3D reconstruction model of atlantoaxial joint gap based on Cxy-Net

Published: Jan. 9, 2025

One of the surgical devices used to treat cervical spine disorders. The atlantoaxial lateral block is fusion device. Atlantoaxial joint space reconstruction one key steps in use device, whereas conventional 3D suffered from low precision and accuracy, as well inability take into account its dynamic properties accurately. To address these issues, this work proposes a parallel segmentation model. By using patient's CT datas input, gap reconstructed by edge detection module model paper, visualized output. In module, an advanced image algorithm based on Cxy-Net adopted optimize extract details gap. average Hausdorff distance (Hd) 10.5211 mm, symmetric surface (ASD) 0.3861 overlap (So) reaches 90.09%, Dice similary coefficient (Dice) 0.8834, accuracy (AC) 0.8914. Compared with modeling, present paper improves coefficient, about 15.37%, 8.96%, 4.84% respectively.

Language: Английский

---

Improved SwinUNet with fusion transformer and large kernel convolutional attention for liver and tumor segmentation in CT images

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: April 24, 2025

Segmentation of both liver and tumors is a critical step in radiation therapy hepatocellular carcinoma. Despite numerous algorithms have been proposed for organ tumor delineation, automatic segmentation livers remains challenge due to their blurred boundaries low tissue contrast compared surrounding organs within CT images. The U-Net-based methods achieved significant success this task. However, they often suffer from the limitation that feature extraction lacks relationships, i.e., context, among adjacent areas, thereby leading uncertainty results. To address with challenge, we incorporate global-local context attention into Swin-UNet. Firstly, introduce Swin-neighborhood Fusion Transformer Block (SFTB) capture global local an image, enabling us distinguish instances effectively. Secondly, design Large-kernel Convolutional Attention (LCAB) two types highlight crucial features. Experiments on LiTS 3D-IRCADb datasets demonstrate effectiveness method, dice scores 0.9559 0.9610 segmentation, 0.7614 0.7138 segmentation. code available at https://github.com/JennieHJN/image-segmentation/tree/master .

Language: Английский

---

Systematic Review: AI Applications in Liver Imaging with a Focus on Segmentation and Detection

Life, Journal Year: 2025, Volume and Issue: 15(2), P. 258 - 258

Published: Feb. 8, 2025

The liver is a frequent focus in radiology due to its diverse pathology, and artificial intelligence (AI) could improve diagnosis management. This systematic review aimed assess categorize research studies on AI applications from 2018 2024, classifying them according areas of interest (AOIs), task imaging modality used. We excluded reviews non-liver non-radiology studies. Using the PRISMA guidelines, we identified 6680 articles PubMed/Medline, Scopus Web Science databases; 1232 were found be eligible. A further analysis subgroup 329 focused detection and/or segmentation tasks was performed. Liver lesions main AOI CT most popular modality, while classification predominant task. Most (48.02%) used only public datasets, 27.65% one dataset. Code sharing practiced by 10.94% these articles. highlights predominance tasks, especially applied lesion imaging, often using imaging. Detection relied mostly external testing code lacking. Future should explore multi-task models dataset availability enhance AI’s clinical impact

Language: Английский

---

TULTS-Net: Local-Global Feature-Aware Transformer and inter-layer feature interaction for medical image processing

Digital Signal Processing, Journal Year: 2025, Volume and Issue: unknown, P. 105195 - 105195

Published: March 1, 2025

Language: Английский

---