Strategically selecting a combination of synthetic tumor sizes and generating synthetic tumors with precise boundaries significantly improves the accuracy of deep learning-based pancreatic tumor segmentation models.
This research aims to investigate the relationships between microstructural and macrostructural features of the cervical spinal cord in a healthy population using quantitative MRI analysis, and to develop a deep learning-based segmentation framework for accurate measurement of macrostructural characteristics.
Leveraging diverse pathological clues, including segmented regions, entities, and report themes, to build fine-grained cross-modal representations and seamlessly transfer them to enhance the quality of generated brain CT reports.
A cascading refinement CNN model, MS-CaRe-CNN, can semantically segment the left and right ventricle, healthy and scarred myocardial tissue, as well as edema from multi-sequence cardiac MRI data, enabling accurate assessment of myocardial viability.
Incorporating both shape and topological priors into a unified latent representation improves the accuracy and anatomical consistency of automated liver vessel segmentation in medical images.
A novel deep learning approach that leverages scale-specific auxiliary tasks and contrastive learning to effectively capture the complex multi-scale geometry of the liver vascular tree.
A deep learning-powered toolbox, CartiMorph Toolbox (CMT), for automated quantification of knee cartilage shape and lesion from medical images.
This study presents an advanced approach to accurately segment lumbar spine structures, including vertebrae, spinal canal, and intervertebral discs, in MRI scans using deep learning techniques. The key innovations include a robust data preprocessing pipeline, a modified U-Net model with architectural enhancements, and a custom combined loss function to effectively handle class imbalance.
A novel clinical prior guided hierarchical vision-language pre-training framework, IMITATE, that aligns multi-level visual features from medical images with the descriptive and conclusive textual features from hierarchical medical reports, outperforming state-of-the-art methods across various medical imaging downstream tasks.
A novel self-supervised approach, MedSASS, that leverages the intrinsic properties of medical images to enhance binary semantic segmentation performance, outperforming existing state-of-the-art self-supervised techniques.