Jilin University to License AI-Powered Dental Imaging Analysis Technology for Oral Surgery at RMB 1.1 Million

Recently, Jilin University released a public notice on the transformation of scientific and technological achievements, proposing to transfer them through negotiated pricing.“Intelligent Analysis System and Method for Dental Imaging in Oral and Maxillofacial Surgery”Relevant patents are licensed to industry partners, with licensing fees set at1.1 million yuan, with a licensing term of two years. The inventor of the patent isXu Zhimin and His Team。

This technology belongs toTechnical Field of Dental Image Analysis, which discloses an intelligent analysis system and method for dental imaging oriented toward oral and maxillofacial surgery. This method integrates multi-scale features with spatial contextual information to enhance the accuracy and robustness of dental image analysis, thereby providing more reliable support for oral and maxillofacial surgical procedures.

The precise execution of oral and maxillofacial surgical procedures relies heavily on meticulous analysis of dental imaging. However, existing technologies present numerous significant pain points in clinical practice, severely compromising diagnostic and therapeutic efficiency and safety.

The anatomical structures of teeth and their surrounding tissues are complex, with significant individual variations. Traditional image analysis largely relies on physicians’ subjective experience, which is not only time- and labor-intensive but also prone to human error, resulting in poor diagnostic consistency and making it difficult to accurately capture the relationship between subtle lesions and complex anatomical structures.

From a technical perspective, existingDigital Image Processing MethodsThere are significant limitations.

On one hand, most techniques focus on shallow feature extraction and fail to adequately mine deep semantic information, resulting in ambiguous lesion localization that cannot meet the stringent precision requirements of oral and maxillofacial surgery. On the other hand, dental images often suffer from blurred boundaries due to factors such as acquisition angles and device resolution. Traditional algorithms exhibit weak anti-interference capabilities, making it difficult to achieve clear segmentation, which hinders subsequent surgical planning.

Furthermore, existing systems lack effective integration of spatial contextual information. In complex scenarios characterized by diverse tooth morphologies and interwoven tissues, feature extraction remains insufficient, further compromising the robustness and reliability of image analysis.

These issues directly lead to inefficiencies in clinical diagnosis and treatment:Physicians must devote substantial time to manually analyzing medical images, thereby prolonging the surgical preparation cycle. Surgical plans developed based on ambiguous imaging are prone to deviations, increasing intraoperative risks and the likelihood of postoperative complications. For procedures demanding extremely high precision, such as dental implantology and orthodontics, existing image analysis technologies struggle to meet clinical needs. There is an urgent demand for an intelligent, high-precision, integrated analysis solution to overcome the core bottlenecks in precise diagnosis and treatment.

The core advantage of the patented technology “Intelligent Analysis System and Method for Dental Imaging in Oral Surgery” lies inConstructing an Integrated Solution with “Deep Multi-Scale Feature Mining + Spatially Adaptive Intelligent Enhancement”, achieving end-to-end technological innovation from feature extraction to semantic segmentation, and thoroughly overcoming the limitations of traditional dental image analysis, namely “low accuracy, poor robustness, and reliance on subjective experience.”

This technology was the first toFeature ExtractionLevelAchieved a disruptive breakthrough, innovativelyAdopting the Swin Transformer backbone network, simultaneously acquiring multi-scale feature maps from the superficial, middle, and deep layers of dental images, thereby breaking through the limitations of traditional algorithms that focus solely on single-level features.

Shallow Feature MapUsed to capture basic information such as edges and color contrast,Mid-level Feature MapEmphasizing abstract structures such as the relative positions of teeth,Deep Feature Mapthereby extracting high-level semantic information, such as the three-dimensional morphology of teeth. These three layers of features comprehensively cover both the details and global structure of the images, laying a data foundation for precise analysis.

Meanwhile, for the mid-level and deep-level feature maps, a design was developedSpatial Adaptive Module. The offset learning module dynamically adjusts the receptive field positions to precisely adapt to the irregular morphology and complex boundaries of teeth; meanwhile, the modulation scalar learning module regulates feature response intensity to highlight critical anatomical structural information. Following bilinear interpolation sampling and position-wise dot product operations, spatially enhanced feature maps are generated, significantly improving the model’s sensitivity and robustness to detailed features.

InEdge Segmentation and Feature Fusionphase, this technology has constructed“Precision Testing + Complementary Integration” Closed-Loop System, to overcome the challenge of blurred boundaries.

First, an edge detection branch was innovatively designed.The shallow feature maps are passed through a convolutional layer to focus on edge features. Subsequently, an edge feature mask is generated using the Sigmoid activation function and element-wise multiplied with the shallow feature maps based on spatial positions, thereby accurately extracting the tooth edge information feature map. This approach effectively addresses the limitation of traditional algorithms in producing unclear edge segmentation.

Secondly, a multi-dimensional feature fusion strategy is adopted.First, spatially enhanced feature maps and edge information feature maps are integrated via skip connection layers to form an initial fused feature map. Subsequently, dilated convolutions with varying dilation rates (3, 5, and 1) are employed to achieve cross-layer structural decomposition; cross-layer geometric feature responses are calculated based on mid-level features, and fusion correction is performed by combining shifted windows with a relative position encoding mechanism. Finally, resolution is restored through upsampling to generate a multi-scale significantly fused feature map, thereby achieving complementary optimization of detailed information and semantic information.

Furthermore, this technology inIn terms of clinical utilityWith“High Adaptability + Strong Reliability”significant advantages.

In terms of adaptability,It supports the input of common oral and maxillofacial surgical imaging data, such as CBCT, and is compatible with imaging data from various scanning parameters and device resolutions. It is widely applicable to multiple surgical scenarios, including dental implantation, orthodontics, and tumor resection.

In terms of accuracy,The classification layer captures subtle anatomical features via pointwise convolution, generates a tooth voxel probability map through the Softmax classifier, and then converts it into a discrete segmentation label map using thresholding, connected component analysis, and morphological operations, thereby ensuring that the segmentation results align with the true anatomical structures.

In terms of efficiency,The fully automated analysis process requires no manual intervention, significantly reducing image processing time and providing efficient support for surgical planning.

Currently, addressing the existing issues in oral and maxillofacial surgery“Insufficient imaging analysis accuracy, reliance on subjective experience, and low efficiency”To address these core pain points, medical technology companies both domestically and internationally have centered their efforts around“AI-powered algorithms, multimodal data fusion, and clinical scenario adaptation”Accelerating deployment in these three key areas, leveraging technological innovation to propel dental imaging analysis from "manual-driven" to "intelligent and precise," thereby meeting the clinical demands of diverse scenarios such as implantology, orthodontics, and oral surgery.

Fusion TechLaunched“Shennong Cloud CT”Equipped with AI-assisted diagnostic capabilities, it leverages AI-driven metal artifact suppression technology to generate high-quality images. Meanwhile, its large 18×20 cm field of view design enables precise capture of teeth and surrounding anatomical structures, providing a reliable data foundation for subsequent analysis.

Its supportingDigital Intelligence Engine DentalX™It achieves deep collaboration among cloud data storage, remote image interpretation, and AI-assisted diagnosis, enabling physicians to rapidly complete tasks such as nerve canal identification and implant site planning.

Dentsply Sirona's Primescan Chairside Intraoral ScannerIntegrated with AI-powered 3D reconstruction technology, it rapidly acquires high-precision intraoral digital models and automatically identifies tooth margins, proximal relationships, and occlusal features, providing data support for the fabrication of surgical guides and the design of prosthetic restorations.

itsCEREC Intelligent Restoration SystemFurther integrated with AI algorithms, it enables automatic generation of tooth morphology and optimization of restorations. The related products have obtained certifications from multiple countries and regions, including the FDA, CE, and NMPA.

It should be noted that itsCariVu Caries Detection DeviceIt is a standalone handheld near-infrared imaging device, not integrated with the Primescan or AI diagnostic modules, primarily used for early caries screening.

Landray Medical (Landray)As a representative enterprise of domestically produced CBCT, it focuses on the deep integration of dental imaging equipment and intelligent algorithms. ItsSmart3D - X Series CBCTEquipped with a proprietary intelligent segmentation engine, it leverages deep learning technology to automatically and precisely segment teeth, jawbones, and the inferior alveolar nerve, generating 3D visualization models. It also provides intelligent alerts for suspicious lesion areas (such as low-density shadows) to assist clinicians in identifying pathologies like periapical periodontitis and jaw cysts, making it suitable for implantology, orthodontics, and surgical planning. This product has obtained NMPA Class III medical device certification.

Align Technology (parent company of Invisalign)Integrating AI deeply into the entire orthodontic workflow. ItsiTero Intraoral Scanner and ClinCheck Pro Software PlatformThrough collaborative efforts, it can automatically generate personalized tooth movement simulation plans based on scan data for dentist review and adjustment.

The company has further advanced multimodal data fusion, enabling the integration of iTero intraoral scan data with third-party CBCT images and facial scan information for comprehensive assessment of complex cases (such as impacted teeth and orthognathic surgery).