This study aimed to explore the risk factors for preoperative deep vein thrombosis (DVT) in patients with traumatic lower-extremity fracture (TLEF), analyze the correlation and predictive efficacy of the systemic immune-inflammation index (SII) in DVT development, and establish a corresponding nomogram prediction model. This study retrospectively analyzed the clinical data of 950 patients who underwent surgery for TLEF at our hospital between May 2022 and April 2024. The enrolled patients were randomly divided into the training group (665 cases) and the validation group (285 cases) at a 7:3 ratio. We performed univariate and multivariate logistic regression analyses in the training dataset to screen independent risk factors for preoperative DVT in TLEF patients. R Studio was used to construct the nomogram model. The goodness-of-fit test and ROC curve were adopted to evaluate the efficacy of the model. Age, fracture site, SII, D-dimer and APTT were identified as independent predictive factors for preoperative DVT. The nomogram exhibited good agreement with the ideal model in both the training and validation datasets. The nomogram prediction model for preoperative DVT risk in TLEF patients was developed by incorporating SII and other risk factors. It can assist clinical practitioners in early assessment of DVT development.
Ai-Ling Yang, Ya-Min Feng, Yang Shen et al.· BMC Musculoskeletal Disorder...· 0 citations
Introduction Ginkgo fruits in natural field environments are characterized by low color saliency, small target size, severe occlusion by branches and leaves, and complex illumination variations. These factors significantly reduce real-time detection accuracy and limit the applicability of existing object detection methods in practical agricultural scenarios. Methods To address these challenges, this study proposes a real-time ginkgo fruit detection method based on an improved YOLOv11n framework. A multi-scenario dataset was constructed by collecting ginkgo fruit images under diverse lighting conditions, occlusion levels, and viewing angles, and data augmentation strategies were applied to improve sample diversity and model generalization. On this basis, a CFNet channel fusion module was embedded into the backbone network, a DynamicHead detection head was introduced to enhance multi-scale feature representation, and the original loss function was replaced with the Efficient IoU (EIoU) loss to improve bounding box regression accuracy. These improvements collectively form the proposed CED-YOLOv11n model, achieving a balanced optimization between detection accuracy and inference efficiency. The effectiveness of the model was validated through Grad-CAM visualization analysis, ablation studies, and comparative experiments with classical object detection models. Results Experimental results show that the proposed CED-YOLOv11n achieves a precision of 94.6%, a recall of 85.4%, and a mean average precision (mAP) of 93.8% on the constructed ginkgo fruit dataset. In addition, the model is lightweight, with a parameter size of only 4.96 MB, and achieves an inference speed of 53.2 FPS, demonstrating strong real-time performance. Compared with mainstream object detection models such as DETR, Faster R-CNN, and YOLOv5, the proposed method achieves superior overall detection performance. Discussion The results indicate that the proposed method effectively enhances the accuracy and efficiency of ginkgo fruit detection in complex natural environments. It provides technical support for the development of vision-based perception systems in intelligent ginkgo harvesting equipment and offers a reference for further optimization and integration of smart agricultural machinery systems.
Zhenyang Lv, Zhentao Wang, Boyuan Tan et al.· Frontiers in Plant Science· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
