一.Combining 3D modeling with positioning aids optimizes the quality of lumbar oblique X-ray images.
Spondylolysis is a significant cause of lower back pain in young adults and adolescents. While Western healthcare systems rely heavily on CT or MRI for diagnosis, traditional X-rays remain the primary diagnostic method in resource-constrained areas. However, textbooks recommend an oblique view angle of 30–50°, a range that is too broad and fails to account for the smaller lumbar lordosis angles characteristic of Asian populations, often leading to inconsistent image quality and the risk of repeat imaging. This study aims to examine the most suitable oblique lumbar view angle for Asian populations and to validate the feasibility of using auxiliary positioning devices to improve image quality.
二.Optimization Strategy for Positioning in Full-Mouth Dental Radiography Using AI
The primary objective of this study was to analyze placement errors in dental PANO (Panel Orthogonal Angiography) radiographs using fusion artificial intelligence (fusion AI) technology. A CNN trained through transfer learning was used to generate fusion predictions of placement errors, providing clinical adjustments for these errors. Therefore, this study integrated six CNN models into a fusion AI system after transfer learning. The probability values of the six types of placement errors were summed to obtain predicted values for each error type. Correlation analysis was then performed to obtain a correlation coefficient matrix, which was used to evaluate the pivotal mechanisms of placement errors and to provide subsequent clinical control strategies and measures for PANO. Experimental results demonstrated that the fusion AI system had good and statistically significant predictive capabilities for different types of placement errors. Furthermore, this study provides a correlation path diagram of placement errors, identifying key pivot points by analyzing the relationships between errors, providing radiologists with correction strategies for placement improvement or image quality control. In conclusion, this study successfully fused multiple AI classification models, which can improve the accuracy of common PANO imaging and provide valuable reference data and strategic suggestions for relevant professionals.