VIVARUS
Diagnostics
Rare Disease Intelligence at the Point of Care
We are building a clinical decision support system that reads the hospital record a patient already has and surfaces the rare diagnoses that fit it — each one shown against the published criteria it matches, the evidence that supports it, and the evidence that argues against it.
The delay is not in the patient. Its in the system.
In the largest European survey of its kind, patients waited an average of 4.7 years from first symptom to confirmed diagnosis. Roughly nine tenths of that time passed after they had already sought medical help. Seventy-three per cent were diagnosed with a different disease at least once along the way.
That gap is not a failure of clinical skill. A hospital haematologist may see a given ultra-rare condition once in a career, and rare presentations arrive disguised as common ones. But by the time the diagnosis is finally made, the evidence for it is usually already in the record — scattered across years of laboratory results, notes, referrals and discharge summaries, and never assembled into a single view.
That assembly is the problem we work on.
Vivarus Diagnostics grew out of years spent working in rare disease therapy development, where we saw a recurring pattern: patients waiting far too long for answers. Despite the dedication of clinicians, the sheer complexity and rarity of these conditions often makes early recognition incredibly difficult. With backgrounds in biochemistry, medicine, and pharma, we started Vivarus Diagnostics to build tools that can quietly support clinicians behind the scenes — making rare feel a little more recognizable.
300M+
people worldwide live with a rare disease — between 3.5% and 5.9% of the population
4.7 years
average time from first symptom to confirmed diagnosis across 41 European countries
~90%
of that delay falls after the first medical contact, not before it
7.6x
higher direct healthcare costs before diagnosis than matched controls, in German claims data
Sources, in order: Nguengang Wakap et al., European Journal of Human Genetics (2020) · EURORDIS Rare Barometer, European Journal of Human Genetics (2024), n=6,507 · same · Orphanet Journal of Rare Diseases (2025), n=1,243.
Mission & Vision
Our mission is to accelerate the diagnosis of rare diseases by equipping clinicians with AI-powered tools that support earlier recognition and smarter decision-making — without adding complexity to their workflow.
We envision a world where no patient waits years for a diagnosis simply because their condition is rare. By bridging biomedical science with intelligent technology, we aim to make rare disease diagnosis faster, more accurate, and accessible across all levels of care.
Our Values
Patients first
Behind every delayed diagnosis is someone who spent years being told the problem was something else, or nothing at all. That is the standard the work is measured against.
Clinical trust
Clinical judgement stays with the clinician. The system is designed to widen what a doctor considers, never to narrow it or decide for them.
Patient-Centered Innovation
Innovation means nothing if it doesn’t reach the right people. We design technology that is intuitive, clinically meaningful, and built to improve lives.
Where this goes
Validation across independent clinical environments, certification, then deployment. First in haematology and oncology, then wider. If you are an investor, a hospital, a research group or a rare disease therapy developer with a stake in patients being found sooner, we would like to hear from you.
Why us
Vivarus Diagnostics came out of years spent in rare disease therapy development, watching the same pattern from the commercial side: therapies reaching approval for patients who were not being identified in time to receive them. The bottleneck had moved upstream of treatment, into recognition.
The founding team combines biochemistry and machine learning, senior operating experience across rare disease pharmaceutical companies, and a medical qualification paired with regulatory and legal practice. We have brought rare disease products through European regulatory systems before. We know what the evidence has to look like before anyone will accept it.
We are building this as a regulated medical device
Vivarus Diagnostics is being developed toward MDR Class IIa conformity assessment as software as a medical device. That means an ISO 13485-aligned quality management system, ISO 14971 risk management, and a clinical evaluation programme built on clinician-adjudicated retrospective data followed by prospective multi-site pilots. Development, validation and evaluation datasets are held separate at the level of the individual patient record, so that nothing we evaluate on has been learned from.
We say this early and plainly because the alternative is worse. A system that offers diagnostic suggestions to clinicians without a regulatory basis is not a faster route to market; it is an unbounded liability sitting inside a hospital. The certification path is not overhead on the product. It is most of the product.
Vivarus Diagnostics is in development. It is not CE-marked, is not currently a medical device, and is not available for clinical use. Nothing on this site is a claim of diagnostic performance.
Our Solution
A second read on the record you already have
Vivarus Diagnostics is developing a clinical decision support system for hospital hematology and oncology. It is designed to work on data the hospital already holds: laboratory results, medications, diagnoses and procedures, together with unstructured clinical notes and reports. We are building it to read them as a trajectory rather than a snapshot, because distinguishes a rare disease is often the shape of the change over time rather than any single abnormal value.
Four commitments shape how it is built. They are also the four things we would want to interrogate in someone else's system.
1. Grounded in guidelines, not a score
Each condition the system surfaces is presented against the published diagnostic criteria it matches: which findings in the record support it, which contradict it, and which evidence is absent that would settle the question. The reasoning is traceable to its source. A ranked list a clinician cannot interrogate is no longer decision support but an instruction, which is the wrong approach for rare disease.
2. A deterministic layer that can override the model
Rule-based logic sits alongside the language model and is not subject to it: eligibility and exclusion checks, contradiction detection, evidence-completeness validation, and suppression of any output that cannot be justified from the record in front of it. These rules are fixed, testable and versioned. Probabilistic reasoning proposes; deterministic clinical logic disposes.
3. Selective activation, not universal flagging
The system is designed to engage rare-disease reasoning only where the record warrants it. This is a clinical decision before it is an engineering one. A tool that raises a rare possibility on every atypical presentation produces alert fatigue, unnecessary testing and cost — and would be abandoned within a month. The value of rare disease work comes from justified escalation, not more escalation.
4. Auditable by construction
Every inference is designed to be reproducible after the fact: the input snapshot, the pipeline version, the model version, the ranked output and the reasoning behind it are recorded together. This is a regulatory requirement, but it is also the only honest basis for a clinician to trust a system of this kind, and the only way to investigate it when it is wrong.
We are starting in one place
Our first indication area is haematology and oncology. Rare and rare-adjacent conditions there are routinely mistaken for more common ones, the data is comparatively well structured, and the cost of a late diagnosis is measured in months of viable treatment. The disease library currently covers approximately 227 conditions in this area
We would rather be genuinely useful in one specialty than plausible across all of them. Further specialties follow once the first has been validated and certified.
News
