Clinical decision support · Hepatocellular carcinoma
Catch liver cancer earlier, on the ultrasound machines clinics already own.
LiverAI is an AI second reader for liver ultrasound. It reviews each scan in seconds, highlights suspicious regions, and returns a clear risk score, giving frontline clinicians the kind of support that normally requires a specialist radiologist on site.
The gap we close
Ultrasound is everywhere. The expertise to read it is not.
Across Uganda and much of Sub-Saharan Africa, ultrasound machines reach far more patients than radiologists ever could. Yet early hepatocellular carcinoma is subtle on a B-mode scan, and reading it confidently demands specialist training that most clinics simply do not have on hand.
The result is a quiet emergency. Patients arrive late, when the disease is already advanced and treatment options have narrowed. LiverAI is built to change where that story begins, by putting reliable second-reader support at the point of care.
How it works
Three steps that fit the workflow clinics already run.
LiverAI adds intelligence to the scan a clinic is already performing. There is no new hardware to buy and no change to how patients are seen.
Scan as usual
A sonographer captures a standard B-mode or contrast-enhanced liver ultrasound, exactly as they would today, then uploads the image through a secure interface.
LiverAI reads it instantly
The model analyses the visual patterns associated with hepatocellular carcinoma, highlights the suspicious region on the image, and returns a probability score within seconds.
The clinician decides
A clear, shareable report supports referral, follow-up imaging, and documentation. LiverAI supports the clinician and never replaces clinical judgement.
The product
Everything a clinic needs to act on a scan with confidence.
LiverAI is a complete decision support workspace, from the moment a patient is registered to the moment a report is shared with a referring specialist.
Lesion heatmaps
Every result shows where the model is looking, overlaying the suspicious region so clinicians can see the reasoning rather than trust a number alone.
Probability scoring
A calibrated risk score and confidence level turn a complex image into a clear signal that supports faster, better referral decisions.
One-tap reports
A professional PDF brings together patient details, risk factors, the annotated image, and the recommendation, ready to share or attach to a referral.
Patient records
Register patients with the clinical risk factors that matter for the liver, including hepatitis B and C, cirrhosis, alcohol use, diabetes, and family history.
Works on existing kit
The tool accepts standard ultrasound images, so a clinic can begin without buying new machines or rebuilding its imaging setup.
Resilient to image quality
The model is designed for the variable, often noisy images that real clinics produce, not the pristine scans of a research laboratory.
Analytics dashboard
A clear overview tracks scan volume, detection rates, and risk factor patterns, helping a facility understand its own caseload over time.
Runs on modest hardware
Inference is light enough to run on the kind of computing a clinic can realistically afford, keeping results fast without a data centre.
Technology
Deep learning that earns clinical trust.
LiverAI is built on a dual-path neural network that learns from both standard and contrast-enhanced ultrasound. The goal is not only accuracy on a benchmark, but reasoning a clinician can see, data a hospital can govern, and results a clinic can rely on.
Dual-path imaging model
A convolutional architecture reads B-mode and contrast-enhanced ultrasound together, capturing detail that a single view alone would miss.
Explainable by design
Each prediction is paired with a visual heatmap, so the model presents interpretable evidence instead of an opaque conclusion.
Privacy first
Images are de-identified and the system can run on a facility's own infrastructure, keeping sensitive patient data under local control.
Performance figures reflect evaluation on held-out de-identified ultrasound data. LiverAI is a decision support aid intended for use by qualified healthcare professionals under clinical supervision, and it does not provide an autonomous diagnosis.
Why LiverAI
Designed for the realities of African clinics.
Global imaging tools are built for hospitals with abundant bandwidth, pristine images, and specialist staff. LiverAI is built for everywhere else, where the need is greatest and the margins are thinnest.
Built for low-resource settings
The system tolerates variable images and limited connectivity, so it performs in the field rather than only in a flagship hospital.
Specialised for the liver
This is a focused oncology tool for hepatocellular carcinoma, not a general scanner adapted to many tasks, which keeps it sharp where it counts.
Affordable at the point of care
Expert-level support arrives at a fraction of the cost of CT or MRI, and it works on the ultrasound a clinic already has.
See it in action
Walk through a live LiverAI analysis.
Explore an interactive demonstration of the diagnostic workspace, from uploading a scan to reviewing the highlighted finding and the generated report. We can also arrange a guided session for your clinical team.
The team
Engineers and clinicians, building side by side.
LiverAI brings together machine learning expertise and frontline clinical insight, so the product is shaped by the people who build the models and the people who care for patients.
Get in touch
Bring LiverAI to your clinic or programme.
Whether you run a clinic, a referral hospital, a screening programme, or an investment that wants to widen access to early cancer detection, we would like to hear from you. Tell us a little about your work and we will follow up.