A5076735410- Advanced MRI Techniques and Applications
- Medical Imaging Techniques and Applications
- Generative Adversarial Networks and Image Synthesis
- Advanced Neuroimaging Techniques and Applications
- Image and Signal Denoising Methods
- Advanced Image Processing Techniques
- Radiomics and Machine Learning in Medical Imaging
- Sparse and Compressive Sensing Techniques
- Atomic and Subatomic Physics Research
- Medical Image Segmentation Techniques
- AI in cancer detection
- Cell Image Analysis Techniques
- Image Processing Techniques and Applications
- Cardiac Imaging and Diagnostics
- Advanced Image and Video Retrieval Techniques
- 3D Shape Modeling and Analysis
- Nuclear Physics and Applications
- Image Retrieval and Classification Techniques
- Domain Adaptation and Few-Shot Learning
- Infrared Thermography in Medicine
- Image Processing and 3D Reconstruction
- Advanced X-ray and CT Imaging
- Computer Graphics and Visualization Techniques
- NMR spectroscopy and applications
- Brain Tumor Detection and Classification
Stanford University
2022-2025
Bilkent University
2018-2022
Palo Alto University
2022
Acquiring images of the same anatomy with multiple different contrasts increases diversity diagnostic information available in an MR exam. Yet, scan time limitations may prohibit acquisition certain contrasts, and some be corrupted by noise artifacts. In such cases, ability to synthesize unacquired or can improve utility. For multi-contrast synthesis, current methods learn a nonlinear intensity transformation between source target images, either via regression deterministic neural networks....
Generative adversarial models with convolutional neural network (CNN) backbones have recently been established as state-of-the-art in numerous medical image synthesis tasks. However, CNNs are designed to perform local processing compact filters, and this inductive bias compromises learning of contextual features. Here, we propose a novel generative approach for synthesis, ResViT, that leverages the sensitivity vision transformers along precision convolution operators realism learning....
Supervised reconstruction models are characteristically trained on matched pairs of undersampled and fully-sampled data to capture an MRI prior, along with supervision regarding the imaging operator enforce consistency. To reduce requirements, recent deep image prior framework instead conjoins untrained priors during inference. Yet, canonical convolutional architectures suboptimal in capturing long-range relationships, based randomly initialized networks may yield performance. address these...
Multi-contrast MRI acquisitions of an anatomy enrich the magnitude information available for diagnosis. Yet, excessive scan times associated with additional contrasts may be a limiting factor. Two mainstream frameworks enhanced efficiency are reconstruction undersampled and synthesis missing acquisitions. Recently, deep learning methods have enabled significant performance improvements in both frameworks. decreases towards higher acceleration factors diminished sampling density at...
In recent years, deep learning models comprising transformer components have pushed the performance envelope in medical image synthesis tasks. Contrary to convolutional neural networks (CNNs) that use static, local filters, transformers self-attention mechanisms permit adaptive, non-local filtering sensitively capture long-range context. However, this sensitivity comes at expense of substantial model complexity, which can compromise efficacy particularly on relatively modest-sized imaging...
Abstract Purpose To develop a framework that jointly estimates rigid motion and polarizing magnetic field (B 0 ) perturbations () for brain MRI using single navigator of few milliseconds in duration, to additionally allow acquisition at arbitrary timings within any type sequence obtain high‐temporal resolution estimates. Theory Methods exist match data low‐resolution single‐contrast image (scout) estimate either or . In this work, called QUEEN (QUantitatively Enhanced parameter Estimation...
Learning-based translation between MRI contrasts involves supervised deep models trained using high-quality source- and target-contrast images derived from fully-sampled acquisitions, which might be difficult to collect under limitations on scan costs or time. To facilitate curation of training sets, here we introduce the first semi-supervised model for contrast (ssGAN) that can directly undersampled k-space data. enable learning data, ssGAN introduces novel multi-coil losses in image,...
Magnetic resonance imaging (MRI) is a common and life-saving medical technique. However, acquiring high signal-to-noise ratio MRI scans requires long scan times, resulting in increased costs patient discomfort, decreased throughput. Thus, there great interest denoising scans, especially for the subtype of diffusion that are severely SNR-limited. While most prior methods supervised nature, training datasets multitude anatomies, scanners, parameters proves impractical. Here, we propose...
Abstract Object Spatio-temporal MRI methods offer rapid whole-brain multi-parametric mapping, yet they are often hindered by prolonged reconstruction times or prohibitively burdensome hardware requirements. The aim of this project is to reduce time using deep learning. Materials and This study focuses on accelerating the volumetric multi-axis spiral projection MRF, aiming for T1 T2 while ensuring a streamlined approach compatible with clinical To optimize time, traditional method first...
Abstract Purpose This study aims to develop a high‐efficiency and high‐resolution 3D imaging approach for simultaneous mapping of multiple key tissue parameters routine brain imaging, including T 1 , 2 proton density (PD), ADC, fractional anisotropy (FA). The proposed method is intended pushing clinical from weighted quantitative can also be particularly useful diffusion‐relaxometry studies, which typically suffer lengthy acquisition time. Methods To address challenges associated with...
Abstract Purpose To develop a 3D spherical EPTI (sEPTI) acquisition and comprehensive reconstruction pipeline for rapid high‐quality whole‐brain submillimeter QSM quantification. Methods For the sEPTI acquisition, k‐space coverage is utilized with variable echo‐spacing maximum k x ramp‐sampling to improve efficiency signal incoherency compared existing approaches. reconstruction, an iterative rank‐shrinking B 0 estimation odd‐even high‐order phase correction algorithms were incorporated into...
Multi-contrast MRI acquisitions of an anatomy enrich the magnitude information available for diagnosis. Yet, excessive scan times associated with additional contrasts may be a limiting factor. Two mainstream approaches enhanced efficiency are reconstruction undersampled and synthesis missing acquisitions. In reconstruction, performance decreases towards higher acceleration factors diminished sampling density particularly at high-spatial-frequencies. synthesis, absence data samples from...
Multi-contrast MRI protocols increase the level of morphological information available for diagnosis. Yet, number and quality contrasts is limited in practice by various factors including scan time patient motion. Synthesis missing or corrupted can alleviate this limitation to improve clinical utility. Common approaches multi-contrast involve either one-to-one many-to-one synthesis methods. One-to-one methods take as input a single source contrast, they learn latent representation sensitive...
Magnetic resonance imaging (MRI) offers the flexibility to image a given anatomic volume under multitude of tissue contrasts. Yet, scan time considerations put stringent limits on quality and diversity MRI data. The gold-standard approach alleviate this limitation is recover high-quality images from data undersampled across various dimensions, most commonly Fourier domain or contrast sets. A primary distinction among recovery methods whether anatomy processed per cross-section. Volumetric...
Learning-based synthetic multi-contrast MRI commonly involves deep models trained using high-quality images of source and target contrasts, regardless whether domain samples are paired or unpaired. This results in undesirable reliance on fully-sampled acquisitions all which might prove impractical due to limitations scan costs time. Here, we propose a novel semi-supervised generative model that instead learns recover directly from accelerated contrasts. To achieve this, the proposed...
Recent developments in spatiotemporal MRI techniques enable whole-brain multi-parametric mapping incredibly short acquisition times through highly-efficient k-space encoding, subspace reconstruction and carefully-designed regularization. However, this comes at the cost of long making such methods difficult to integrate into clinical practice. This abstract proposes a framework denoted SMILR (pronounced smile-r) reduce from multiple hours few minutes machine learning. To evaluate performance,...