Lung diseases, such as lung cancer, chronic obstructive pulmonary disease and pneumonia, constitute a significant global health threat, with millions of individuals affected every year. Early diagnosis and treatment are thus of significant importance for improving public health. Bronchoscopy, a minimally invasive diagnostic and therapeutic procedure, has emerged as an essential tool for detecting, treating, and managing various lung diseases. However, the availability of bronchoscopic services is predominantly constrained by the need for more experienced doctors in underdeveloped regions.
The research team led by Professors LU Haojian, WANG Yue, and XIONG Rong from the Zhejiang University College of Control Science and Engineering developed an AI co-pilot bronchoscope robot. This technology empowers novice doctors to adeptly conduct lung examinations, significantly reducing dependence on experienced doctors and effectively enhancing the diagnostic capabilities for lung diseases in underdeveloped regions. This groundbreaking technology holds great promise for revolutionizing the field of lung disease examination and diagnosis. Their findings were published online in the journal Nature Communications on January 4.
“Bronchoscopy calls for a remarkable level of skills and experience, leading to a substantial disparity in the quality of care provided by experts and novice doctors. Therefore, we hope to integrate robotics technology and artificial intelligence to reduce the medical burden and compromise the demand for highly skilled doctors,” said LU Haojian.
Fig. 1: Overview of our AI co-pilot bronchoscope robot deployed in a clinical setting for bronchoscopic procedures.
To address this issue, the team developed a plug-and-play bronchoscope robot system based on magnetic adsorption for the quick connection and replacement of the catheter. They designed two types of catheters, with diameters of 3.3 mm (with a 1.2 mm working channel) and 2.1 mm (without a working channel), suitable for patients of various ages and with the potential for a detailed examination of narrower passages in the patient’s lungs. In addition, compared to imported medical robot systems that cost millions of dollars, the hardware cost of this robot is economically friendly.
Fig. 2: AI–human shared control algorithm and training strategy.
To reduce the risk of catheter damage to lung tissues and improve efficiency during bronchoscopy examinations, the team also developed an AI-human shared control algorithm. This algorithm, based on style transfer and the embodied intelligence approach of simulation-to-reality, enables the trained strategy network to predict a steering action (pitch and yaw angles) for the robot’s orientation based on bronchoscope images and coarse-grained human commands (up, down, left, right, or forward).
“By incorporating AI technology into the bronchoscope robot system, doctors do not need to intervene throughout the process, remarkably reducing their time and efforts channeled for bronchoscopy examinations,” said WANG Yue.
Fig. 3: Results of in vivo experiments.
Ultimately, the team evaluated the performance of the AI co-pilot bronchoscope robot in a minipig, whose bronchial structure closely resembles that of the human bronchus. The results showed that novice doctors, guided by an AI co-pilot, outperformed experts with decades of experience. Thanks to AI technology, the bronchoscope head could stay at the center of the airway in real-time, ensuring an unobstructed field of view and reducing collisions with the airway wall.
“Once this cutting-edge technology is utilized in the future, it will appreciably lower the threshold for operational experience during bronchoscopy medical examinations. This technology is expected to alleviate the issue of insufficient diagnostic capabilities for lung diseases in economically underdeveloped regions,” LU Haojian pointed out.
（From ZJU NEWSROOM）