their own deep specialist agent on the Pro plan, on top of the full pathology scope on Basic.
No. It is physician-facing clinical decision support — a cited thinking partner. It does not diagnose, does not treat, and gives no patient-facing advice. The decision always stays with you.
Every answer is grounded only in champion-authored, cited pathology knowledge — traceable, not hallucinated — and always states the decision rule plus what data is missing, instead of a confident guess.
HeyHippocrates