Alibaba’s DAMO RADAR: The AI Model Changing the Conversation Around Cancer Detection

 By Pamoja AI | September 2026

What if one AI model could look at a single abdominal CT scan and help doctors identify not just one disease, but 146 different clinical findings across 18 organs?

That is the idea behind DAMO RADAR, a new medical artificial intelligence model developed by Alibaba’s DAMO Academy and research partners.

The model has attracted attention after research describing it was published in Science, with Alibaba also making the model publicly available as open source.

For a continent where cancer diagnosis, specialist capacity and access to advanced healthcare remain major challenges, the development raises an important question:

Could AI become another layer of diagnostic capacity for African healthcare?

What exactly is DAMO RADAR?

DAMO RADAR is not simply a “cancer detector.”

It is a generalist medical imaging AI model designed to analyse contrast-enhanced abdominal CT scans. Rather than focusing on one cancer or one organ, it examines multiple abdominal organs and can identify a wide range of abnormalities.

The model covers 146 clinical findings across 18 abdominal organs and anatomical structures, including findings associated with malignant tumours.

That distinction matters.

A headline saying that an AI model can “detect 146 diseases” can make the technology sound like a digital doctor. In reality, DAMO RADAR is a diagnostic support system. Its output still needs to be interpreted within the clinical context by healthcare professionals.

The research team describes the system as an expert-level generalist AI for abdominal CT diagnosis.

And the scale behind it is significant.

DAMO RADAR was trained using more than 400,000 contrast-enhanced abdominal CT examinations and approximately 15 million anatomy-aware image-text pairs. Instead of relying entirely on manually labelled images, the researchers used CT scans together with clinical reports to teach the model relationships between anatomy, imaging patterns and medical descriptions.

From one cancer to many conditions

Medical AI has traditionally tended to be highly specialised.

One model might be trained to identify breast cancer. Another might focus on lung nodules. Another might analyse liver lesions.

DAMO RADAR represents a different direction.

Instead of asking:

“Can AI identify this particular cancer?”

the researchers are asking:

“Can one generalist AI system understand a much broader range of abnormalities across an entire anatomical region?”

That is potentially important because real patients do not arrive at hospitals with perfectly labelled diseases.

A CT scan may contain several findings at once.

A patient may have a tumour, inflammation, an anatomical abnormality or another condition that was not necessarily the original reason for the scan.

A generalist model could potentially help clinicians identify findings that might otherwise require additional attention.

The numbers behind the model

In testing involving nearly 40,000 real-world examinations, DAMO RADAR achieved an average AUC of 0.913 across 146 clinical findings.

AUC, or area under the receiver operating characteristic curve, is one measure used to evaluate how well a diagnostic system distinguishes between different clinical states. A score closer to 1 indicates stronger discrimination in that evaluation setting.

The researchers also evaluated the model across external datasets and clinical settings.

In a reader study involving 26 radiologists, the researchers reported that DAMO RADAR performed better than 23 of the participating radiologists on average for the evaluated tasks.

But perhaps the more interesting result was what happened when humans and AI worked together.

According to the reported study results, radiologists using DAMO RADAR experienced approximately a 10% reduction in missed diagnoses and a more than 30% reduction in diagnostic time.

That changes the conversation.

The question is no longer simply whether AI can outperform a human on a benchmark.

It becomes:

Can AI help a human specialist work faster and miss fewer abnormalities?

Why this matters for cancer

Cancer remains one of the world's biggest health challenges.

The World Health Organization estimates that cancer caused nearly 10 million deaths globally in 2024. It also notes that many cancers can be cured when detected early and treated effectively.

In Kenya, the latest GLOBOCAN 2024 estimates cited by the National Cancer Institute of Kenya indicate approximately 35,867 new cancer cases and 22,888 cancer deaths annually.

The five leading cancers by new cases in Kenya were reported as breast, cervical, prostate, oesophageal and colorectal cancer.

These numbers highlight an important point.

Cancer care is not only about discovering new treatments.

It is also about finding disease earlier, interpreting diagnostic information accurately and connecting patients to treatment quickly.

AI cannot solve all of those problems.

But diagnostic AI could potentially become one component of a larger healthcare system.

The African question: What happens when the specialist is far away?

This is where the DAMO RADAR story becomes particularly interesting for Africa.

Healthcare access is not evenly distributed.

A patient in Nairobi may have access to advanced imaging and specialist expertise that is much harder to obtain in a smaller town or rural setting.

The challenge is therefore not simply:

“Does the hospital have a CT scanner?”

It is also:

“Who is available to interpret the scan?”

AI could potentially become an additional layer between medical imaging and specialist interpretation.

Imagine a future workflow where:

1. A patient receives a CT scan.
2. The scan is analysed by an AI system.
3. Potential abnormalities are flagged.
4. A radiologist reviews the findings.
5. The wider clinical team combines the imaging with laboratory results, medical history and other evidence.
6. The patient is referred for appropriate confirmatory testing and treatment.

That is very different from saying:

“AI diagnoses cancer.”

The more realistic vision is:

AI assists healthcare professionals in finding important signals faster.

Why open source matters

One of the most interesting aspects of DAMO RADAR is not only its performance.

It is the fact that Alibaba has released the project publicly.

The official repository includes the model's code, training and inference components, documentation and supporting resources. The project is released under the Apache 2.0 licence.

That opens an interesting door for researchers and developers.

Instead of medical AI remaining entirely inside a private laboratory, researchers around the world can inspect the technology, reproduce experiments, evaluate performance and investigate how it might be adapted.

For Africa, this could be significant.

African researchers could potentially investigate questions such as:

• How does the model perform on African patient populations?
• Does it generalise across hospitals with different imaging equipment?
• How does it perform on data from Kenyan hospitals?
• Can it support hospitals with limited specialist capacity?
• What infrastructure would be required to deploy it safely?
• How should patient privacy and medical data governance be handled?
• Could similar approaches work with other imaging modalities?

These questions cannot be answered simply by looking at Alibaba's benchmark results.

They require local validation.

But AI is not a magic cancer machine

This is perhaps the most important part of the conversation.

DAMO RADAR's results are impressive, but they should not be interpreted as meaning that patients can upload a CT scan to an AI system and receive a definitive cancer diagnosis.

Medical diagnosis is complicated.

A radiological finding may require additional imaging, laboratory tests, clinical assessment or pathology.

For many cancers, histopathology remains essential for confirming what a suspicious lesion actually is.

There are also questions around dataset differences, equipment, demographics, clinical workflows and regulatory approval.

An AI model performing strongly in one research environment does not automatically mean it will perform identically in every hospital in Kenya, Nigeria, Ghana, Tanzania or elsewhere.

That is why independent validation matters.

Could this help Kenya?

Potentially — but the opportunity is bigger than simply importing a model.

Kenya already has national cancer screening and early-diagnosis guidelines, including updated guidelines published in 2024.

The next phase of healthcare AI could involve building systems around those existing healthcare structures.

Imagine AI being used to support:

Early detection → imaging → specialist review → referral → treatment → follow-up

rather than treating AI as an isolated technology.

For example, an AI system could potentially help prioritise scans requiring urgent specialist review.

That could become particularly useful in environments where radiologists are managing large workloads.

But before such systems become part of routine clinical care, questions around validation, regulation, cybersecurity, patient consent, data protection, infrastructure and accountability must be addressed.

The bigger AI trend

DAMO RADAR is also part of a much larger movement.

Alibaba's DAMO Academy has previously worked on AI systems targeting specific medical problems, including cancers involving the liver, pancreas, stomach and colorectal system.

The new model takes a broader approach.

Instead of building an AI system for one disease at a time, researchers are exploring whether a general-purpose medical model can understand multiple conditions and organs simultaneously.

This resembles a broader trend happening across AI.

We have moved from:

One model → one task

towards:

One model → many tasks.

Healthcare may be one of the most consequential areas where this transition happens.

What Pamoja AI thinks

At Pamoja AI, we believe the most interesting question is not:

“Will AI replace doctors?”

It is:

“How can AI expand human capability?”

DAMO RADAR provides an interesting example.

If the reported research results translate into real-world clinical environments, medical AI could potentially help specialists analyse large volumes of medical images more efficiently.

For Africa, that possibility deserves serious research.

The continent does not necessarily need to reproduce Silicon Valley's healthcare infrastructure before benefiting from AI.

It can explore models that fit African realities.

That could mean AI-assisted radiology, telemedicine, clinical decision support, automated patient triage and other tools designed around existing healthcare systems.

But innovation must be accompanied by evidence.

African healthcare should not become a testing ground for unvalidated AI.

The technology should be independently evaluated on local populations, integrated with qualified healthcare professionals and governed by appropriate medical and data-protection standards.

The future may be collaboration, not replacement

The DAMO RADAR story is ultimately bigger than Alibaba.

It represents a direction in which AI is moving from narrow tools toward generalist systems capable of handling multiple medical tasks.

If that trajectory continues, the future radiology department may not be:

Doctor vs AI.

It could be:

Doctor + AI + medical imaging + clinical data + human judgement.

And that distinction matters.

The technology may become faster.

The models may become more capable.

The amount of medical data available to AI will continue to grow.

But the responsibility for patient care will still require people, institutions, evidence and accountability.

For countries such as Kenya, the opportunity is to participate in that future — not simply as consumers of imported AI, but as researchers, developers, healthcare professionals, regulators and innovators building AI that works for African patients.

DAMO RADAR is one more signal that the medical AI race is accelerating.

The next question is whether Africa will only watch it happen — or help shape what comes next.

---