A5033643177- Medical Image Segmentation Techniques
- Computer Graphics and Visualization Techniques
- Generative Adversarial Networks and Image Synthesis
- Advanced Image Fusion Techniques
- Hip disorders and treatments
- Anomaly Detection Techniques and Applications
- Radiomics and Machine Learning in Medical Imaging
- Total Knee Arthroplasty Outcomes
- Neural Networks and Applications
- 3D Shape Modeling and Analysis
- AI in cancer detection
- Functional Brain Connectivity Studies
- Brain Tumor Detection and Classification
- MRI in cancer diagnosis
- Lower Extremity Biomechanics and Pathologies
- Image Retrieval and Classification Techniques
- Multiple Sclerosis Research Studies
- Medical Imaging Techniques and Applications
- Adversarial Robustness in Machine Learning
- Gaussian Processes and Bayesian Inference
- Image and Signal Denoising Methods
- Bone and Joint Diseases
- Advanced MRI Techniques and Applications
- COVID-19 diagnosis using AI
University of Basel
2024-2025
Abstract Purpose: Trochlear dysplasia (TD) is a common malformation in adolescents, leading to anterior knee pain and instability. Surgical interventions such as trochleoplasty require precise planning correct the trochlear groove. However, no standardized preoperative plan exists guide surgeons reshaping femur. This study aims generate patient-specific, pseudo-healthy MR images of region that should theoretically align with respective patient’s patella, potentially supporting...
Denoising diffusion models have recently achieved state-of-the-art performance in many image-generation tasks. They do, however, require a large amount of computational resources. This limits their application to medical tasks, where we often deal with 3D volumes, like high-resolution three-dimensional data. In this work, present number different ways reduce the resource consumption for and apply them dataset images. The main contribution paper is memory-efficient patch-based model...
Monitoring diseases that affect the brain's structural integrity requires automated analysis of magnetic resonance (MR) images, e.g., for evaluation volumetric changes. However, many tools are optimized analyzing healthy tissue. To enable scans containing pathological tissue, it is therefore required to restore tissue in areas. In this work, we explore and extend denoising diffusion models consistent inpainting 3D brain We modify state-of-the-art 2D, pseudo-3D, methods working image space,...
Magnetic resonance (MR) images from multiple sources often show differences in image contrast related to acquisition settings or the used scanner type. For long-term studies, longitudinal comparability is essential but can be impaired by these differences, leading biased results when using automated evaluation tools. This study presents a diffusion model-based approach for harmonization. We use data set consisting of scans 18 Multiple Sclerosis patients and 22 healthy controls. Each subject...
This paper is a contribution to the "BraTS 2023 Local Synthesis of Healthy Brain Tissue via Inpainting Challenge". The task this challenge transform tumor tissue into healthy in brain magnetic resonance (MR) images. idea originates from problem that MR images can be evaluated using automatic processing tools, however, many these tools are optimized for analysis tissue. By solving given inpainting task, we enable featuring lesions, and further downstream tasks. Our approach builds on...
The high performance of denoising diffusion models for image generation has paved the way their application in unsupervised medical anomaly detection. As diffusion-based methods require a lot GPU memory and have long sampling times, we present novel fast detection approach based on latent Bernoulli models. We first apply an autoencoder to compress input images into binary representation. Next, model that follows noise schedule is employed this space trained restore representations from...
Due to the three-dimensional nature of CT- or MR-scans, generative modeling medical images is a particularly challenging task. Existing approaches mostly apply patch-wise, slice-wise, cascaded generation techniques fit high-dimensional data into limited GPU memory. However, these may introduce artifacts and potentially restrict model's applicability for certain downstream tasks. This work presents WDM, wavelet-based image synthesis framework that applies diffusion model on wavelet decomposed...
The human brain undergoes rapid development during the third trimester of pregnancy. In this work, we model neonatal infant in age range. As a basis, use MR images preterm- and term-birth neonates from developing connectome project (dHCP). We propose neural network, specifically an implicit representation (INR), to predict 2D- 3D varying time points. order subject-specific process, it is necessary disentangle subjects' identity latent space INR. two methods, Subject Specific Latent Vectors...
This paper contributes to the "BraTS 2024 Brain MR Image Synthesis Challenge" and presents a conditional Wavelet Diffusion Model (cWDM) for directly solving paired image-to-image translation task on high-resolution volumes. While deep learning-based brain tumor segmentation models have demonstrated clear clinical utility, they typically require scans from various modalities (T1, T1ce, T2, FLAIR) as input. However, due time constraints or imaging artifacts, some of these may be missing,...
Purpose: Trochlear Dysplasia (TD) is a common malformation in adolescents, leading to anterior knee pain and instability. Surgical interventions such as trochleoplasty require precise planning correct the trochlear groove. However, no standardized preoperative plan exists guide surgeons reshaping femur. This study aims generate patient-specific, pseudo-healthy MR images of region that should theoretically align with respective patient's patella, potentially supporting pre-operative...