Rachit Saluja
A5048402168- Brain Tumor Detection and Classification
- Radiomics and Machine Learning in Medical Imaging
- Glioma Diagnosis and Treatment
- Brain Metastases and Treatment
- Medical Imaging and Analysis
- Medical Imaging Techniques and Applications
- Meningioma and schwannoma management
- Explainable Artificial Intelligence (XAI)
- Automotive and Human Injury Biomechanics
- Advanced Neural Network Applications
- Infant Development and Preterm Care
- Machine Learning in Healthcare
- Cerebral Palsy and Movement Disorders
- Infant Health and Development
- Medical Image Segmentation Techniques
- Child Development and Digital Technology
- Neonatal and fetal brain pathology
- Children's Physical and Motor Development
- Image and Signal Denoising Methods
- Scientific Computing and Data Management
- Human Pose and Action Recognition
- Machine Learning and Data Classification
- Traumatic Brain Injury and Neurovascular Disturbances
- Prosthetics and Rehabilitation Robotics
- Digital Holography and Microscopy
Cornell University
2024-2025
University of California, San Francisco
2022-2024
Weill Cornell Medicine
2024
Hospital of the University of Pennsylvania
2022
California University of Pennsylvania
2021
University of Pennsylvania
2019-2020
Indian Institute of Space Science and Technology
2017
PES University
2016
An infant's risk of developing neuromotor impairment is primarily assessed through visual examination by specialized clinicians. Therefore, many infants at for go undetected, particularly in under-resourced environments. There thus a need to develop automated, clinical assessments based on quantitative measures from widely-available sources, such as videos recorded mobile device. Here, we automatically extract body poses and movement kinematics the at-risk (N = 19). For each infant,...
Radiosurgery(UCSF-BMSR) dataset is a public, clinical, multimodal brain MRI consisting of 560 MRIs from 412 patients with expert annotations 5136 metastases.Data consists registered and skull stripped T1 post-contrast, pre-contrast, FLAIR subtraction (T1 precontrast -T1 post-contrast) images voxelwise segmentations enhancing metastases in NifTI format.The also includes patient demographics, surgical status primary cancer types.The UCSF-BSMR has been made publicly available the hopes that...
We describe the design and results from BraTS 2023 Intracranial Meningioma Segmentation Challenge. The Challenge differed prior Glioma challenges in that it focused on meningiomas, which are typically benign extra-axial tumors with diverse radiologic anatomical presentation a propensity for multiplicity. Nine participating teams each developed deep-learning automated segmentation models using image data largest multi-institutional systematically expert annotated multilabel multi-sequence...
Over the last decade, computer science has made progress towards extracting body pose from single camera photographs or videos. This promises to enable movement detect disease, quantify performance, and take out of lab into real world. However, current tracking algorithms fall short needs science; types data that matter are poorly estimated. For instance, metrics currently used for evaluating use noisy hand-labeled ground truth do not prioritize precision relevant variables like...
Clinical monitoring of metastatic disease to the brain can be a laborious and time-consuming process, especially in cases involving multiple metastases when assessment is performed manually. The Response Assessment Neuro-Oncology Brain Metastases (RANO-BM) guideline, which utilizes unidimensional longest diameter, commonly used clinical research settings evaluate response therapy patients with metastases. However, accurate volumetric lesion surrounding peri-lesional edema holds significant...
Interpretability for machine learning models in medical imaging (MLMI) is an important direction of research. However, there a general sense murkiness what interpretability means. Why does the need MLMI arise? What goals one actually seek to address when needed? To answer these questions, we identify formalize and elements MLMI. By reasoning about real-world tasks common both image analysis its intersection with learning, five core interpretability: localization, visual recognizability,...
An important challenge in segmenting real-world biomedical imaging data is the presence of multiple disease processes within individual subjects. Most adults above age 60 exhibit a variable degree small vessel ischemic disease, as well chronic infarcts, which will manifest white matter hyperintensities (WMH) on brain MRIs. Subjects diagnosed with gliomas also typically some abnormal T2 signal due to WMH, rather than just tumor. We sought develop fully-automated algorithm distinguish and...
Abstract Meningiomas are the most common primary intracranial tumors and can be associated with significant morbidity mortality. Radiologists, neurosurgeons, neuro-oncologists, radiation oncologists rely on brain MRI for diagnosis, treatment planning, longitudinal monitoring. However, automated, objective, quantitative tools non-invasive assessment of meningiomas multi-sequence MR images not available. Here we present BraTS Pre-operative Meningioma Dataset, as largest multi-institutional...
We describe the design and results from BraTS 2023 Intracranial Meningioma Segmentation Challenge. The Challenge differed prior Glioma challenges in that it focused on meningiomas, which are typically benign extra-axial tumors with diverse radiologic anatomical presentation a propensity for multiplicity. Nine participating teams each developed deep-learning automated segmentation models using image data largest multi-institutional systematically expert annotated multilabel multi-sequence...
Delineation and quantification of normal abnormal brain tissues on Magnetic Resonance Images is fundamental to the diagnosis longitudinal assessment neurological diseases. Here we sought develop a convolutional neural network for automated multiclass tissue segmentation MRIs that was robust at typical clinical resolutions in presence variety lesions. We trained 3D U-Net full from prior atlas-based method an internal dataset consisted 558 T1-weighted (453/52/53; training/validation/test)...
Neural networks were trained for segmentation and longitudinal assessment of posttreatment diffuse glioma. A retrospective cohort (from January 2018 to December 2019) 298 patients with glioma (mean age, 52 years ± 14 [SD]; 177 men; 152 glioblastoma, 72 astrocytoma, 74 oligodendroglioma) who underwent two consecutive multimodal MRI examinations randomly selected into training (n = 198) testing 100) samples. tumor three-dimensional nnU-Net convolutional neural network multichannel inputs (T1,...
Gliomas are the most common malignant primary brain tumors in adults and one of deadliest types cancer. There many challenges treatment monitoring due to genetic diversity high intrinsic heterogeneity appearance, shape, histology, response. Treatments include surgery, radiation, systemic therapies, with magnetic resonance imaging (MRI) playing a key role planning post-treatment longitudinal assessment. The 2024 Brain Tumor Segmentation (BraTS) challenge on glioma MRI will provide community...
Abstract An infant’s risk of developing neuromotor impairment is primarily assessed through visual examination by specialized clinicians. Therefore, many infants at for go undetected, particularly in under-resourced environments. There thus a need to develop automated, clinical assessments based on quantitative measures from widely-available sources, such as video cameras. Here, we automatically extract body poses and movement kinematics the videos at-risk (N=19). For each infant, calculate...
We present a keypoint-based foundation model for general purpose brain MRI registration, based on the recently-proposed KeyMorph framework. Our model, called BrainMorph, serves as tool that supports multi-modal, pairwise, and scalable groupwise registration. BrainMorph is trained massive dataset of over 100,000 3D volumes, skull-stripped non-skull-stripped, from nearly 16,000 unique healthy diseased subjects. robust to large misalignments, interpretable via interrogating...
The 2024 Brain Tumor Segmentation Meningioma Radiotherapy (BraTS-MEN-RT) challenge aims to advance automated segmentation algorithms using the largest known multi-institutional dataset of radiotherapy planning brain MRIs with expert-annotated target labels for patients intact or post-operative meningioma that underwent either conventional external beam stereotactic radiosurgery. Each case includes a defaced 3D post-contrast T1-weighted MRI in its native acquisition space, accompanied by...
Abstract PURPOSE 1-Highlighting the Brain Tumor Segmentation (BraTS) challenge; past, present and future. 2- Emergence of ASNR-MICCAI BraTS-METS 2023 challenge with discussing our experience. 3-Discussing how to overcome it. 4- Highlighting administrative technical challenges for sharing data.4- Reviewing educational initiatives challenge. CLINICAL RELEVANCE/APPLICATION challenge: History BraTS evolution its opensource datasets, The components. workflow. How we used cumulative experience run...
Abstract PURPOSE Clinical monitoring of metastatic disease to the brain using magnetic resonance imaging (MRI) can be laborious and time-consuming, particularly when multiple small metastases are involved assessments performed manually. METHODS AND MATERIALS The BraTS-METS 2023 dataset is acquired from varying MRI quality across different vendors. scans pre-processed algorithms refined by a pool annotators with expertise. Two independent board-certified neuroradiologists finally reviewed...
We propose and demonstrate a compressive sensing (CS) framework for correlation holography. This is accomplished by adopting the principle of thresholding in two-point intensity correlation. The measurement matrix that are required applying CS here systematically extracted from random illuminations laser speckle data. Reconstruction results using CS, with thresholding, compared. Our study reveals liminal requires far fewer samples reconstruction hologram has wide application image reconstruction.
The idea behind Compressive Sensing(CS) is the reconstruction of sparse signals from very few samples, by means solving a convex optimization problem.In this paper we propose compressive sensing framework using Two-Step Iterative Shrinkage/Thresholding Algorithms(TwIST) for reconstructing speech signals.Further, compare with two other algorithms, l 1 Magic and Gradient Projection Sparse Reconstruction(GPSR).The performance our demonstrated via simulations exhibits faster convergence rate...
The University of California San Francisco Brain Metastases Stereotactic Radiosurgery (UCSF-BMSR) dataset is a public, clinical, multimodal brain MRI consisting 560 MRIs from 412 patients with expert annotations 5136 metastases. Data consists registered and skull stripped T1 post-contrast, pre-contrast, FLAIR subtraction (T1 pre-contrast - post-contrast) images voxelwise segmentations enhancing metastases in NifTI format. also includes patient demographics, surgical status primary cancer...
Interpretability for machine learning models in medical imaging (MLMI) is an important direction of research. However, there a general sense murkiness what interpretability means. Why does the need MLMI arise? What goals one actually seek to address when needed? To answer these questions, we identify formalize and elements MLMI. By reasoning about real-world tasks common both image analysis its intersection with learning, five core interpretability: localization, visual recognizability,...