Zhang Xun: Artificial intelligence assists in the diagnosis of small nodules in the lungs, and this is a path that we have to take

Video loading...

Q: You are the first surgical method in the world to treat the giant pharyngeal esophageal diverticulum, known as Zhang's surgery. With your experience of pioneering Zhang's surgery, can you share how to innovate in the field of thoracic surgery?

A: First, clinical research should start from the perspective of clinical practical problems and find scientific research entry points from the perspective of solving the clinical needs of patients. For example, the laparoscopic resection of the huge pharyngophageal diverticulum that I did, a patient who could not find a suitable safe and effective solution according to the traditional methods of the past.

Because the diverticulum is huge, partly in the neck and partly in the chest, according to traditional practice, whether it is from the neck or from the chest opening, there is no way to solve the problem of complete resection, and the trauma is very large. In order to solve this problem, I am constantly thinking about how a new idea can be adopted, which shows that clinical research should start from practical needs.

We designed a new surgical method that was minimally invasive, safe and effective, which is truly an international first. Practice has proved that this procedure is very effective from the long-term follow-up of patients who have already undergone surgery. It is necessary to translate clinical needs into scientific research content, rather than making things behind closed doors.

We have many problems that have not been effectively solved in the clinic, which is an entry point for us to engage in scientific research. Unfortunately, some people miss this entry point, have a good opportunity, but do not find an effective way.

Some people innovate for the sake of innovation, and make a new treatment or a surgical formula that is divorced from clinical needs, which may actually complicate a simple problem. Therefore, this is an important point, clinical research must find an entry point from clinical needs.

Secondly, challenges and opportunities coexist. When solving problems in the clinic, no one has done it before, which is a challenge for us, but it is also an opportunity, giving us the opportunity to innovate and solve problems that have not been solved by previous generations.

These two points are what we should pay attention to when we do clinical research, especially young doctors. I think that if we can do these two things, solve the clinical reality, and seize this opportunity to face the difficulties, we will often come up with practical, effective, fruitful, and high-level scientific research projects.

Q: What is the current progress in the imaging diagnosis of AI-assisted lung nodules? Can you talk to us about its application prospects and clinical value?

We've been doing ai-aided work on the diagnosis of small lung nodules for over five years. So far, artificial intelligence-assisted lung nodules have been analyzed in our hospital more than 60,000 cases, and has become a diagnostic method that is very much needed and used in the clinical work of our hospital and the First Hospital of Xiamen University, whether it is imaging, internal oncology, and thoracic surgery. The problems it solves are:

First, locate the nodule. Whether the patient has a small nodule is a big problem at present, because the number of people participating in the census is increasing, and a considerable number of people find that the lungs have small nodules through CT examination, which will bring a lot of mental pressure to people.

Whether there are small nodules, how many, and more importantly, whether the small nodules are benign or malignant, especially for very small nodules in sub-centimeters, it is difficult for doctors to make a very accurate judgment. This puts a lot of pressure on patients to do CT re-examinations frequently and everywhere, consulting doctors, which creates the problem of over-diagnosis and over-treatment.

Artificial intelligence assists the image diagnosis of small nodules, one is that it can be accurately located. In addition, it can relatively accurately assist doctors in diagnosing the degree of malignancy of small nodules, which is what we call the probability of malignancy.

According to the statistics of a clinical data we have just completed recently, in the past two years of 810 small nodule surgeries in the department of thoracic surgery in our chest hospital, preoperative AI gave a diagnosis of malignant probability, and we compared it with postoperative pathology to see its diagnostic accuracy. Because AI diagnostics are not just benign and malignant, either/or, it suggests a malignant probability from 0 to almost 100%.

Of these 810 cases, lung cancer pathological diagnosis accounted for 640. According to our findings, the average probability of malignancy in these 640 cases of pathological diagnosis is 86%. Of the cases with an AI-diagnosed probability of malignancy of small nodules, 94% were malignant, 5% were precancerous lesions, and only 1% were diagnosed as benign lesions. AI can largely solve our judgment of the preoperative benign and malignant nature of small nodules. Then group it, according to the size of the nodule (diameter) is sub-centimeter, 1 to 2 centimeters, or 2 to 3 centimeters, and finally found that the smaller the diameter of the part, the greater the proportion of the probability of MALIgnancy of AI diagnosis of more than 86%.

In addition, it is evaluated by dividing it into different densities. Now clinical nodules are divided into pure ground glass, sub-solid and solid. Subunits are further divided into small nodules with a solid component greater than 50% and less than 50%.

According to the analysis of the results, the more pure ground glass-like, the solid composition of less than 50% of the small nodule, the higher the probability of AI judging its malignancy. Malignant tumors are very small, and they are pure ground glass and sub-solid is a difficult point in our clinical diagnosis. It is precisely on this difficult point that AI provides us with a more reliable basis for auxiliary diagnosis. Therefore, AI will become a very important diagnostic means to improve the accuracy of clinical diagnosis of small nodules in the lungs.

In addition, the question of localization, for the surgeon, which lung segment it is in, and what is the relationship between the surrounding vascular trachea, is also what we need to be clear when we take sub-lobectomy, wedge resection or lung segment resection surgery.

We now also integrate three-dimensional imaging technology into the clinic, such as the need to perform surgery on a small nodule, its malignant probability is relatively high, we can call up its three-dimensional imaging at any time, see the location of the lung segment where it is located, the relationship between the surrounding blood vessels.

Now some companies are dedicated to doing these jobs, the cost is high, and it will take a long time. And we're now injecting it directly into the AI system, ready to move, with no cost or time difference. This is also convenient for surgeons, and now the proportion of subpulmonary resection surgeries is getting higher and higher, and AI is also of great help to surgeons.

So from the work we have done for more than five years, we can say with great confidence that ai-assisted diagnosis of small lung nodules is completely feasible clinically, and it is a path we must take. AI is very effective in improving the early diagnosis and treatment of lung cancer and preventing over-diagnosis and over-treatment. Therefore, the country is now also taking artificial intelligence as a national strategy, and all countries have invested a lot of manpower and material resources in artificial intelligence, and one of the most important pieces is the medical field.

For thoracic surgery, AI is geared toward small nodules, and we are now using it more than just imaging for diagnostics. We call it the AI-assisted lung cancer diagnosis and treatment integrated solution, which includes AI-assisted diagnosis of lung small nodules, AI-assisted diagnosis of pathological diagnosis, and AI-assisted diagnosis and treatment decision-making. That is to say, after the diagnosis of lung cancer, whether it should be surgery, radiation therapy and chemotherapy, targeted therapy, or immunotherapy, it can be automatically grabbed by the patient's data.

Because this server is placed in the network management of our hospital, after grabbing the patient's data at any time, it automatically prompts the doctor what further tests the patient should do, and if the diagnosis is confirmed, the treatment method should be taken.

One of the main bases of AI is the guide, which can compare the guide, and the second is the literature, which can prompt the doctor with the latest literature. The other is clinical data. In addition to AI-assisted diagnostic and diagnostic decision-making, there is also a point that AI structures clinical data after automatically collecting it. We clinicians basically focus on doing a good job in clinical work, and the entire data collection, collation, and follow-up can be completed by artificial intelligence.

This not only completes the clinical work, but also provides a very good resource and platform for our scientific research. Therefore, artificial intelligence is promising in our clinic, and there is still a lot of room for future development.

Professor Zhang Xun

Professor Zhang Xun

Director of Thoracic Surgery Department of Tianjin Chest Hospital

Professor, doctoral supervisor and postdoctoral cooperative teacher of Tianjin Medical University

Experts who enjoy special government allowances from the State Council

Chairman of the Thoracic and Cardiovascular Surgery Branch of the Chinese Medical Association

President of Thoracic Surgeons Branch of Chinese Medical Doctor Association

Chairman of Thoracic and Cardiovascular Surgery Branch of Tianjin Medical Association

Deputy Leader of the Esophageal Specialty Group of the Thoracic and Cardiovascular Surgery Branch of the Chinese Medical Association

Vice Chairman of the Minimally Invasive Thoracic Surgery Expert Committee of the Chinese Medical Doctor Association

Associate Editor of Chinese Clinical Journal of Thoracic and Cardiovascular Surgery, a member of the Expert Committee on Clinical Pathways of Thoracic Surgery of the Ministry of Health

Associate Editor of the Journal of Thoracic and Cardiovascular Surgery (Chinese Edition).

Note: This article is based on the interview video content of the "First China Thoracic Surgery Specification and Innovative Surgery Summit Exhibition" of the People's Health Network's China Thoracic "Everyone" Talk Series Interview with Professor Zhang Xun.