Platform / Technology / AI-assisted diagnostics
Pillar guide · AI-assisted diagnostics
AI-Assisted Diagnostics for Fluorescent Point-of-Care Cancer Testing
OncoFirm™ is building AI-assisted decision support into its fluorescent lateral flow platform. The reader turns every test strip into measured, time-stamped data. Validated software then checks the run, calculates a standardized result and, as the menu grows, helps make sense of multiplex panels and results over time. Clinicians stay in charge of every diagnosis.
At a glance
What it is
Software that checks, quantifies and organizes test data for a qualified professional
Where it runs
The OncoFirm™ handheld fluorescence reader and connected software
First role
Strip-level quality assurance for the CEA and PSA programs
Next roles
Multiplex interpretation, serial-result trend support, connected records
Status
In development · not FDA cleared or approved · first products research use only
On this page
Definition
What are AI-assisted diagnostics?
AI-assisted diagnostics use validated software models to check, quantify and organize diagnostic data, then present the result, with quality indicators, to a qualified healthcare professional who makes the clinical decision.
In laboratory medicine the most useful AI is often the least visible. It catches a test that did not run correctly, applies the same calibration every time, and lays out a series of results so a change is easy to see. That is the kind of AI OncoFirm is building: practical, measurable and accountable.
What it is designed to do
What it is not designed to do
Confirm each strip ran correctly before a result is shown
Make a cancer diagnosis or treatment decision
Turn fluorescence signal into a standardized, quantitative value
Replace laboratory confirmation or imaging
Flag results that need a repeat or a closer look
Hide how a result was produced
Show serial results side by side for a clinician
Change its own behavior in the field without validation
Foundations
Why AI starts with a measured fluorescent signal
Software can only be as consistent as the data it receives. A conventional rapid test produces a colored line that a person judges by eye, so two people in two rooms can read the same strip differently. OncoFirm replaces that judgment with measurement.
Whole-strip imaging
The handheld reader images the entire strip, not a single spot, recording the full fluorescence profile across the test, reference and control lines.
Timed reads
Signal is captured at several time points, so the shape of the development curve can reveal flow or release problems that a single snapshot would miss.
Built-in reference line
Each strip carries a fluorescent reference line of known brightness. The test-to-reference ratio corrects for reader-light and temperature drift inside every test.
The result is structured, traceable data: a value, a time series and a quality record for every test, which is the starting point for any trustworthy analysis. Read how fluorescent and gold-nanoparticle assays compare.
How it works
From strip to standardized result in five steps
Every step is designed to be logged, so a result can be traced back to the strip lot, calibration and checks that produced it. See the digital diagnostic reader and the fluorescent lateral flow platform for the hardware side.
01
Sample and assay
A small blood, serum or plasma sample is applied to an OncoFirm™ fluorescent strip designed for the target biomarker, such as CEA or PSA.
02
Signal capture
The reader excites the fluorescent labels and images the whole strip at several time points.
03
Signal processing
Background correction, test-to-reference ratio and the lot calibration curve convert raw light into a standardized value, planned to trace to WHO reference materials.
04
Model-assisted checks
Validated models review the signal profile, flag abnormal flow or reagent release, and confirm the reference and control lines behaved as expected.
05
Decision support output
The professional sees a quantitative result with quality indicators and, where relevant, prior results for comparison. Interpretation stays with them.
Where it's headed
AI roadmap: growing with the test menu
AI capability is planned to expand in step with the platform. Each phase is validated on its own before the next one is released, and timing depends on partnerships and funding.
Phase 1
Quality assurance
Run-validity checks on every strip: reference-line verification, flow and release anomaly detection from timed reads, and automatic flagging before a result is reported. Built alongside the CEA and PSA programs.
Phase 2
Quantitation and multiplex interpretation
Consistent quantitation across reagent lots, then separate, cross-checked results for planned panels such as the bloodborne virus panel and the cardiac marker panel.
Phase 3
Serial-result trend support
Side-by-side display of a patient’s results over time, starting with tumor markers like CEA where change between measurements is what clinicians watch.
Phase 4
Connected data and research insight
Secure exchange with laboratory and health record systems, and de-identified datasets that let research partners study biomarker patterns at scale.
Applications
Where AI-assisted analysis adds value
Cancer biomarker monitoring
Multiplex panels
Separate, quality-checked values for each target on one strip, from the bloodborne virus panel to the cardiac marker panel.
Infectious disease screening
The same reader and checks across the assay pipeline, giving consistent reads in clinics, field programs and low-resource settings.
Research and clinical studies
Traceable, structured data for translational research and clinical collaborations, including biomarker discovery work.
Responsible AI
Principles built into the design
These principles follow the WHO guidance on Ethics and governance of artificial intelligence for health and FDA expectations for AI-enabled device software.
Human in the loop
Results support, and never replace, the judgment of qualified laboratory and clinical staff.
Validated, locked models
Models are fixed and validated before release; any update follows a documented change-control process.
Data quality and fairness
Training and test data are planned to span sample types, reagent lots and diverse populations, with performance checked across subgroups.
Transparency
Each result shows its quality indicators and how it was produced, not just a final number.
Traceability
Calibration is planned to trace to WHO reference materials, and every result links back to its strip lot and checks.
Privacy and security
Access controls, audit trails, encrypted transfer and de-identified research datasets are part of the design.
Regulatory path
How AI-assisted features will be brought to market
AI-assisted functions on the OncoFirm platform are part of the device software, so they are developed and validated with the assays and reader they run on. The planned path is:
- Research use only first. Early products and software are intended for research and development settings.
- Analytical validation. Precision, linearity, lot-to-lot performance and interference testing for each assay and software function.
- Clinical evaluation. Studies in intended-use populations before any clinical claim is made.
- FDA review. Submissions for clinical use, including a plan for how validated software updates will be managed.
FAQ
Frequently asked questions
What are AI-assisted diagnostics?
AI-assisted diagnostics use validated software models to check, quantify and organize diagnostic data, then present the result with quality indicators to a qualified professional. The software supports the decision; it does not make the diagnosis.
Does the OncoFirm platform diagnose cancer on its own?
No. The platform is designed to report a measured biomarker value, such as CEA in ng/mL, with quality flags. Interpreting that value alongside history, imaging and other tests remains the job of the clinician.
Why does AI need a fluorescent reader rather than a visual test?
A visual line gives a yes/no impression that varies with lighting and the reader's eye. The OncoFirm reader images the whole strip at several time points and records numbers, which is the kind of consistent, structured data software can learn from and check.
What will AI do first on the OncoFirm platform?
The first role is quality assurance: confirming each strip ran correctly, checking the fluorescent reference line and flagging abnormal flow before a result is reported. Quantitation support and multiplex interpretation follow as the test menu grows.
Is OncoFirm's AI-assisted software cleared by the FDA?
No. The platform and its software are in development and have not been cleared or approved by the FDA. First products are planned for research use only. Any future clinical claim would require analytical and clinical validation and FDA review.
How will patient data be protected?
The data architecture is being designed with privacy and cybersecurity requirements in mind: de-identified research datasets, access controls, audit trails and encrypted transfer to laboratory and health record systems.
Keep reading
Related guides
Sources
References
- Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine. 2019;25:44–56. doi.org/10.1038/s41591-018-0300-7
- Esteva A, Robicquet A, Ramsundar B, et al. A guide to deep learning in healthcare. Nature Medicine. 2019;25:24–29. doi.org/10.1038/s41591-018-0316-z
- Bera K, Schalper KA, Rimm DL, Velcheti V, Madabhushi A. Artificial intelligence in digital pathology — new tools for diagnosis and precision oncology. Nature Reviews Clinical Oncology. 2019;16:703–715. doi.org/10.1038/s41571-019-0252-y
- Liu X, Faes L, Kale AU, et al. A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging. The Lancet Digital Health. 2019;1:e271–e297. doi.org/10.1016/S2589-7500(19)30123-2
- U.S. Food and Drug Administration. Artificial Intelligence-Enabled Medical Devices. www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-enabled-medical-devices
- World Health Organization. Ethics and governance of artificial intelligence for health. Geneva: WHO; 2021. www.who.int/publications/i/item/9789240029200
- National Cancer Institute. Artificial Intelligence (AI) and Cancer. www.cancer.gov/research/infrastructure/artificial-intelligence
Development status. The OncoFirm™ platform, its assays and its AI-assisted software are in development. They have not been cleared or approved by the FDA and are not available for sale. First products are planned for research use only. This page describes design goals, not demonstrated performance.
Build the next generation of diagnostic intelligence with us
We are looking for research, clinical, data-science and commercial partners. Start a collaboration inquiry or contact us directly.