Over 20 years ago, a single mother accompanied her only son in his 20s for treatment—lung cancer, multiple metastases, severe pleural effusion. At that time, medications were not advanced, and radiation technology was not mature, providing limited benefit to the patient, resulting in the tragic outcome of parents outliving their child.

China Medical University Hospital has implemented AI to delineate critical organs in under a minute, allowing more time for patient care. Provided by the Radiation Oncology Department, China Medical University Hospital.

Reflecting on a case encountered during his residency, Dr. Ji-An Liang, Director of the Radiation Oncology Department at China Medical University Hospital, still expresses some regret, “As radiation oncologists, we feel gratified when old patients return healthy. Some survive luckily, but because the treatment technology wasn’t very good back then, the irradiated area was large, affecting not only tumors but also normal organs, leaving side effects. This makes us reflect, prompting the need to develop and improve to give patients a better quality of life.”

Since 1982, cancer has topped the list of the 10 leading causes of death. In recent years, about 120,000 people in Taiwan are diagnosed with cancer annually, with more than half requiring radiotherapy. The importance of radiotherapy for cancer patients is undeniable.

Traditionally, patients undergo CT (Computer Tomography) and MRI (Magnetic Resonance Imaging) scans to obtain a series of images, after which radiation oncologists manually delineate critical organs on these cross-sectional images one by one. Due to the unclear tumor boundaries, physicians need to cross-reference images for a long time, sometimes taking 4 to 5 hours per case. With the introduction of AI into medical technology, this time-consuming and labor-intensive step has found a solution.

In 2019, Dr. Ti-Hao Wang, a physician at the Radiation Oncology Department at China Medical University Hospital with an interest in information and programming, integrated AI into the delineation of critical organs. “We collaborated with several hospitals, collecting data from over 10,000 patients with head and neck cancer, esophageal cancer, lung cancer, etc. Subsequently, 20 medical centers nationwide have implemented the system.”

Although similar systems existed abroad, they were not yet mature, and Taiwan’s development was not too late. Considering self-development offered greater potential, “As medical personnel, we have advantages that general hardware manufacturers might not anticipate.” The most challenging part of development was data collection, as training an AI model requires a large amount of high-quality, standardized medical imaging data, involving patient privacy and requiring de-identification. Each hospital has its regulations and administrative processes, which Dr. Ti-Hao Wang admits is not easy. “We collaborate with academia and industry, ensuring all procedures pass individual hospitals’ IRBs (Institutional Review Boards).”

Another challenge is the variability of medical images. Cancer patients, due to metastasis, surgery, or chemotherapy, have significant differences in anatomical structure and pathological features, requiring the team to use various deep learning models and algorithms for comprehensive training, introducing cross-domain learning techniques to enhance model generalization capability. Clinical trials and feedback help optimize the model continuously.

With initial results, China Medical University Hospital transferred clinical data, practical application scenarios, and initial AI algorithm development to Ever Fortune. AI Co., Ltd., continuously improving technology and software system development. Together, they completed the “Automatic Organ Delineation System,” which has now been certified by the US FDA and Taiwan TFDA.

“This system can handle MRI or PET-CT images as long as they are in DICOM format for conversion. The software has built-in programs, allowing physicians to automatically add organ information before formulating radiotherapy plans effortlessly,” Ji-An Liang contrasts past and present. Automated delineation is completed in under a minute, applicable to various cancer types, saving time to care for patients.

University Hospital includes experienced radiation oncologists, specialist nurses, oncological physicists, and oncological technologists. Provided by the Radiation Oncology Department, China Medical University Hospital.

More importantly, the system reduces damage to normal organs. For example, accurately delineating tumors in the head and neck increases treatment efficacy while protecting the salivary glands, reducing dry mouth symptoms. Similarly, for lung cancer metastasis patients like the young man 30 years ago, using the automatic delineation system with automated planning can achieve hippocampal avoidance in whole-brain radiation, reducing damage to memory regions.

As cancer is increasingly considered a chronic disease, treatment approaches avoid severe harm to healthy tissue. With technological progress, care is becoming more precise and efficient, providing radiation oncologists more opportunities to see old patients return in good health.