“It’s not luck”: How stage 1 lung cancer was caught on my first X-ray over 70
It’s the afternoon of May 5 at Seongnam City Medical Center in Gyeonggi Province. Dr. Hwa-Young Song, a radiologist, reads the chest X-ray of a patient who had a checkup earlier in the day. After checking the electronic medical record (EMR) on the leftmost monitor, Dr. Song heads to the other two monitors. They look similar at first glance, but the X-ray image on the left monitor clearly shows a red lesion in the rib cage just below the right collarbone. If you look closely, you can also see the number 23%. This is how Seongnam Medical Center utilizes Lunit Insight CXR, an artificial intelligence (AI) image analysis solution for chest X-rays introduced by Lunit (328130) earlier this year, in the clinic.
◇ 10 abnormal findings in chest X-ray images…up to 99 percent accuracy
Lunit Insight CXR uses artificial intelligence (AI) to analyze chest X-ray images in real time to help doctors read them. It identifies the 10 most common abnormalities found on chest X-rays, including lung nodules, atelectasis (collapsed lungs), calcification (abnormal deposits of calcium in lung tissue), cardiac enlargement, pulmonary cirrhosis, fibrosis, mediastinal enlargement, pleural effusion, and pneumothorax with 97-99% accuracy.
It can also be used to screen for tuberculosis. The AI determines the likelihood of a lesion being present and presents it as a probability value and changes the color of the display according to the threshold value set by the reader. For this patient, the AI solution means “there is a 23% probability of one suspicious lesion in the lung area near right ribs 3 and 4”. “It was difficult to identify the lesion based on the X-ray image because it overlapped with the right rib,” says Song. “When we compared it to the CT (computed tomography) scan results without seeing the readout, there was indeed a lung lesion. Fortunately, it’s not malignant, but it’s a case that needs to be followed up.”
AI-assisted image reading is literally a diagnostic aid. There was a time when the idea of AI replacing doctors in the near future seemed like something out of a science fiction movie. Twelve years ago, IBM’s supercomputer Watson defeated a human champion on the U.S. TV quiz show Jeopardy and began his residency at Memorial Sloan Kettering Cancer Center, where he challenged himself to treat lung cancer patients. Later, Watson for Oncology, trained at MD Anderson Cancer Center, appeared and received a lot of attention in Korea as it was introduced to leading medical institutions including Gacheon Daegil Hospital. However, Watson revealed its fatal limitations and disappeared from the medical field after less than five years. Experts say that even in radiology, one of the most active areas of medical AI use, it is not practical to replace a skilled specialist. This is because a chest X-ray cannot be used to diagnose many of the conditions that can occur in the lungs. The more advanced the technology, the more important the role of an experienced specialist in the field.
“When the probability of a suspected lesion is only 10 to 20 percent, it is often difficult for even radiologists to make a judgment,” said Dr. Song. “AI solutions that quantify and represent lesions that are difficult to filter out with the naked eye are highly effective in improving accuracy in detecting conditions that are on the borderline between normal and abnormal.”
◇ With the help of AI solutions…focus on reading hidden ‘suspicious lesions’
Seongnam Medical Center, which has been open for about three years, has many elderly patients who have never had a chest X-ray before. Due to the nature of public hospitals, there is a high demand for medical examinations. This is the biggest difference from private university hospitals, where most patients are diagnosed with lung cancer or other diseases at primary or secondary medical institutions.
A patient takes an X-ray at Seongnam City Medical Center in Gyeonggi-do on May 5. Reporter Cho Hyung-joo
“There were many times when we were overwhelmed by the idea of providing an accurate reading based on only one chest X-ray image in a situation where it was difficult to compare it with the previous image,” said Song. “Since introducing the AI solution, we have been able to focus more on reading suspicious lesions with more confidence in the findings of normal people.” In a recent case, Seongnam Medical Center detected early-stage (stage 1)메이저놀이터 lung cancer in a woman in her 70s who had a chest X-ray for the first time in her life. “Based on the AI result of ‘16% probability of suspected lesion,’ we requested an additional CT (computed tomography) scan, and lung cancer was found,” said Song. “When we heard that the patient underwent surgery and was discharged safely, we realized the usefulness of AI solutions once again.”
◇ AI solutions are useful, but not yet recognized…”It will have a great complementary effect on public healthcare infrastructure”
AI image reading is being recognized for its usefulness in the field, but there is no separate fee for it. This makes it even more difficult for public hospitals with tight budgets to adopt it, let alone private hospitals. Seongnam Medical Center, which was contemplating adoption while using the demo version, benefited from the Functional Enhancement Project for Regional Public Hospitals, which the Ministry of Health and Welfare launched at the end of last year with the aim of strengthening public healthcare competitiveness by introducing advanced medical devices such as AI. Ten local medical centers nationwide, including Seongnam Medical Center, are in the process of introducing Lunit Insight MMG, which helps diagnose breast cancer, along with Lunit Insight CXR.
Dr. Song Hwa-young, a radiologist at Seongnam City Medical Center in Gyeonggi-do, Gyeonggi-do, introduces the experience of image reading using AI solutions on May 5. Hyungju Choi Reporter
Digital healthcare services, including medical AI, are gaining attention as an alternative to improve the quality of public healthcare services and bridge regional disparities that have been exposed as vulnerable during the COVID-19 pandemic. Lunit has proactively proposed a model to set up AI screening centers at public health centers nationwide that combine AI medical technology and cloud platforms. In this model, AI technology analyzes chest images taken by X-ray machines at each health center in real time to help medical staff determine the presence or absence of major lung diseases. If the screening results show any abnormalities, they can be referred to a healthcare program or a community hospital for treatment. The Australian government also introduced Lunit Insight MMG to its national breast cancer screening program earlier this year.