BRENNEN

Industries / Health / Healthcare / Medical labs

Blood and chemistry analyzers

Machines that count blood cells and measure salts, sugars, proteins, hormones and drugs in blood, from the hospital's main lab to a handheld reader at the bedside.

7

Milestones achieved, from 1953 to 2026.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1953

    Counting cells by current

    Wallace Coulter's patent, filed in 1949 and granted on 20 October 1953, passes a dilute suspension through a tiny opening that carries an electric current. Each particle that passes changes the resistance, so each pulse counts one particle and its size shows in the pulse; the patent names counting red blood cells as a use. By 1993 a Coulter counter was the reference for hematocrit (the share of blood made of red cells) in a University of Pennsylvania evaluation.

    patents.google.compubmed.ncbi.nlm.nih.gov

    Credit

    Wallace H. Coulter

  2. 1957

    A machine runs the chemistry

    Blood chemistry used to mean adding a reagent to serum, waiting for the proteins to settle out, filtering and measuring, by hand for each sample. Leonard Skeggs, in Cleveland, built a machine that pumped samples and reagents through tubing in a continuous stream, removed proteins by dialysis through a membrane, and kept each sample apart with air bubbles on its way to a colorimeter. A Technicon salesman saw his prototype in February 1954; Skeggs published the method in September 1957, and Technicon sold it as the AutoAnalyzer.

    pmc.ncbi.nlm.nih.goveuropepmc.org

    Credit

    Leonard T. Skeggs Jr.; Technicon Instruments

  3. 1978

    Chemistry on a dry slide

    H. G. Curme and colleagues described thin, dry films that carry all the reagents for a test: a drop of about 10 microliters of serum spreads through a porous top layer, reacts in the layers below, and the color is read by reflected light, with no liquid reagents to mix. Their glucose film varied by 1.5% on control serum over 20 days. By 1993 a dry-slide analyzer, the Kodak Ektachem 700, served as the main-lab reference in a University of Pennsylvania evaluation.

    pubmed.ncbi.nlm.nih.govpubmed.ncbi.nlm.nih.gov

    Credit

    H. G. Curme and colleagues

  4. 1993

    A lab in the hand

    The i-STAT analyzer is a handheld reader with disposable cartridges whose sensors are thin-film electrodes made on silicon chips, the way computer chips are made. An evaluation at the Hospital of the University of Pennsylvania, in the emergency department and the urgent-test lab, found sodium results repeatable to within 0.46% to 0.89%. Its chemistry results agreed with the main lab's analyzer except for chloride, where the narrow range of values limited the comparison, and its hematocrit read differently from the lab's Coulter counter.

    pubmed.ncbi.nlm.nih.gov

    Credit

    K. A. Erickson and P. Wilding, who ran the evaluation

  5. 2005

    Software sorts white cells

    The CellaVision DM96 photographs the cells on a stained blood smear and uses a neural network to sort white cells into types for a technologist to confirm. In an evaluation at Massachusetts General Hospital, technologists agreed with 82% of its first classifications. After the authors set aside differences they judged insignificant, it caught every smear with blasts (immature cells that can signal leukemia) and wrongly flagged 6% of smears without them.

    pubmed.ncbi.nlm.nih.gov

    Credit

    CellaVision AB of Lund, Sweden; Alexander Kratz and colleagues, who ran the evaluation

  6. 2017

    A vitamin fools the tests

    Many immunoassays hold their antibodies in place with a link between biotin, a B vitamin, and a protein called streptavidin, as Roche's FDA file for one troponin test describes. On 28 November 2017 the FDA warned that biotin taken as a supplement, often sold for hair, skin and nails, can cause "clinically significant incorrect lab test results". It reported that one patient taking high doses died after a falsely low troponin result, the main blood test for a heart attack, and urged clinicians to ask patients about biotin.

    LessonTests must be checked against what patients actually take, and a result that does not fit the patient calls for a question about supplements.

    pharmacytimes.comaccessdata.fda.gov

    Credit

    US Food and Drug Administration

  7. 2026

    Bone marrow read on screen

    Hematologists at the Gemelli university hospital in Rome ran 350 bone marrow samples from routine work through the Scopio X100HT, which scans the whole smear at high magnification and pre-sorts cells with AI, and compared its counts with the microscope. After expert review the counts matched closely for the main cell types (a correlation of 0.93 for two of them, on a scale where 1 is a perfect match). The machine's own pre-sort matched less well for blast counts above 5% (0.59), and 16 very dense samples could not be counted. Siemens Healthineers, whose name the system carries in the study, helped fund it.

    LessonAI pre-sorting supports an expert but does not replace one for rare and abnormal cells.

    pubmed.ncbi.nlm.nih.gov

    Credit

    Gina Zini and colleagues, Fondazione Policlinico Universitario A. Gemelli

Watch for

  1. Mass spectrometry goes routine

    Mass spectrometry identifies and measures molecules by their weight and is often treated as the reference method for steroid hormones, vitamin D and some drugs, but it has needed specialist staff. Roche says its cobas Mass Spec solution now runs 39 tests, fully automated beside its routine analyzers, the broadest menu of any automated mass spectrometry platform, and shortens turnaround times; one of its tests is the first mass spectrometry test placed in the US "moderate complexity" category, which lets non-specialists run it. A first outside evaluation, at La Paz University Hospital in Madrid, found good repeatability (variation of 3.4% at most) but large differences from the immunoassays it would replace, such as -36.5% for estradiol and -28.5% for aldosterone.

    Achieved whenlabs outside Roche publish turnaround times, staff time and failure rates from routine use.

    roche.compubmed.ncbi.nlm.nih.gov

    Claimed by

    Roche, December 2025

Open

  1. 01

    An open table of interferences

    A free, searchable table of which substances disturb which blood tests, and at what levels: hemoglobin from burst red cells, bilirubin, fats, biotin and common drugs. Build it with AI models from the FDA's public 510(k) decision summaries, the review documents for each cleared test, and have trained people check a sample.

    The brief
    Credit · unclaimed
  2. 02

    A hormone method switch table

    An open table that shows, for each hormone and each brand of immunoassay, how far results sit from mass spectrometry at the levels where doctors act, drawn from published method-comparison studies, so that a clinic knows what changes when its lab switches method.

    The brief
    Credit · unclaimed

Open 01 · the brief

An open table of interferences

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A free, searchable table of which substances disturb which blood tests, and at what levels: hemoglobin from burst red cells, bilirubin, fats, biotin and common drugs. Build it with AI models from the FDA's public 510(k) decision summaries, the review documents for each cleared test, and have trained people check a sample.

Why it is still open

Each decision summary lists the substances tested, but in PDF tables whose columns come apart when the text is extracted, as in Roche's troponin T summary K162895. A 2026 study notes that package inserts often leave out how interference was tested, and some do not state the hemoglobin level that affects a result. The largest compilation, Effects on Clinical Laboratory Tests, built on Donald Young's reference books, is a paid service sold through Wiley, and its makers say they have no plans to let users add entries. A search found no open, machine-readable compilation.

accessdata.fda.govhrcak.srce.hrclinfx.wiley.comhealthcare.uiowa.edu

Built with what exists

the FDA 510(k) database and its decision summary PDFs; pdftotext or another PDF table reader; AI models that read tables and scanned pages, such as Claude or Gemini, or an open-weight model such as Mistral Large 4 (see the Note "A trillion open parameters"); a simple web page for lookups. The advance is AI agents that can work through hundreds of long technical documents and cite the page for each value, as in Enveda's Claude-based review of drug-making records (see the Note "Seven more flagged").

  1. First step

    Download the decision summaries for ten troponin and thyroid tests, extract their interference tables with one model, and check them by hand.

  2. Done looks like

    At least 300 decision summaries for chemistry and immunoassay tests, with every interference row (substance, highest level without a significant bias, bias limit, test level) linked to its page. Two people with lab training check a random 15% and agree with the table on at least 98% of fields. Table and code are published under an open license, with a page that answers which listed tests biotin affects at a given level.

Open 02 · the brief

A hormone method switch table

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open table that shows, for each hormone and each brand of immunoassay, how far results sit from mass spectrometry at the levels where doctors act, drawn from published method-comparison studies, so that a clinic knows what changes when its lab switches method.

Why it is still open

Outside studies keep finding large gaps. A 2020 study of six steroid hormones on routine analyzers found immunoassays differed from mass spectrometry by -31% to +137%, and the 2026 La Paz evaluation of Roche's automated system found -36.5% for estradiol and called for close communication between clinicians and the lab. The US Centers for Disease Control and Prevention's Hormone Standardization Program certifies testosterone and estradiol tests that come within set limits of its reference method, such as a mean bias within 6.4% for testosterone, but it covers two hormones, gives a pass or fail, and lists only labs that choose to be listed. The results sit in separate papers, and a search found no compiled open table.

pmc.ncbi.nlm.nih.govpubmed.ncbi.nlm.nih.govcdc.gov

Built with what exists

PubMed and Europe PMC for open-access papers; OpenAlex for citations; AI models to pull slopes, intercepts and biases from tables and figures; R or Python to compute bias at decision levels; two human checkers.

  1. First step

    Extract the bias tables from ten open-access comparisons of testosterone immunoassays with mass spectrometry.

  2. Done looks like

    At least 100 method-comparison studies covering testosterone, estradiol, cortisol, aldosterone, progesterone and vitamin D, each row linked to its paper. Two checkers verify 20% of rows with at least 98% agreement, and the table is published openly and updated each year.