InferVision's products have reached roughly 40 countries and over 2,000 institutions globally, with a significant share of deployments in public health screening in developing regions, China Economic Net (CEN) reported on Saturday.

Chinese solutions bundle software, hardware, and training together, and these packages are comparatively much more cost-effective," said Dr. Muhammad Shahbaz, president of the China-Pakistan Medical Association, who has been involved in advancing multiple Chinese medical technology partnerships in Pakistan.

In Pakistan, integrated with a battery-operated ultra-portable X-ray unit, the product helped screen over 8,000 individuals among miners and people in remote areas in 2023.

The picture is not limited to TB. In Pakistan, a Chinese company, Landing Med, established an AI medical diagnosis laboratory with Akbar Niazi Teaching Hospital that began operating in 2023, providing free cervical cancer screening to 10,000 women in its first phase.

The lab uses AI-assisted cytology to analyse samples that would otherwise require a specialist pathologist — a resource Pakistan's public health system has in limited supply outside major cities.

In a remote mining community in northeast Nigeria, tuberculosis screening no longer begins in a hospital. It begins outdoors, where workers line up beside an ultra-portable X-ray machine and health staff read the images with the help of artificial intelligence.

The device was supplied by a Chinese medical AI company.
This is one instance of a wider pattern. Chinese medical AI is becoming visible in parts of the global south.

It is screening: sorting large populations for probable disease, flagging who needs further testing, operating at volumes no human workforce can match on the budgets that public-health programmes in developing countries actually have.

Why screening comes first: Screening is one of the few areas of medical AI that can scale relatively quickly in lower-resource settings because the task is narrow, repeatable and high-volume. Unlike more complex clinical applications, screening does not require AI to make a final diagnosis.

It flags likely abnormalities, sorts patients by risk and identifies who needs confirmatory testing. That makes it easier to insert into existing workflows — and it is where Chinese companies have found their opening.

The underlying need is acute. There is severe brain drain in low- and middle-income countries: trained specialists leave, or too few graduate each year.

AI systems can be trained on the expertise of the best medical specialists and then perform on par with them in specific tasks — effectively bringing expert judgment to places those experts will not go.

The numbers bear this out. WHO estimates that 10.7 million people fell ill with TB in 2024, but only 8.3 million were newly diagnosed and treated. 87% of new cases were concentrated in 30 countries, including Pakistan and Nigeria.

Yet radiologist density in low-income countries stands at roughly 1 per million people, against 93 per million in high-income countries. Films are often read by general clinicians, not specialists.

Echoing that logic, some deputies to China’s National People’s Congress have argued in recent media reports that AI’s most immediate value in healthcare lies in supporting grassroots doctors and expanding access in under-served areas.

"In LMICs, we see the same pattern, with radiology being the number one application of medical AI," said Dr. Zhi Zhen Qin, founder at the Global Impact Solutions, and former digital health specialist at the UNOPS Stop TB Partnership.

In June 2025, WHO approved six computer-aided detection (CAD) products of TB on chest X-ray, including one from China’s InferVision, alongside products from the Netherlands and South Korea etc.

Chinese companies in the field: Chinese medical AI is becoming visible across parts of the global south, with its earliest overseas gains in tuberculosis screening. In 2021, InferVision's product entered the Stop TB Partnership's Global Drug Facility catalogue — the principal procurement mechanism for donor-funded TB programmes worldwide.

The entry followed an open tender process: InferVision submitted a bid, met the quality-assurance and technical criteria, and was added to an approved list from which recipient countries make their own purchasing decisions.

The GDF package included a perpetual licence for unlimited reads, online and offline image processing, required hardware, installation, training, and one year of support. The competing Dutch product, CAD4TB, was listed at $16,700; InferVision entered at $5,000.

That price difference matters, but so does the pricing structure. Many AI products use pay-per-scan models that suit private hospitals charging patients.

They do not suit large donor-funded programmes, where budgets are fixed in advance for one, three or six years. A perpetual licence removes that friction.

Scaling medical AI in the global south: A vast majority of medical AI remains at the pilot stage in low- and middle-income countries, and the barriers to wider roll-out are often practical rather than technical.

"Accuracy is good enough across many of these products. The differences in accuracy among products are often very minimal. So it becomes more about other factors," said Dr. Qin. Price and certification status compliant with global QA criteria are the two she returns to most.

For the next stage, Dr. Qin sees greater promise in more decentralised tools. Smartphones can already support AI-assisted cervical cancer screening; handheld ultrasound devices and digital stethoscopes costing a few hundred dollars require no radiation and run on batteries.

"If the endpoint is a smartphone, which many people in LMICs can access, then that is already enough. You don't have to buy another dedicated medical device."

Credit: Independent News Pakistan (INP) — Pak-China