China Focus: AI diagnosis boosts accuracy in clinical blood cell testing

CHONGQING -- A Chinese medical research team has developed a patented model using artificial intelligence technology to speed up and automate blood analysis, allowing laboratories to provide faster and more accurate feedback to doctors.

Hospital laboratories generally evaluate millions of blood cells for disease diagnosis every day, which is a repetitive task usually performed manually by trained physicians who inspect blood samples through microscopes, one after another.

The new AI model has been hailed as a groundbreaking development, as it has the potential to revolutionize how laboratories analyze blood samples, streamlining the evaluation process and flagging potential abnormalities with exceptional efficiency.

A blood smear is a conventional examination used to look for abnormalities in blood cells. It is an important clinical diagnosis of blood and lymphatic system disorders. However, the current analysis of blood smear images under a microscope is time-consuming and mainly relies on the skills and experience of the physician. Many Chinese clinics at the grassroots level are lacking well-trained professionals, resulting in a low rate of diagnosis.

Based on 2,500 microscopic blood smear images of patients at the Chongqing-based Xinqiao Hospital under the Army Medical University from 2018 to 2022, researchers developed the AI-assisted model, which turned detection experiences, accumulated by experts, into algorithms. The AI has learned to recognize which cells are normal or abnormal with an accuracy rate of over 90 percent in a shorter time than human eyes, said Peng Xiangui, leader of the research team and chief examiner at the hospital's hematology center.

"Physicians spend half an hour or more in examining a blood sample and generating a report, while the AI model just needs a few minutes per case," said Peng, who has more than 30 years of experience in the field of hematology.

Peng went on to explain that in order to train, validate and test the AI model, his team fed it with data involving 300,000 red blood cells collected over the past five years, thereby resolving a challenging identification problem.

The results showed that it can accurately identify 49 types of abnormal blood cells, covering most of the common diseases seen in the clinic.

In addition to reducing manual labor, Peng believes that such a medical AI model will assist inexperienced physicians in reaching the level of an expert and boost their work efficiency.

"A doctor can only examine a limited number of cases in his or her lifetime, whereas the AI database can accumulate unlimited information, making diagnosis more scientific and accurate," Peng told Xinhua.

This is the latest example of Chinese researchers using AI technology to assist physicians in clinical diagnosis. From diagnostic screening to surgical robots, AI-assisted medical products are gaining momentum in hospitals across China.

Zhang Xi, head of the hospital's hematology center, said the AI device was granted a Class II medical device license in September last year and has obtained 32 patents. Several domestic medical institutions have expressed their interest in collaborating with the research team.

"We will explore further potential applications of the AI system to assist more primary care hospitals and serve more patients," Zhang added.