Coreline Soft and INFINITT Healthcare are integrating chest CT AI into hospital PACS environments in the US
Series Overview
September 4 | HIRA Takes Its First Step in AI Claims Review as Medical Imaging AI Enters National Infrastructure
September 11 | The Competition to Build Integrated Medical Imaging AI Platforms
September 18 | Lung Cancer Screening AI and Post-Screening Patient Management
September 25 | How Breast Cancer and Stroke AI Improve Patient Outcomes
October 2 | DICOM Standardization and the Expansion of Medical Imaging AI
October 9 | Performance Monitoring and Security for Medical Imaging AI
October 16 | Linking Medical Imaging AI Reimbursement to Clinical Outcomes
October 23 | A Global Market Strategy for Korean Medical Imaging AI
Medical imaging artificial intelligence is entering a new phase of competition. Hospitals increasingly need an operating layer capable of connecting multiple AI applications to their existing clinical systems, rather than separate software for every disease.
At the center of this shift is the picture archiving and communication system, or PACS, which stores, retrieves and displays medical images. The radiology information system, or RIS, manages examination schedules, reporting and radiology workflows, while the electronic medical record, or EMR, connects imaging findings with broader patient information and treatment decisions.
An AI platform must work across these systems without forcing radiologists and other clinicians to move between multiple screens. It must identify the appropriate examination, send the images to the relevant algorithm, return the results to the reading worklist and notify the right clinical team when urgent action may be needed.
The commercial contest is therefore expanding beyond algorithm accuracy. Platform providers are competing over integration with hospital systems, access to radiology workflows, the management of multiple algorithms and the data needed to evaluate performance after deployment.
This development follows the transition examined in the first article of this series. On Sept. 1, South Korea’s Health Insurance Review and Assessment Service began using AI-generated interpretations of knee images in the review of claims for osteoarthritis treatment. The move showed that medical imaging AI was beginning to operate within a national health insurance process.
The second stage concerns how hospitals will manage a growing number of AI applications. If every product requires its own server, interface, viewer, alert system and maintenance agreement, the burden of integration may outweigh the time saved by automation. A platform capable of coordinating different products could become the main gateway through which imaging AI reaches clinical practice.
PACS companies and independent platforms compete to control the AI gateway
PACS vendors have a natural advantage in this market because their systems already sit inside the radiologist’s daily working environment. Images, prior examinations, reading lists and reports pass through the PACS, giving its operator a central position from which to incorporate AI.
Sweden-based Sectra has developed an AI service within its enterprise imaging environment that allows hospitals to access applications from Sectra and outside developers. Its approach is to place AI selection, deployment and monitoring within the existing imaging infrastructure so that hospitals can add applications without building a separate connection for each product.
An examination can be routed to an appropriate algorithm according to factors such as imaging modality, body region and clinical purpose. The resulting information can then be returned to the radiology environment, where it may support case prioritisation, measurements or image interpretation.
Independent AI platform companies are pursuing a different route. Israel-based Aidoc positions its aiOS platform as an operating layer that can analyse imaging examinations and deliver notifications to multidisciplinary care teams. The company’s portfolio focuses heavily on acute conditions, where the value of AI depends on whether a suspected finding reaches the clinician responsible for treatment in time.
For example, identifying a possible pulmonary embolism or intracranial haemorrhage on a CT scan has limited operational value if the result remains inside a separate AI dashboard. It becomes clinically relevant when it is reflected in the reading queue and communicated through a hospital-approved pathway to emergency physicians, neurologists, surgeons or other specialists.
The two approaches reveal a broader contest. PACS companies can extend an installed imaging system into an AI distribution platform, while independent platform providers can connect several algorithms and clinical departments across different PACS environments. Medical equipment manufacturers and large healthcare technology companies are also incorporating AI orchestration into their imaging and informatics portfolios.
Hospitals must consequently decide who controls the routing of their images, the presentation of AI results and the records generated during clinical use. The platform operator may also influence which third-party applications can be installed, how easily they can be replaced and whether performance data can be exported for independent analysis.
A multitechnology statement led by Professor Adrian Brady, former president of the European Society of Radiology, warned that a proliferation of complex interfaces could reduce both the utility and adoption of imaging AI. The statement, jointly published by five radiology organisations, called for standards-based integration and stressed that AI results should be available within clinicians’ normal working environments.
That warning has become more relevant as hospitals move from one or two pilot applications to larger portfolios. A hospital may use separate tools for lung nodules, breast cancer, fractures, stroke, pulmonary embolism and image-quality assessment. Without a common operating structure, clinicians may face duplicated alerts, inconsistent result formats and uncertainty over which system is responsible when an analysis fails.
A platform should therefore be assessed on more than the number of algorithms in its catalogue. Important criteria include compatibility with DICOM and healthcare data exchange standards, integration with PACS, RIS and EMR systems, support for different vendors, cybersecurity controls, audit logs, version management and procedures for service interruption.
Hospitals should also be able to determine where patient data are processed and retained, who can access them and how information is handled when a contract ends. The ability to add or replace an algorithm without rebuilding the entire workflow will be particularly important as the market consolidates and products are updated or withdrawn.
Korean companies need to connect specialised AI with medical information systems
South Korea has developed globally competitive algorithms in chest imaging, cancer detection and neurological disease. The next challenge is to ensure that these products can operate within the information systems used by hospitals in Korea and overseas.
Coreline Soft provides a useful example. Its AVIEW platform analyses low-dose chest CT scans for conditions including lung nodules, emphysema and coronary artery calcification. This approach allows several findings to be assessed from a single examination, an important consideration for lung cancer screening programmes in which CT images can contain information relevant to more than one disease.
Coreline Soft has also worked with INFINITT Healthcare to integrate chest CT analysis into PACS environments used by medical institutions in the United States. INFINITT’s role as a medical imaging and information-system provider offers a route through which a specialised Korean algorithm can enter a hospital’s established reading workflow.
The significance of this relationship lies in its structure. Coreline Soft supplies disease-specific imaging analysis, while INFINITT provides the system through which images and results are managed in clinical practice. Their cooperation illustrates how Korean AI developers and healthcare information-system companies could jointly approach overseas hospitals.
Other Korean companies have established specialised positions in the imaging AI market. Lunit has developed products for chest radiography and mammography, while VUNO has commercialised AI applications for chest radiographs and other medical images. JLK and Neurophet have concentrated on neurological imaging and quantitative brain analysis.
Detailed comparisons of these companies’ international partnerships, regulatory approvals, hospital deployments and platform strategies will be addressed in the next article. At this stage, their collective challenge is to connect specialised algorithms with hospital workflows without increasing the burden on clinicians or creating additional data silos.
Korean PACS and medical information-system providers could serve as distribution and integration partners for these applications. Conversely, AI developers could strengthen the competitiveness of Korean hospital systems by adding clinically validated functions for specific diseases.
This model will require clear contractual arrangements. Hospitals should specify the cost of initial integration and continuing maintenance, permitted uses of clinical data, responsibility for software updates, response times during failures and access to performance records. They should also establish what happens to stored results and audit logs when a product is replaced or a service agreement expires.
Government procurement and hospital evaluation systems can encourage this development by requiring interoperability and measurable clinical outcomes. Products that communicate through recognised standards and allow hospitals to retain control of their operational data will be easier to manage across multiple institutions.
South Korea has several structural advantages: widespread use of digital medical imaging, a national health insurance system, nationwide screening programmes and domestic companies active in both medical AI and hospital information technology. These assets could support a national imaging AI infrastructure if they are connected through common technical and governance rules.
The platform competition will ultimately be decided inside the clinical workflow. Hospitals will favour systems that deliver the right result to the right professional at the right time, allow algorithms to be changed without disrupting care and provide evidence that their use improves efficiency or patient outcomes.
PACS has consequently become more than an archive and image viewer. It is emerging as the control point through which imaging data are routed to AI, results are returned to clinicians and clinical performance is documented. The companies that can manage this process reliably may shape how medical imaging AI is purchased, deployed and evaluated.
The third article will examine the platform and overseas market strategies of leading Korean and international medical imaging AI companies, including how specialised algorithm developers are building partnerships with PACS vendors, hospital systems and global distributors.
