A5054796268- Medical Image Segmentation Techniques
- Advanced Neuroimaging Techniques and Applications
- Retinal Imaging and Analysis
- Functional Brain Connectivity Studies
- Optical Coherence Tomography Applications
- Advanced MRI Techniques and Applications
- Domain Adaptation and Few-Shot Learning
- Neonatal and fetal brain pathology
- Brain Tumor Detection and Classification
- Advanced Neural Network Applications
- Medical Imaging Techniques and Applications
- Dementia and Cognitive Impairment Research
- Machine Learning in Healthcare
- Neural Networks and Applications
- Digital Imaging for Blood Diseases
- Image Retrieval and Classification Techniques
- Glaucoma and retinal disorders
- Retinal and Optic Conditions
- Advanced Graph Neural Networks
- 3D Shape Modeling and Analysis
- Visual Attention and Saliency Detection
- Retinal Diseases and Treatments
- Fetal and Pediatric Neurological Disorders
- Advanced Vision and Imaging
- COVID-19 diagnosis using AI
Athinoula A. Martinos Center for Biomedical Imaging
2023-2025
Harvard University
2023-2025
Massachusetts General Hospital
2023-2025
École de Technologie Supérieure
2018-2023
Indian Institute of Technology Hyderabad
2016-2019
International Institute of Information Technology, Hyderabad
2017-2018
The University of Texas Southwestern Medical Center
2009-2010
University of Florida
2003
Recently, dense connections have attracted substantial attention in computer vision because they facilitate gradient flow and implicit deep supervision during training. Particularly, DenseNet that connects each layer to every other a feed-forward fashion has shown impressive performances natural image classification tasks. We propose HyperDenseNet, 3-D fully convolutional neural network extends the definition of connectivity multi-modal segmentation problems. Each imaging modality path,...
Retinal swelling due to the accumulation of fluid is associated with most vision-threatening retinal diseases. Optical coherence tomography (OCT) current standard care in assessing presence and quantity image-guided treatment management. Deep learning methods have made their impact across medical imaging, many OCT analysis been proposed. However, it currently not clear how successful they are interpreting on OCT, which lack standardized benchmarks. To address this, we organized a challenge...
Automated and accurate segmentation of cystoid structures in optical coherence tomography (OCT) is interest the early detection retinal diseases. It is, however, a challenging task. We propose novel method for localizing cysts 3-D OCT volumes. The proposed work biologically inspired based on selective enhancement cysts, by inducing motion to given slice. A convolutional neural network designed learn mapping function that combines result multiple such motions produce probability map cyst...
Segmentation of retinal layers from Optical Coherence Tomography (OCT) volumes is a fundamental problem for any computer aided diagnostic algorithm development. This requires preprocessing steps such as denoising, region interest extraction, flattening and edge detection all which involve separate parameter tuning. In this paper, we explore deep learning techniques to automate these handle the presence/absence pathologies. A model proposed consisting combination Convolutional Neural Network...
Abstract The relationship between CSF biomarkers of Alzheimer’s disease and neurodegenerative effects is not fully understood. This study investigates neurodegeneration patterns across biomarker groups, the association brain volumes with amyloid tau status, sex differences in these relationships a clinical neurology sample. MRI data were analyzed 306 patients MassGeneral Brigham healthcare system aged 50+ (mean age = 68.4 ± 8.8 years; 43.1% female), who had lumbar punctures within one year...
Brain surface analysis is essential to neuroscience, however, the complex geometry of brain cortex hinders computational methods for this task. The difficulty arises from a discrepancy between 3D imaging data, which represented in Euclidean space, and non-Euclidean highly-convoluted surface. Recent advances machine learning have enabled use neural networks spaces. These facilitate yet pooling strategies often remain constrained single fixed-graph. This paper proposes new learnable graph...
Abstract The emerging field of geometric deep learning extends the application convolutional neural networks to irregular domains such as graphs, meshes and surfaces. Several recent studies have explored potential for using these techniques analyse segment cortical surface. However, there has been no comprehensive comparison approaches one another, nor existing Euclidean methods, date. This paper benchmarks a collection traditional models on phenotype prediction segmentation sphericalised...
Surface-based analysis of the cerebral cortex is ubiquitous in human neuroimaging with MRI. It crucial for cortical registration, parcellation, and thickness estimation. Traditionally, these analyses require high-resolution, isotropic scans good gray-white matter contrast, typically a 1mm T1-weighted scan. This excludes most clinical MRI scans, which are often anisotropic lack necessary T1 contrast. To enable large-scale studies using vast data, we introduce recon-all-clinical, novel method...
A variety of imaging modalities have been used for developing diagnostic aids glaucoma assessment. Structural such as colour fundus and optical coherence tomography (OCT) investigated automatically estimating key parameters assessment cup to disc diameter ratio thickness the retinal nerve fibre layer (RNFL). OCT-based angiography or OCTA is a new modality which provides structural an-giographic information about layers. We present method detection using images. Specifically, capillary...
The analysis of the brain surface modeled as a graph mesh is challenging task. Conventional deep learning approaches often rely on data lying in Euclidean space. As an extension to irregular graphs, convolution operations are defined Fourier or spectral domain. This domain obtained by decomposing Laplacian, which captures relevant shape information. However, decomposition across different graphs causes inconsistencies between eigenvectors individual domains, causing algorithm fail. Current...
Synthetic data have emerged as an attractive option for developing machine-learning methods in human neuroimaging, particularly magnetic resonance imaging (MRI)-a modality where image contrast depends enormously on acquisition hardware and parameters. This retrospective paper reviews a family of recently proposed methods, based synthetic data, generalizable machine learning brain MRI analysis. Central to this framework is the concept domain randomization, which involves training neural...
A simultaneous segmentation and registration model is proposed for functional MR (fMR) image alignment that significant minimizing motion artifacts on fMRI data analysis. Due to T/sub 2/* weighted signal loss decreased resolution, in fMR images, the images can't be aligned reliably by only using information. Our approach uses shape of a contour pre-segmented high resolution find unknown each time series spatial transform maps interface given contour. This achieved an energy depending...
Neuronal cell bodies mostly reside in the cerebral cortex. The study of this thin and highly convoluted surface is essential for understanding how brain works. analysis data is, however, challenging due to high variability cortical geometry. This paper presents a novel approach learning exploiting directly across domains. Current approaches rely on geometrical simplifications, such as spherical inflations, popular but costly process. For instance, widely used FreeSurfer takes about 3 hours...
3D imaging modalities are becoming increasingly popular and relevant in retinal owing to their effectiveness highlighting structures sub-retinal layers. OCT is one such modality which has great importance the context of analysis cystoid subretinal Signal noise ratio(SNR) images obtained from less hence automated accurate determination a challenging task. We propose an method for detecting/segmenting cysts volumes. The proposed biologically inspired fast aided by domain knowledge about...
This paper presents an atlas-based segmentation method for subcortical structures in magnetic resonance brain images. utilizes our previously introduced two-step registration comprising affine transformation applied to the entire area followed by a local nonrigid structures. is compared three existing methods using two objective indices, namely dice and relative error on volume. The IBSR database considered which expert segmented are available. results obtained show that proposed outperforms methods.