Originally published on the Health & Medical Network

Lung cancer is often referred to as the “new national disease.” According to the latest cancer statistics report from the Health Promotion Administration of the Ministry of Health and Welfare, nearly 70% of lung cancer cases are diagnosed at a late stage, with an average two-year survival rate of only about 15%. Early screening is the key to improving prognosis.

Professor  Cheng-Yu Chen, Vice President of Taipei Medical University and a board member of the Radiological Society of the Republic of China, stated that with continuous advancements in technology, Taipei Medical University Hospital has developed the “Lung Cancer Clinical AI Decision Support System.” By integrating AI technology, the system enables faster and more comprehensive patient assessments for diagnosis, treatment, and prognosis decision-making. This not only enhances diagnostic accuracy but also enables precise treatment, ultimately improving survival rates.

70% of Patients Are Diagnosed at a Late Stage, Highlighting the Importance of Early Screening

Medical experts agree that lung cancer ranks third in incidence among the top ten cancers but has the highest five-year mortality rate. Since early-stage lung cancer presents no symptoms, approximately 70% of patients in Taiwan are already at an advanced stage when first diagnosed.

Vice President Cheng-Yu Chen explained that late-stage lung cancer means surgery is no longer an option—only early-stage cases can be treated surgically. Performing surgery at a late stage does not improve survival rates, making additional therapies necessary, such as radiation therapy, chemotherapy, immunotherapy, and targeted therapy. If these treatments prove ineffective, patients may need to consider clinical trials to explore new drug options.

Vice President Cheng-Yu Chen and His Research Team. From left: CIO Tzu-Hao Chang, CDO Min-Huei Hsu, Professor Chi-Long Chen, Vice President Cheng-Yu Chen, Lecturer Nguyen Quoc Khanh Le, and Director Shih-Hsin Hsiao

Limitations and Challenges of Traditional Chest X-rays and CT Scans

Vice President Cheng-Yu Chen stated, “Early-stage lung cancer may present no symptoms, which makes diagnosis challenging.” During routine health check-ups, most people undergo chest X-rays. However, X-rays have a low detection rate for early-stage lung cancer, particularly for ground-glass opacities, which are often invisible on X-ray images.

Additionally, parts of the lungs may be obscured by the heart and ribs, creating blind spots that make it difficult to detect lesions. Despite these limitations, chest X-rays still play an important role in health screenings, as they can also help evaluate heart size, lung infections, and other conditions.

Therefore, modern lung cancer screening primarily relies on low-dose computed tomography (LDCT), which is extremely fast—completing a scan in about ten seconds—with no blind spots. This method overcomes the limitations of traditional symptom-based diagnostics, such as physician auscultation and chest X-rays.

However, Vice President Cheng-Yu Chen pointed out that an LDCT scan of the lungs generates over 300 images, each approximately 1 millimeter thick—about the width of a human hair. The sheer volume of images presents a significant challenge for physicians, as the lungs contain numerous blood vessels that appear as small dots. These can sometimes be mistaken for nodules, leading to false positives and unnecessary follow-up examinations for patients.

Vice President Cheng-Yu Chen (second from the left) at the 2021 Future Tech Expo.

Taipei Medical University Integrates AI Technology for Fast and Accurate Detection of Gene Mutations and Malignancy

Early lung cancer screening is crucial, requiring the most efficient diagnostic tools to enhance accuracy and reduce false positives. In response, Taipei Medical University Hospital has developed the Lung Cancer Clinical AI Decision Support System, which can precisely identify nodule locations from over 300 images in just a few seconds. The system also generates 3D representations to outline nodule edges.

Vice President Cheng-Yu Chen explained, “By using radiomics technology, nodules captured in imaging can be extracted into tens of thousands of data points. With a database of 4,000 to 5,000 analyzed lung cancer cases for radiomics comparison, AI machine learning can be trained to automatically determine whether a patient’s lung nodule is malignant.”

Vice President Cheng-Yu Chen further stated that the newly developed software can even determine whether a tumor observed in imaging carries cancer-causing gene mutations. For late-stage patients who are not eligible for surgery, treatment typically begins with chemotherapy or targeted therapy to shrink the tumor to a certain extent before surgery, potentially extending survival.

Targeted therapy, in particular, requires tumor genetic analysis to identify mutations. Among lung cancer patients, the most common mutation is in the EGFR gene, making its detection crucial, as it allows for a wider range of precision medicine treatment options.

With the integration of AI technology in LDCT scans, the system has a 75% accuracy rate in detecting gene mutations, an 80% accuracy rate in distinguishing between benign and malignant nodules, and a 90% accuracy rate in pinpointing lung cancer locations.

AI can assist doctors in monitoring patients by evaluating lung nodules, which vary in size and have the potential to become cancerous. The American College of Radiology’s Lung-RADS guidelines assign scores based on nodule size. A score of 2 indicates a low likelihood of malignancy, requiring only annual follow-ups, while a score of 4 suggests a higher risk, necessitating more frequent monitoring every three months or a biopsy.

Physicians rely on sharp observation skills to assess nodules, but they cannot visually determine whether a tumor has an EGFR gene mutation. Differences in physicians’ interpretations of Lung-RADS may lead to varying follow-up recommendations. Scoring is complex and challenging to evaluate manually.

The AI software developed by Taipei Medical University automates Lung-RADS scoring, enhancing efficiency and accuracy in patient assessments.

Enhancing Diagnostic Accuracy and Precision Treatment to Assist in Medical Decision-Making

Vice President Cheng-Yu Chen stated Clinical Decision Support System-Shared Decision Making (CDSS-SDM) helps physicians and patients make informed treatment decisions. Taipei Medical University’s AI is not limited to LDCT scans; it also utilizes large-scale databases and genetic information to predict patient survival rates. Additionally, AI can assess the risk of brain metastasis in lung cancer patients using big data analysis. If a high risk is detected, brain MRI scans can be performed early, allowing for timely medication to prevent metastasis.

Will AI eventually replace doctors? Vice President Cheng-Yu Chen explains that AI is not infallible and may sometimes over-diagnose, leading to unnecessary tests. Ultimately, clinical decisions should be made through discussions between doctors and patients. AI serves as a supportive tool, providing data-driven insights and reminders to enhance diagnostic accuracy and efficiency. The integration of AI technology significantly improves, accelerates, and strengthens the diagnostic and treatment process.