With the deepening application of artificial intelligence (AI) technology in the field of medical devices, pre-trained models have increasingly become a crucial engine driving innovation in intelligent healthcare due to their efficiency, generalization capability, and transfer learning performance. However, potential risks associated with pre-trained models—such as issues related to source diversity and quality controllability —pose new challenges to the safety and effectiveness of AI-based medical devices. Against this background, the National Medical Products Administration (NMPA) released the sectoral standard YY/T 1833.5-2024 Artificial Intelligence Medical Devices—Quality Requirements and Evaluation-Part 5: Pre-trained Models in September 2024. This standard provides essential technical guidance and a regulatory framework for standardizing the application of pre-trained models in medical devices, marking a milestone in ensuring the safety and efficacy of AI-powered medical products. This article offers an in-depth interpretation and analysis of the standard's background, orientation, and key technical consideration. It elucidates specific requirements regarding documentation for pre-trained models, definitions of critical quality attributes, and conformity assessment pathways. Furthermore, the practical and far-reaching implications of the standard are discussed, including its role in enhancing quality assurance throughout the entire lifecycle of AI-based medical devices and guiding technological innovation and healthy development within the industry. Additionally, by dissecting the standard, this article aims to support the industry in conducting prudent evaluations during model selection phases, thereby reducing the application of low-quality and high-risk models.

Objective:To determine the direction of traction that has the least influence on temporomandibular joint (TMJ) during mandibular distraction osteogenesis(MDO).Methods:(1) The three-dimensional finite element model was established after the mandible data were obtained by cone beam computed tomography (CBCT) scanning, and the validity of the model was verified. (2) Based on the verified three-dimensional finite element model, the distraction osteogenesis of mandible was carried out with six different directions. (3) The effects of different traction directions on disc pressure, osteotomy displacement, temporal bone pressure and condylar pressure were measured.Results:(1)In the thinnest area of the articular disc, the direction of the minimum stress on the upper and lower surfaces of the disc was "along the direction of the mandible, parallel to the surface of the mandible" . (2) Under the same load, the displacement of each osteotomy surface along the traction direction was different, but the difference was not big, and the effect was basically the same. (3) "Along the direction of the mandible, parallel to the surface of the mandible" and "along the direction of the mandible and parallel to the sagittal plane" had less pressure on the temporal bone than other situations, and the two directions of traction were parallel to the mandible. After the force decomposition, the direction of the condyle was the smallest, so the pressure on the condyle was smaller.Conclusions:A satisfactory TMJ model can be obtained by DICOM. The traction force of "along the direction of the mandible body, parallel to the surface of the mandible body" and "along the direction of the mandible body and parallel to the sagittal plane" had the least effect on the TMJ. When designing the traction direction of MDO, we should not only consider the influence of surgery on the shape of jaw and upper respiratory tract, but also weigh in the influence of traction direction on the TMJ, and determine the direction of traction which has the least influence on the TMJ.

Objective:To determine the direction of traction that has the least influence on temporomandibular joint (TMJ) during mandibular distraction osteogenesis(MDO).Methods:(1) The three-dimensional finite element model was established after the mandible data were obtained by cone beam computed tomography (CBCT) scanning, and the validity of the model was verified. (2) Based on the verified three-dimensional finite element model, the distraction osteogenesis of mandible was carried out with six different directions. (3) The effects of different traction directions on disc pressure, osteotomy displacement, temporal bone pressure and condylar pressure were measured.Results:(1)In the thinnest area of the articular disc, the direction of the minimum stress on the upper and lower surfaces of the disc was "along the direction of the mandible, parallel to the surface of the mandible" . (2) Under the same load, the displacement of each osteotomy surface along the traction direction was different, but the difference was not big, and the effect was basically the same. (3) "Along the direction of the mandible, parallel to the surface of the mandible" and "along the direction of the mandible and parallel to the sagittal plane" had less pressure on the temporal bone than other situations, and the two directions of traction were parallel to the mandible. After the force decomposition, the direction of the condyle was the smallest, so the pressure on the condyle was smaller.Conclusions:A satisfactory TMJ model can be obtained by DICOM. The traction force of "along the direction of the mandible body, parallel to the surface of the mandible body" and "along the direction of the mandible body and parallel to the sagittal plane" had the least effect on the TMJ. When designing the traction direction of MDO, we should not only consider the influence of surgery on the shape of jaw and upper respiratory tract, but also weigh in the influence of traction direction on the TMJ, and determine the direction of traction which has the least influence on the TMJ.