# Artificial Intelligence Moves Deep Into Hospitals - IT조선

*Источник: IT조선*
*Дата: 2026-07-26*
*Язык: en*

**Кратко:** Before a chest X-ray taken at a hospital reaches a physician, artificial intelligence may already have marked areas suspected of containing lung nodules, pneumonia or pneumothorax. When a patient suspected of having a stroke undergoes a computed tomography scan in the emergency room, AI analyzes vascular blockages and blood flow to identify those who require the most urgent treatment.

Before a chest X-ray taken at a hospital reaches a physician, artificial intelligence may already have marked areas suspected of containing lung nodules, pneumonia or pneumothorax.
When a patient suspected of having a stroke undergoes a computed tomography scan in the emergency room, AI analyzes vascular blockages and blood flow to identify those who require the most urgent treatment. In general wards, AI continuously reviews patients’ blood pressure, pulse, respiratory rate and body temperature, alerting medical staff when the risk of cardiac arrest rises.
In this way, AI is becoming deeply embedded in medical care without patients necessarily realizing it.
Until only a few years ago, medical AI was largely regarded as an experimental technology whose performance was tested in hospital research laboratories or pilot programs. It is now becoming a practical clinical tool that supports medical professionals in examination rooms, health screening centers, emergency departments and general wards.
The most widely used medical AI systems in South Korea are image-reading support solutions that analyze chest X-rays, mammograms and other medical images. More recently, their applications have expanded to emergency stroke care and inpatient risk monitoring, extending patients’ exposure to AI from examination rooms to hospital wards.
Patients, however, rarely receive AI-generated findings directly. In most cases, AI presents physicians with risk scores or highlights suspicious areas, while doctors make the final diagnosis and treatment decisions.
From a patient’s perspective, it can therefore be difficult to know whether an X-ray, scan or set of vital signs has been analyzed by an AI system.
X-ray Rooms Become Patients’ Most Frequent Point of Contact With A
The places where patients are most likely to encounter medical AI are examination rooms equipped with X-ray and mammography systems.
Lunit INSIGHT CXR, developed by South Korean medical AI company Lunit, identifies and marks suspected abnormalities in chest X-rays, including lung nodules, pulmonary consolidation and pneumothorax.
The AI does not make a definitive diagnosis. Instead, it helps radiologists conduct an additional review of lesions they might otherwise overlook.
Chest X-rays are widely used not only in health screenings but also in emergency departments, outpatient clinics and inpatient care. They are performed more frequently than many other imaging examinations, and their AI-generated results can be connected relatively easily to existing picture archiving and communication systems, or PACS.
Chest X-ray AI is therefore regarded as one of the medical AI product categories most frequently encountered by patients in South Korea.
Lunit INSIGHT CXR detects multiple types of suspected chest abnormalities and assists medical professionals in interpreting the images.
AI is also being used in breast cancer screening.
Lunit INSIGHT MMG marks areas suspected of containing cancer in mammograms and provides a score indicating the probability of malignancy. Mammograms can be particularly difficult to interpret in women with dense breast tissue because normal tissue and lesions may overlap. AI helps physicians review suspicious areas more closely.
Hospitals including Seoul National University Hospital and Seoul Medical Center have piloted Lunit’s image-analysis solutions through publicly funded medical AI projects.
After initially expanding through major tertiary hospitals and large health screening centers, Lunit is extending the supply of its products to regional medical hubs.
Under a government program supporting AI-based clinical systems at regional accountable care hospitals, Lunit’s chest X-ray and mammography AI systems are being installed at six institutions: Dankook University Hospital, Chungnam National University Sejong Hospital, Ulsan University Hospital, Jeju National University Hospital, Jeonbuk National University Hospital and Chonnam National University Hwasun Hospital.
Medical AI, previously concentrated in large hospitals in Seoul and the surrounding metropolitan region, is now spreading to national university hospitals and regional medical centers.
Patients receiving chest X-rays or mammograms at these hospitals may have their images analyzed by AI as part of the examination process.
AI Measures Lung Nodules and Tracks Their Growth
For chest CT examinations, hospitals are using AI systems such as AVIEW LCS, developed by Coreline Soft.
The solution automatically identifies lung nodules in low-dose chest CT images and measures their size and volume. When previous CT images are available, it can also compare them to determine how much a nodule has grown.
Some products also analyze other abnormalities that can be detected through chest CT scans, including emphysema and coronary artery calcification.
While X-ray AI is particularly useful for rapidly screening images that may contain abnormalities, CT-based AI focuses on quantifying small lung nodules and tracking changes over time.
It can be particularly useful for patients who undergo repeated examinations at intervals of several months or years after lung nodules are detected during lung cancer screening or routine health checkups.
Stroke AI Supports Time-Critical Decisions in Emergency Rooms
The emergency room is one of the hospital settings in which medical AI can have the most direct effect on patient treatment.
A stroke occurs when a blood vessel in the brain becomes blocked or ruptures. In cases of acute ischemic stroke caused by a blockage in a major cerebral artery, the speed at which thrombectomy is performed can determine a patient’s chances of survival and the severity of long-term disability.
JLK-CTP, developed by JLK, analyzes CT perfusion images and helps distinguish between areas of the brain experiencing reduced blood flow and tissue that may still be recoverable.
JLK-LVO detects suspected large-vessel occlusions in CT angiography images.
JLK has established a portfolio of AI products designed to support the full stroke care pathway by analyzing CT scans, magnetic resonance imaging and angiographic images.
Konkuk University Medical Center and Chung-Ang University Hospital have adopted stroke AI solutions, including JLK-CTP, under subscription-based arrangements.
Hospitals can use the AI findings to support decisions on whether thrombectomy is required and which patients should receive priority treatment.
JLK has also said it was selected as a supplier to 11 of the 17 institutions participating in the government’s regional accountable care hospital support program. Products supplied under the initiative include JLK-CTP and JLK-LVO.
The government is investing a total of 14.2 billion won in the program.
Installing stroke AI systems at regional hospitals could help narrow disparities in medical care between major hospitals in the Seoul metropolitan area and hospitals elsewhere in the country.
During nighttime hours or in regional emergency rooms where stroke specialists may be unavailable, AI can help medical staff rapidly identify suspected patients and determine whether they should be transferred to another hospital.
AI Watches Over Patients After They Enter the Ward
Patients remain within the reach of medical AI even after examinations are completed and they are admitted to a hospital ward.
VUNO Med-DeepCARS, developed by VUNO, analyzes four vital signs—blood pressure, pulse, respiratory rate and body temperature—to calculate a patient’s risk of cardiac arrest within the next 24 hours on a scale ranging from zero to 100.
The product is designed to identify patients in general wards whose condition may suddenly deteriorate. Unlike intensive care units, general wards do not usually keep patients continuously connected to multiple monitoring devices.
Patients may not know that the vital signs periodically measured by nurses are being entered into an AI system. On medical staff screens, however, a risk score is displayed for each patient.
When the score rises above a certain threshold, medical professionals can reassess the patient and consider blood tests, imaging examinations or transfer to an intensive care unit.
As of April 2025, DeepCARS had been deployed across more than 48,000 hospital beds in South Korea, including beds at more than 20 tertiary hospitals, according to the company.
While imaging AI is activated whenever an examination is performed, DeepCARS operates repeatedly throughout a patient’s hospital stay.
Research findings have also suggested that the system can improve patient outcomes in real-world clinical settings.
VUNO said an analysis of approximately 160,000 inpatients at Kangdong Sacred Heart Hospital, Shihwa Medical Center and Incheon Naeun Hospital found that cardiac arrests in general wards fell by 21 percent following the introduction of DeepCARS. Among patients with sepsis, cardiac arrests decreased by 29 percent.
AI Supports Doctors Rather Than Replacing Them
Most medical AI experienced by patients in hospitals is used for clinical decision support rather than autonomous medical care.
Even when chest X-ray AI highlights an area suspected of containing pneumonia, the final interpretation is made by a radiologist.
When an AI system reports a high cardiac-arrest risk score, medical professionals must still examine the patient’s consciousness, symptoms and test results.
Similarly, even when stroke AI identifies a blood-flow abnormality, a specialist decides whether thrombectomy or drug treatment should be performed.
Medical AI companies emphasize the accuracy of their products, but hospitals consider more than accuracy when deciding whether to adopt them.
Other important factors include false-positive and false-negative rates, analysis speed, integration with PACS and electronic medical records, and whether the system increases the workload of medical staff.
When AI classifies too many patients as high-risk, medical professionals may become fatigued by excessive alerts. Conversely, if AI incorrectly classifies a high-risk patient as normal, treatment could be delayed.
Hospitals must also consider that AI performance may vary depending on patient characteristics and the clinical environment at each institution.
Medical AI Expands Beyond Seoul’s Major Hospitals
The adoption of medical AI in South Korea has historically been concentrated in major hospitals in Seoul and the surrounding metropolitan area.
Hospitals with large volumes of medical imaging data, well-developed information systems and sufficient specialist personnel have found it easier to validate AI performance and integrate the technology into existing systems.
More recently, government-backed projects have introduced chest X-ray, mammography, lung CT and stroke AI systems into regional accountable care hospitals.
The aim is to help hospitals in areas with shortages of medical professionals identify emergency patients more quickly and reduce the burden of interpreting medical images.
Government support may provide the initial momentum needed for hospitals to adopt AI. Whether hospitals continue paying for the systems after the publicly funded projects end, however, is a separate question.
For AI systems to remain in use, developers and hospitals must demonstrate that they shorten interpretation times, accelerate emergency treatment and reduce cardiac arrests and mortality among inpatients.
National Health Insurance reimbursement is another critical variable.
Even when an AI medical device receives approval from the Ministry of Food and Drug Safety, hospitals must bear the cost if there is no separate reimbursement for its use.
Products for which hospitals can charge patients directly as non-reimbursed services may spread more rapidly, but they also raise concerns about patients’ financial burden and whether adequate information is being provided.
“Medical AI has already moved beyond the research and development stage and is now being used in the actual process of testing and caring for hospitalized patients,” an industry official said.
“In the future, competitiveness will not be determined by how many hospitals have installed a product, but by whether it can improve patient outcomes while becoming naturally integrated into the workflows of medical professionals.”
simalo@chosunbiz.com

[Оригинал](https://it.chosun.com/news/articleView.html?idxno=2023092166614)