Taiwan broadens eligibility while Australia focuses on participation and regional access
Germany formalises AI-assisted reading as domestic and global companies compete on platforms and follow-up management
Series Overview
- September 4 | Medical Imaging AI Enters National Infrastructure as HIRA Launches AI-Assisted Claims Review
- September 11 | The Medical Imaging AI Platform Race
- September 18 | Lung Cancer Screening AI and National Follow-Up Care
- September 25 | Breast AI Cuts Interval Cancers and Speeds Diagnosis
- October 2 | Stroke AI Connects Image Analysis to Treatment and Patient Recovery
- October 9 | DICOM Standardization and Performance and Security Management for Medical Imaging AI
- October 16 | Linking Medical Imaging AI Reimbursement to Clinical Outcomes
- October 23 | Global Market Strategies for Korean Medical Imaging AI
The success of lung cancer screening depends on years of follow-up after the initial low-dose chest computed tomography scan. National programmes must identify eligible people, encourage participation and connect nodule findings to repeat examinations at intervals of three months, six months, one year or two years, depending on clinical risk.
Medical imaging artificial intelligence is also expanding from nodule detection to prior-image comparison, measurement of changes in nodule size and volume, follow-up scheduling and structured reporting. As participation grows, health systems must manage new examinations alongside an accumulating volume of repeat scans. Reading accuracy, clinician workload, missed examinations, image exchange between institutions and long-term data management have become central performance measures.
Medical imaging artificial intelligence is also expanding from nodule detection to prior-image comparison, measurement of changes in nodule size and volume, follow-up scheduling and structured reporting. As participation grows, health systems must manage new examinations alongside an accumulating volume of repeat scans. Reading accuracy, clinician workload, missed examinations, image exchange between institutions and long-term data management have become central performance measures.
Medical imaging AI companies are pursuing different strategies around integrated platforms and disease-specific applications. Israel-based Aidoc connects emergency imaging analysis with care coordination across hospitals, while Sweden’s Sectra uses its picture archiving and communication system (PACS) to integrate algorithms from multiple suppliers. Korean companies Lunit, Coreline Soft and VUNO are applying disease-specific technologies to overseas screening and clinical workflows. INFINITT Healthcare is building integrated environments that connect AI with hospital operations through PACS and imaging information systems.
A company’s market position depends on the accuracy of its algorithms and their ability to operate reliably within national and hospital systems. Patient selection, scanning, interpretation, additional testing, treatment referral and long-term follow-up are all part of that assessment.
Taiwan Expands Eligibility, Australia Improves Access and Germany Formalises AI-Assisted Reading
South Korea launched its National Lung Cancer Screening Program in 2019, making it one of the earlier adopters in Asia. The programme offers low-dose chest CT every two years to high-risk people aged 54 to 74 with a smoking history of at least 30 pack-years. A pack-year represents smoking one pack of cigarettes a day for one year. Korea’s national health insurance system, cancer-screening data and widespread hospital PACS infrastructure provide a foundation for applying AI to follow-up management.
Other countries are revising both eligibility and programme operations. Taiwan incorporated low-dose CT lung cancer screening into its national cancer programme in 2022 and expanded eligibility in 2025. The smoking threshold now covers people aged 50 to 74 with at least 20 pack-years. Non-smokers may also qualify if a parent, child or sibling has had lung cancer, with eligibility beginning at 40 for women and 45 for men and continuing through age 74.
Taiwan’s policy reflects the incidence of lung cancer among women and non-smokers in East Asia. Korea would need to examine false positives, overdiagnosis, radiation exposure and cost-effectiveness before adopting similar criteria. It should also assess whether age and smoking history alone identify enough high-risk people and study risk-based models that include family history and environmental exposure.
Australia launched its National Lung Cancer Screening Program in July 2025. It covers high-risk people aged 50 to 70 who have smoked at least 30 pack-years and either continue to smoke or quit within the previous 10 years. Nearly 100,000 people participated during its first year, when the programme detected more than 230 primary lung cancers.
Australia has focused on participation and regional access. The programme combines national public information efforts with mobile CT services for Aboriginal and Torres Strait Islander communities and people living in rural and remote areas. It also manages the next round of screening for people with normal results and additional examinations for those whose findings require closer assessment.
Australian Health Minister Mark Butler said many lung cancers are diagnosed at an advanced stage because symptoms appear late. “If it is found early, treatment options are greater and more likely to be effective,” he said when announcing the programme’s first-year results.
Germany began annual low-dose CT lung cancer screening under statutory health insurance in April 2026. The programme covers current and former smokers aged 50 to 75 who meet defined requirements for duration and cumulative tobacco exposure. About two million people could qualify, although nationwide implementation requires trained clinicians and authorised facilities.
Germany has incorporated computer-aided detection into the reading process. A radiologist first reviews a scan without the software and then checks it again using computer-assisted analysis. A second specialist independently follows the same procedure when a case requires follow-up or more detailed assessment.
The software must support nodule detection and volumetric measurement, volume-doubling-time calculations, prior-image comparison and structured documentation. The rules also cover minimum annual reading volumes, multicentre quality control and access to previous examinations. Germany has therefore made computer-assisted analysis part of the system used to maintain consistent screening quality.
Bernhard van Treeck, a member of Germany’s Federal Joint Committee (G-BA), said “a little more patience” would be required before the programme could be delivered at the same level nationwide. Full implementation requires clinician training, facility accreditation, reading infrastructure and quality-control procedures.
Japan Advances LDCT Screening and SaMD Review
Japan has one of the world’s highest levels of computed tomography (CT) equipment per capita. Lung cancer screening has traditionally relied on chest X-rays, but the country is developing the use of low-dose chest CT (LDCT) for high-risk smokers. Its existing CT infrastructure provides a foundation for earlier lung cancer detection and long-term follow-up.
Japan’s National Cancer Center raised its recommendation level for LDCT lung cancer screening among high-risk smokers in 2025. Five municipalities are conducting LDCT screening pilot programmes in 2026. Evidence on eligibility, participation, false positives, additional testing and cost-effectiveness could inform the design of future screening policy.
Expanding LDCT screening will increase both initial and follow-up examinations. Clinicians must compare the size and volume of pulmonary nodules with previous scans and arrange further examinations according to risk. Medical imaging AI can support this process through automated nodule detection, volumetric measurement, prior-scan comparison and structured reporting.
Japan is also developing its regulatory framework for software as a medical device (SaMD). The Pharmaceuticals and Medical Devices Agency (PMDA) has operated a permanent priority-review pathway for SaMD since 2026 and is refining its approach to devices that use real-world data and generative AI. The policy challenge is to shorten review times while maintaining oversight of software updates and real-world performance.
The market-entry route for medical devices in Japan depends on the product’s function and risk classification. Some products covered by established certification standards can obtain third-party certification from a registered certification body. Devices with novel functions or higher risks undergo a PMDA review of safety, effectiveness, risk management and quality-control evidence before approval by the Ministry of Health, Labour and Welfare.
Coreline Soft said its computer-aided detection (CAD) product for pulmonary nodules recently received ministry approval following a PMDA review. The case applies the company’s experience in LDCT analysis and pulmonary nodule follow-up from Korea’s national lung cancer screening programme to Japanese regulatory requirements. Its commercial progress in Japan can be assessed through the number of installed and active clinical sites, scan volumes, reading times and follow-up completion rates.
Regulatory approval must be followed by local distribution, hospital-system integration, Japanese-language reading environments, clinician training, technical support and reimbursement. Radiologists need to review AI findings alongside previous scans within their existing picture archiving and communication systems (PACS) if the technology is to support repeat screening. Korean companies’ competitiveness in Japan will depend on their ability to maintain these operations after approval.
Korea now needs to manage participant identification, screening uptake, reading quality, post-positive referral, follow-up compliance and interval cancers within one performance system. AI can support interpretation and follow-up, while clinicians and public institutions retain operational responsibility and final clinical judgement.
Screening Operations Determine Global Competitiveness
Overseas platform companies are expanding businesses that integrate and manage AI across hospitals. Aidoc uses emergency CT triage technology as the basis for a platform that runs multiple algorithms, delivers alerts to clinical teams and tracks subsequent action. Its care-coordination functions could also support screening programmes by helping prevent patients with detected nodules from being lost to follow-up.
Sectra connects external AI through its PACS and enterprise imaging system. Hospitals can use algorithms from different suppliers and review the results within existing worklists and reading environments. This approach allows hospitals to retain control over algorithm selection and imaging data. Aidoc concentrates on care coordination and follow-up action, while Sectra focuses on operating multiple algorithms within the hospital’s imaging environment.
Both companies have expanded from lesion detection into AI integration, deployment and performance management. Running separate servers, screens and alert systems for each product increases cost and administrative burden. Platform providers manage algorithm routing, access controls, incident response, version changes and performance records within a shared infrastructure.
Korean companies are also entering international screening and hospital-operations markets with disease-specific technologies and overseas clinical experience. Lunit supplies chest X-ray and mammography AI to hospitals, imaging-equipment companies and screening programmes. Chest X-ray is widely available and relatively inexpensive, supporting population-level triage for tuberculosis and other lung diseases. High-risk lung cancer assessment and nodule follow-up still require low-dose CT and a connected clinical pathway.
Coreline Soft is using its experience in Korea’s national lung cancer screening programme to expand overseas. Its low-dose CT software analyses pulmonary nodules, emphysema and coronary artery calcification and tracks changes in nodules identified on previous scans. According to company data, Coreline Soft had supplied more than 60 of an estimated 200 AI-ready hospitals in Taiwan as of the first half of 2026. Deployments include National Taiwan University Hospital and Chang Gung Memorial Hospital.
In Australia, the company estimates that its number of AI analyses during the same period was equivalent to about 40% of the national programme’s examination volume. The figure is not an official market share because national participant numbers and AI analysis volumes use different counting methods. In Germany, Coreline Soft said it secured 77 new hospital contracts by July 2026, including Charité – Universitätsmedizin Berlin among the publicly identified customers.
These deployments show how experience gained through Korea’s national screening programme is being applied in different health systems. The company is supplying repeat-screening and nodule-tracking technologies to major Taiwanese hospitals, Australia’s national programme and German healthcare providers. Continued reporting of active deployment scope, longitudinal screening data, reading time, review rates and follow-up compliance would provide stronger evidence of clinical and operational performance.
A 2026 study in European Radiology used Coreline Soft software to analyse data from the UK Lung Cancer Screening Trial. Researchers compared baseline and three-month follow-up low-dose CT scans from 361 participants. The system automatically matched 83.5% of 339 nodules that persisted across both examinations. The matching rate was 91.8% for patients with one nodule and 72.8% for those with more than five.
The study showed that AI can identify the same nodule across repeated examinations and connect it with the prior scan. However, the sample and follow-up period were limited, and the study did not evaluate national programme performance or patient outcomes. Coreline Soft provided the research software, and a company representative was among the co-authors. Radiologists must still verify the automated matches.
INFINITT Healthcare is developing hospital AI integration based on its PACS and imaging information infrastructure. Its experience connecting different hospital systems and imaging equipment allows it to incorporate external AI into existing reading workflows. In global markets, its competitiveness will depend on an open architecture that supports both its own products and third-party algorithms, standardised integration and overseas customer support.
VUNO has expanded from imaging AI for chest CT and X-ray into biosignal-based risk prediction. Its lung cancer screening capabilities should be assessed across nodule detection, prior-image comparison, structured reporting, multicentre reading and follow-up management. Technology that analyses imaging and inpatient biosignals could also support care after diagnosis, although clinical effectiveness requires product-specific studies and real-world data.
India’s Qure.ai and China’s Infervision are expanding through population screening and public healthcare markets. Qure.ai supplies chest X-ray tools for tuberculosis and lung-disease triage, alongside CT applications for pulmonary nodules and neurological emergencies. Infervision has expanded chest CT and lung-nodule analysis from its base in China’s large imaging market. Public data remain insufficient to compare their use volumes and patient outcomes across countries on a consistent basis.
Multicentre Real-World Evidence Will Determine Overseas Competitiveness
The number of medical imaging AI products and regulatory clearances has increased, but evidence of clinical impact remains limited. A 2025 European Radiology review found that the share of CE-marked medical imaging AI products supported by peer-reviewed research rose from 36% in 2020 to 66% in 2023. Studies examining clinical decision-making, patient outcomes and socioeconomic effects accounted for about 24% of the evidence base. Prospective and multinational studies also remained limited.
The researchers found that studies of technical performance had increased while independent evidence of clinical effects and patient outcomes remained insufficient. Hospital deployments and AI analysis volumes show market reach. Mortality, stage at diagnosis, follow-up compliance, time to treatment and healthcare costs measure clinical and economic performance. Overseas expansion should be assessed using both sets of indicators.
Korean companies need to support their overseas deployment figures with multicentre real-world data and long-term follow-up results. They can apply experience from Korea’s national programme in overseas healthcare systems and bring proven integration and quality-control procedures back into the domestic screening programme. Such evidence can strengthen product credibility and support purchasing decisions by hospitals and governments.
Governments and healthcare providers should include prior-image transfer and storage, patient follow-up, product updates, performance monitoring and incident-response times in AI procurement contracts. They should also define data-handling and algorithm-replacement conditions after a contract ends. Standardised data structures and interoperability would allow national screening to continue when a vendor or product changes.
Lung cancer screening relies on previous scans throughout subsequent testing and care. A CT scan taken today becomes the reference for an examination months or years later, while changes in a detected nodule help determine whether further testing or treatment is required. AI used in national screening must reliably support lesion detection, prior-image comparison, follow-up scheduling, clinician verification and subsequent care.
Korea has a national lung cancer screening programme, nationwide health insurance, extensive PACS adoption and domestic medical imaging AI companies. Connecting participant management, interpretation, repeat examinations, treatment referral and outcome evaluation would allow Korean companies to export both individual algorithms and national screening operations technology. In the lung cancer screening AI market, the ability to manage patients after the first scan will determine long-term competitiveness.
Part 4 will examine how medical imaging AI for breast cancer and stroke affects patient outcomes and clinical decision-making.
