How FullPanel reads your bloodwork
Most tools either flag everything against a population average or hand your labs to a chatbot that invents ranges. FullPanel does neither. Here’s exactly how it works.
Extract
Your results are read from a pasted panel or a lab PDF/photo into structured markers — value, unit, and the lab's own reference range. Units are normalized to SI (mmol/L, nmol/L), so US and Canadian reports read the same.
Map
Every row is matched to one of FullPanel's canonical markers by a deterministic dictionary of names and synonyms — so 'SGPT', 'ALT', and 'Alanine aminotransferase' all resolve to the same marker.
Contextualize
A fixed layer of clinical rules evaluates each marker against context-adjusted ranges for your mode and sex — and reads markers together (AST with GGT, testosterone with LH) the way a knowledgeable human would.
Explain
Each marker is grouped into act, watch, or within-range, shown against both the context-adjusted and the standard lab range, with a plain-language note. A language model phrases the explanation; it never sets the verdict.
The Context Engine
The core idea: a number only means something in context. A testosterone of 38 nmol/L is alarming for an untreated man and completely expected on TRT. A “normal” ferritin can still be low for a heavy blood donor. FullPanel encodes that context as rules and range sets rather than leaving it to a generic model to guess each time.
Two modes, one engine. Optimization reads against optimal sub-ranges for general health. Enhancedreads against expectations for someone on exogenous hormones, so it doesn’t panic over suppressed LH or supraphysiologic testosterone — but still escalates a genuinely dangerous value like a very high hematocrit.
The model is on a leash.Reference ranges and flagging thresholds are fixed, human-authored data. The language model extracts values and writes prose — it cannot change a range, invent one, or override a rule. That’s the difference between this and pasting your labs into a chatbot.
What FullPanel will not do
It does not diagnose disease, and it never recommends or adjusts a medication, compound, or dose. Urgent values are escalated with clear “see a physician now” language. The goal is to make your data legible and tell you what’s worth a conversation with a doctor — nothing more.
About the ranges
FullPanel currently covers 66 markers across the common panels. Standard ranges reflect conventional laboratory references; context-adjusted ranges are estimates drawn from clinical and harm-reduction literature and are pending formal clinical review. We show both, always, so you can see the difference yourself.
FAQ
No. Ranges and the rules that flag markers are fixed, human-authored data — never generated on the fly. A language model is used only to extract values from a PDF and to phrase the explanation; it cannot override a range or invent one.
No. It is an educational tool that organizes and contextualizes your results and tells you what's worth discussing with a physician. It does not diagnose disease or recommend treatment.
Enhanced-mode ranges encode what is expected for someone on exogenous hormones — e.g. suppressed LH, supraphysiologic testosterone — drawn from harm-reduction literature. They exist to prevent false alarms, not to set a target. All ranges are pending formal clinical review.