Beijing Shijitan Hospital Seeks $1.4M Transfer of AI-Powered Laparoscopic Visualization Technology

Recently, Beijing Shijitan Hospital, Capital Medical University, released a public notice on the conversion of scientific and technological achievements, planning to transfer the patent rights of its independently developed“A Focus-Tracking AI-Based Endoscopic Visualization Method, System, and Storage Medium”Patent transferred to industry partners, with transaction amount reaching up to10 million yuan。

This technology leveragesAI-Enhanced Visualization for Endoscopic Surgery, enabling precise tracking of lesion sites, presentation of a full wide-angle field of view, and automatic obstacle avoidance functionality, thereby effectively enhancing the precision and safety of minimally invasive surgery. This innovative achievement in medical AI will inject new momentum into the intelligent development of minimally invasive surgery. The inventors of this patent areZhang Qian and His Team。

In an era of rapid advancement in surgical medical technology, minimally invasive surgery has become the mainstream treatment modality across multiple specialties—including neurosurgery, urology, and ophthalmology—owing to its core advantages of minimal trauma, rapid postoperative recovery, and reduced patient discomfort.

Especially in complex surgical scenarios such as endoscopic neurosurgery, laparoscopic delicate dissection in urology, and vitrectomy in ophthalmology, the precision of surgical maneuvers directly determines the success or failure of treatment. Lesion sites are typically closely adherent to surrounding healthy tissues; even millimeter-level operational errors can damage critical structures such as nerves and blood vessels, thereby leading to severe postoperative complications including infection and functional impairment.

Taking vitrectomy in ophthalmology as an example, critical structures such as the retina are extremely fragile and require precise tracking to avoid iatrogenic injury, which imposes stringent requirements on the real-time performance, clarity, and comprehensiveness of the surgical field of view.

Current clinical applications of traditional laparoscopic techniques face numerous intractable pain points:

First,Limitations in Field of View Coverage. Traditional endoscopes mostly adopt a single-angle view, making it difficult to simultaneously present both a "broad field of global perspective" and "small-scale details." When focusing on the lesion site, physicians may easily overlook the spatial relationships of surrounding tissues, thereby increasing the risk of operational errors.

Secondly,Manual Focus AdjustmentDuring surgery, the lesion site may shift due to factors such as changes in patient positioning and tissue retraction. In such cases, surgeons need to manually adjust the focus and angle of the endoscope by turning knobs, which not only distracts them from the procedure but also introduces adjustment latency, making it difficult to adapt to rapidly changing surgical scenarios.

Third,Insufficient Light Adaptability. Tissues in different surgical areas exhibit varying light reflection and absorption characteristics, making fixed illumination patterns prone to causing overexposure or insufficient brightness in certain regions, thereby impairing the identification of pathological details.

However, existing assistive technologies also suffer from significant drawbacks. Some robotic arm-assisted endoscopic operating systems focus solely on the mechanical control of the endoscope’s position and angle, lacking intelligent recognition of lesions and automatic focus tracking capabilities, thus failing to lock onto key anatomical structures in real time. The few devices equipped with focus adjustment functions either exhibit poor tracking accuracy, making it difficult to precisely identify minute lesions, or have sluggish response speeds that cannot keep pace with the surgical workflow. Furthermore, traditional video stitching techniques are plagued by issues such as edge distortion and inaccurate feature point matching, often resulting in discontinuities or blurring in the stitched field of view, thereby failing to provide surgeons with a coherent and clear global perspective.

Furthermore, existing technologies generally fail to address the risk of collision between the endoscope and surgical instruments, requiring surgeons to divert additional attention to monitoring the endoscope’s position, which further compromises surgical efficiency and safety. These technological shortcomings have, to some extent, constrained the advancement of complex minimally invasive surgeries toward greater precision and efficiency.

To address the shortcomings of traditional endoscopic techniques in clinical applications, such as field-of-view coverage, focus tracking, and obstacle avoidance safety, this patented technology employsArtificial Intelligence AlgorithmsAs the core driving force, it has achieved a significant leap from "passive assistance" to "active intelligence" in the field of endoscopic visualization, with its core innovations and advantages precisely aligning with the practical needs of complex minimally invasive surgeries.

First, AI-powered endoscopes canAutomatic Adaptation to Surgical Scenarios, flexibly adjust the angle and brightness while acquiringWide-Angle Video(Global Perspective) andNon-Wide-Angle Video(Local Details). By leveraging refined algorithms such as feature point matrix similarity matching and edge distortion correction, the system stitches together seamless, full-dimensional videos of the surgical field. This approach overcomes the limitations of traditional endoscopes, which often force a trade-off between field of view and detail, enabling surgeons to simultaneously maintain awareness of the global environment and observe local pathological details.

Second,Applying a Deep Learning Algorithm with a Feature Pyramid Network Architecture, efficiently extract multi-scale features from keyframe images, synchronously generate lens tracking signals and obstacle avoidance control signals, and explicitly set the priority of the tracking signal higher than that of the obstacle avoidance signal. In this way, it can ensure precise tracking of the lesion throughout the procedure without loss, while not affecting the smoothness of surgical operations.

Third, throughSimplified Geometric Models of Endoscopic and Surgical Instruments, Preset Safety Distances, by integrating surgical instrument motion trajectory prediction with path-searching techniques such as the RRT algorithm, it enables proactive collision risk assessment and safe avoidance. This eliminates the need for surgeons to manually adjust the endoscope position, thereby reducing operational burden.

Finally, the brightness, color, and distribution pattern of the LED light source are automatically adjusted based on the tissue types in the surgical field (such as mucosa, blood vessels, and muscle tissue), the position of surgical instruments, and preset illumination standards. This addresses the challenge of adapting lighting to different tissues, ensuring that pathological details remain clearly visible even in complex environments.

The core advantage of this technology lies in its ability to “precisely address clinical pain points” by leveraging AI-driven automation to replace manual intervention. This approach not only mitigates attentional distraction and operational latency associated with manual adjustments but also comprehensively enhances the practicality and safety of the surgical field through a synergistic design featuring wide-angle visualization, precise tracking, and intelligent obstacle avoidance. This technology provides “intelligent visual assistant”-level support for complex minimally invasive procedures in neurosurgery, urology, and other specialties, effectively improving surgical efficiency and precision.

As minimally invasive surgery accelerates toward precision and intelligence, traditional endoscopic equipment faces challenges such as limited field of view, manual focus adjustment, and risk of instrument collision. These pain points have become key bottlenecks restricting the success rate of complex surgeries. Meanwhile, a batch of marketed products and ongoing research projects both domestically and internationally are being implemented, jointly propelling global endoscopic visualization technology into a new era of “AI autonomous decision-making.”

Intuitive Surgical da Vinci Xi/X Surgical Robot: As a benchmark product in global minimally invasive surgical robotics, it features AI-driven dynamic field-of-view optimization, lesion focus locking, and instrument collision warning capabilities. It automatically maintains a clear view of the lesion area during abdominal and urological surgeries, thereby reducing the frequency of manual adjustments by surgeons.

FUJIFILM ELUXEO 7000 4K Endoscopy System: With its core advantages of a 170° ultra-wide field of view, AI-driven dynamic lighting adjustment, and lesion-focused tracking, this system is suitable for abdominal and gastrointestinal surgeries. This product has obtained NMPA certification.

Tinavi Tiandi 2.0 Surgical Robot: As the first orthopedic surgical robot approved in China, its endoscopic module leverages AI-based vertebral lesion tracking and multi-camera video stitching technology to achieve precise registration between preoperative CT scans and intraoperative images, supporting wide-angle field-of-view fusion. It employs feature point matching algorithms for video stitching, with AI tracking accuracy reaching the millimeter level.

MicroPort Robotics Laparoscopic Surgical Robot: As a representative product of domestically produced abdominal surgical robots, it applies “tracking signal priority” and “obstacle-avoidance path planning” technologies to predict the motion trajectory of surgical instruments. It automatically triggers avoidance when the distance falls below the 5 mm safety threshold, while ensuring that the lesion focus remains within the field of view at all times.

Zhuhai People's Hospital: AI-Based Bronchoscope 3D Real-Time Navigation System: By adopting a coupled system integration innovation approach, this solution innovatively combines traditional endoscopic image acquisition systems with a self-developed non-invasive hardware system. Without affecting the original image acquisition system, the self-developed hardware can capture real-time endoscopic images, thereby enabling reconstruction of the 3D bronchial tree from CT data and achieving clinical functions such as preoperative lesion localization, path planning, and intraoperative real-time navigation. Meanwhile, the system is compatible with 90% of endoscopic devices available on the market.

From achieving breakthrough innovations in patented technologies to the successful commercial deployment of products both domestically and internationally, and further to the accelerated advancement of multi-disciplinary R&D projects, FocusTrack’s AI-powered endoscopic visualization technology is reshaping the “visual logic” of minimally invasive surgery. In the future, with continuous optimization of AI algorithms, in-depth application of multimodal image fusion technology, and independent innovative upgrades of domestically manufactured equipment, endoscopic visualization will evolve toward being “more precise, more intelligent, and safer.” This progress will not only provide physicians with stronger technical support but also deliver a medical experience characterized by less trauma and faster recovery for patients.