Chinese scientists have applied artificial intelligence to help doctors more precisely separate tumors from surrounding healthy tissue — a process they liken to removing pits from fruits — improving the accuracy of radiotherapy while limiting damage to vital organs.

During a recent session at the Sun Yat-sen University Cancer Center in Guangzhou, doctors demonstrated the approach on a patient with nasopharyngeal carcinoma, a type of cancer that starts in the nasopharynx — the upper part of the throat behind the nose.

After the patient was scanned using an integrated CT imaging and treatment machine, detailed images of the tumor appeared on a computer screen. Using what researchers call "digital dissection", the AI system quickly outlined the area that needed radiation treatment.

Doctors then reviewed and adjusted the AI-generated outline before finalizing the treatment plan.

The machine proceeded with adaptive radiotherapy — a method that adjusts treatment in real time based on the patient's current condition — completing the session in less than 30 minutes.

Ma Jun, vice-president of the cancer center and an academician of the Chinese Academy of Sciences, said radiation therapy is the primary treatment for nasopharyngeal carcinoma.

If the radiation field is too small, parts of the tumor may be missed, increasing the risk of recurrence. If it is too large, it can harm nearby critical structures such as the brainstem, temporal lobe, middle ear and optic nerve.

Such damage can lead to complications including headaches, memory loss, hearing problems and vision impairment, significantly affecting a patient's quality of life, Ma said.

Before radiotherapy begins, doctors must identify and outline both the tumor and nearby healthy organs on medical images — a process known as target volume delineation. This step ensures radiation is delivered accurately.

Sun Ying, a professor at the center, said the task was previously time-consuming and demanding.

Doctors often had to spend three to six hours on each patient carefully marking the target areas while maintaining intense concentration, she said.

The challenge is compounded by the nature of the disease. Du Xiaojing, a chief physician in the radiotherapy department, said patients typically undergo more than 30 treatment sessions over six to seven weeks.

During that time, tumors may shrink and patients may lose weight, causing shifts in the tumor's position. As a result, images taken days earlier may no longer accurately reflect the tumor's location at the time of treatment, Du added.

To address this issue, a research team led by both Ma and Sun spent more than a decade developing the "digital dissection" technique.

The system uses large datasets to analyze how tumors tend to grow and change over time, and applies AI to automatically generate precise treatment outlines. This helps doctors adapt plans quickly as the patient's condition evolves.

Zhou Guanqun, another chief physician at the center, said the technology now achieves a level of accuracy exceeding that of 50 percent of specialist physicians.

"It has reduced differences between doctors' outlines by 50 percent, improved efficiency by more than fivefold, and shortened treatment time per case to about 30 minutes," Zhou said.