A5089226438- Radiomics and Machine Learning in Medical Imaging
- Glioma Diagnosis and Treatment
- Medical Imaging Techniques and Applications
- MRI in cancer diagnosis
- Advanced MRI Techniques and Applications
- Brain Tumor Detection and Classification
- AI in cancer detection
- Cancer, Lipids, and Metabolism
- Ultrasound and Hyperthermia Applications
- Infectious Encephalopathies and Encephalitis
- Advanced Neural Network Applications
- Ultrasound Imaging and Elastography
- Criminal Justice and Corrections Analysis
- Advanced Electron Microscopy Techniques and Applications
- Neonatal and fetal brain pathology
- Medical Image Segmentation Techniques
- Cell Image Analysis Techniques
- Computational Drug Discovery Methods
University of California System
2023-2024
University of California, San Francisco
2021-2024
Graduate Theological Union
2023-2024
University of California, Berkeley
2023-2024
Abstract Hypoxic‐ischemic encephalopathy (HIE) is a common neurological syndrome in newborns with high mortality and morbidity. Therapeutic hypothermia (TH), which standard of care for HIE, mitigates brain injury by suppressing anaerobic metabolism. However, more than 40% HIE neonates have poor outcome, even after TH. This study aims to provide metabolic biomarkers predicting the outcomes hypoxia‐ischemia (HI) TH using hyperpolarized [1‐ 13 C] pyruvate magnetic resonance spectroscopy....
Although fully automated volumetric approaches for monitoring brain tumor response have many advantages, most available deep learning models are optimized highly curated, multi-contrast MRI from newly diagnosed gliomas, which not representative of post-treatment cases in the clinic. Improving segmentation treated patients is critical to accurately tracking changes therapy. We investigated mixing data (
Prior characterization of treatment-effect and tumor recurrence using deep learning approaches have not optimized for spatial classification within a single lesion, which could improve surgical planning treatment. 10mm patches pre-surgical anatomical physiological images surrounding the locations histopathologically-confirmed tissue samples were used to train our models. Including images, pretraining on unlabeled data in an autoencoding task, training with alternative cross-validation...
Using pre-radiotherapy anatomical, diffusion, and metabolic MRI from 42 patients newly-diagnosed with GBM, we first used Random Forest models to identify voxels that later exhibit either contrast-enhancing or T2 lesion progression. We then applied convolutional encoder-decoder neural networks imaging segment subsequent tumor progression found the resulting predicted region better covered actual while sparing normal brain compared standard uniform 2cm expansion of anatomical define radiation...
Abstract INTRODUCTION Although AI-based approaches have been applied to discriminate between tumor recurrence (rTumor) and treatment-induced effects (TxE) from chemoradiotherapy in patients with recurrent glioma, these models typically either do not account for within-lesion mixtures of rTumor TxE or generalize independent test-sets >80% accuracy. We hypothesize that incorporating imaging features normal-appearing brain (NAB) into existing leverage tissue-samples known locations on...
Abstract BACKGROUND Machine-learning models have demonstrated great promise in predicting tumor infiltration beyond anatomical margins glioblastoma. Yet standard-of-care (SOC) radiation therapy (RT) planning only utilizes a uniform, isotropic 2cm-expansion of the T1-post-contrast-lesion or T2-lesion to generate clinical target volume (CTV), without considering heterogeneity infiltration. We hypothesize that using novel deep-learning approach predict regions progression with metabolic and...
Motivation: Radiopathomic mapping of glioma could improve standard care by helping guide surgical resection and subsequent treatment. Most current methods for predicting tumor pathology using MRI neglect intra-tumoral heterogeneity. Goal(s): We aim to use multi-parametric deep learning spatially map treatment naïve glioma. Approach: utilized histopathologically analyzed tissue samples taken during with known coordinates on pre-surgical semi-supervised ensemble networks. Results: Our...
Motivation: Noninvasive identification of malignant regions in glioma can help guide diagnosis and subsequent treatment planning. Goal(s): This study aims to create models predict elucidate limitations radiopathomic mapping invasiveness using multiparametric physiologic metabolic MRI. Approach: A large, unique MRI dataset with tissue is leveraged compare various machine learning %ki-67 cellularity (cells/mm2). Results: : The best binary model achieved a CV-AUC =0.82 = 0.75 for binarized...
Background: T2-weighted Single Shot Fast Spin Echo (SSFSE) scans at 3 Tesla (3T) are increasingly used to image fetal pathology due their excellent tissue contrast resolution and signal-to-noise ratio (SNR). Temperature changes that may occur in response radio frequency (RF) pulses for these sequences 3T have not been studied human brains. To evaluate the safety of SSFSE brains 3T, magnetic resonance (MR) thermometry was measure relative temperature a typical clinical brain MR exam. Methods:...
Using spectrum obtained at the spatial location of 549 tissue samples from 261 newly diagnosed patients with glioma, we trained and tested an support vector regression (SVR) model on individual metabolites, a 1D-CNN whole spectrum, to predict tumor biology such as cellularity, Ki-67, aggressiveness. A based using entire pre-trained similar classification task outperformed SVR metabolite peak heights.
Treatment-induced effects can mimic tumor recurrence and pose a challenge to accurately assessing treatment response. We aim provide machine learning framework identify important imaging features for discriminating treatment-induced injury from recurrent glioblastoma at biopsy level resolution. Our best model performs with mean AUC of .77+/-0.11 across 4 fold cross-validation 108 tissue samples. rCBV, choline-to-NAA index (CNI), normalized lipid levels were the top three most import in...
Abstract INTRODUCTION Spatial heterogeneity in the glioma microenvironment is difficult to capture through singular biopsy samples. Identification of malignant tumor regions can help guide diagnosis and treatment planning. This study compares radiopathomic maps from 2 strategies derived anatomical diffusion-weighted MRI: 1) a support vector machine model trained on an MRI dataset with tissue samples known spatial coordinates generate probability KI-67 cellularity, 2) bagged random-forest 40+...
Abstract INTRODUCTION Standard-of-care (SOC) radiation therapy (RT) planning only utilizes a fairly 1.5-2cm uniform expansion of T2-weighted-FLAIR MRI lesion to generate clinical target volume (2cm-CTV), without considering the spatial heterogeneity and infiltrative nature glioblastomas. This study aimed use multi-parametric with 2 artificial intelligence (AI)-based approaches predict regions subsequent tumor progression compare resulting predictions 2cm-CTV, hypothesis that applying deep...
Abstract INTRODUCTION Discriminating between glioma tumor recurrence (rTumor) and treatment-induced effects (TxE) from chemo radiotherapy using MRI is an ongoing challenge in neuro-oncology. Although deep learning has been used to characterize suspicious lesions, these models typically do not account for lesions that are comprised of mixtures rTumor TxE components. We hypothesize approach includes tissue samples with known location on MRI, incorporating perfusion-weighted diffusion-weighted...
Abstract Although physiologic (diffusion-weighted and perfusion-weighted) MRI has shown promise in identifying regions of recurrent tumor (rTumor) patients with glioblastoma suspected progression, distinguishing treatment-induced effects (TxE) from rTumor on anatomical remains a challenge. Whereas prior larger-scale machine learning (ML)-based studies mostly utilize imaging alone and/or perform lesion-level predictions, this study aimed to develop non-invasive, radiopathomic tool for...
Abstract INTRODUCTION Noninvasive, radiopathomic mapping of tumor aggressiveness can benefit patients with glioma by guiding the selection tissue samples for diagnosis, increasing extent resection, and non-invasively characterizing residual burden subsequent treatment. Although prior studies have demonstrated utility metabolic metrics quantified from 1H-MR Spectroscopy (MRS) in probing pathology, this study evaluated using entire 1D-spectrum deep learning intratumoral cellularity,...
Abstract INTRODUCTION Pathologically aggressive tumor biology can extend beyond the contrast-enhancing or non-enhancing anatomical lesions in patients with glioma. Identification of malignant regions help guide diagnosis and subsequent treatment planning. This study leverages a unique multi-parametric MRI dataset tissue samples known spatial coordinates to noninvasively predict cellular proliferation (KI-67) novel index aggressiveness (TAI), that combines proliferation, cellularity,...
Abstract Although current advances for automated glioma lesion segmentation and volumetric measurements using deep learning have yielded high performance on newly-diagnosed patients, response assessment in neuro-oncology still relies manually-drawn, cross-sectional areas of the tumor because these models do not generalize to patients post-treatment setting, where they are most needed clinic. Surgical resections, adjuvant treatment, or disease progression can alter characteristics lesions...