A5085138985- Radiomics and Machine Learning in Medical Imaging
- Advanced Neural Network Applications
- Glioma Diagnosis and Treatment
- Brain Tumor Detection and Classification
- AI in cancer detection
- Medical Imaging and Analysis
- Scoliosis diagnosis and treatment
- Sarcoma Diagnosis and Treatment
- Prostate Cancer Diagnosis and Treatment
- COVID-19 diagnosis using AI
- Hepatocellular Carcinoma Treatment and Prognosis
- Artificial Intelligence in Healthcare and Education
- Machine Learning and ELM
- Medical Image Segmentation Techniques
- Spinal Fractures and Fixation Techniques
- Domain Adaptation and Few-Shot Learning
- MRI in cancer diagnosis
- Pancreatic and Hepatic Oncology Research
- Retinal Imaging and Analysis
- Cholangiocarcinoma and Gallbladder Cancer Studies
- Digital Imaging for Blood Diseases
- Acute Ischemic Stroke Management
- Hospital Admissions and Outcomes
- Machine Learning in Healthcare
- Biosensors and Analytical Detection
SickKids Foundation
2020-2024
Mental Health Research Canada
2023-2024
University of Toronto
2020-2024
Vector Institute
2022-2024
Hospital for Sick Children
2021-2024
Canada Research Chairs
2024
Nvidia (United States)
2023
University College London
2021
Great Ormond Street Hospital
2021
Lunenfeld-Tanenbaum Research Institute
2020-2021
Brain tumor is one of the leading causes cancer-related death globally among children and adults. Precise classification brain grade (low-grade high-grade glioma) at an early stage plays a key role in successful prognosis treatment planning. With recent advances deep learning, artificial intelligence–enabled grading systems can assist radiologists interpretation medical images within seconds. The performance learning techniques is, however, highly depended on size annotated dataset. It...
<h3>BACKGROUND AND PURPOSE:</h3> <i>B-Raf proto-oncogene, serine/threonine kinase</i> (<i>BRAF</i>) status has important implications for prognosis and therapy of pediatric low-grade gliomas. Currently, <i>BRAF</i> classification relies on biopsy. Our aim was to train validate a radiomics approach predict fusion V600E mutation. <h3>MATERIALS METHODS:</h3> In this bi-institutional retrospective study, FLAIR MR imaging datasets 115 patients with gliomas from 2 children's hospitals acquired...
Characterizing complex biofluids using surface-enhanced Raman spectroscopy (SERS) coupled with machine learning (ML) has been proposed as a powerful tool for point-of-care detection of clinical disease. ML is well-suited to categorizing otherwise uninterpretable, patient-derived SERS spectra that contain multitude low concentration, disease-specific molecular biomarkers among dense spectral background biological molecules. However, can generate false, non-generalizable models when data sets...
Despite transarterial chemoembolization (TACE) for hepatocellular carcinoma (HCC), a significant number of patients will develop progression on the liver transplant (LT) waiting list or disease recurrence post-LT. We sought to evaluate feasibility pre-TACE radiomics model, an imaging-based tool predict these adverse outcomes.We analyzed computed tomography images LT. The primary endpoint was combined event that included waitlist dropout tumor radiomic features were extracted from largest HCC...
Receiver operating characteristic (ROC) curve is an informative tool in binary classification and Area Under ROC Curve (AUC) a popular metric for reporting performance of classifiers. In this paper, first we present comprehensive review AUC metric. Next, propose modified version that takes confidence the model into account at same time, incorporates Binary Cross Entropy (BCE) loss used training Convolutional neural Network tasks. We demonstrate on three datasets: MNIST, prostate MRI, brain...
Purpose: Scoliosis is a deformity of the spine, and as measure scoliosis severity, Cobb angle fundamental to diagnosis deformities that require treatment. Conventional measurement assessment usually done manually, which inherently time-consuming, associated with high inter- intra-observer variability. While there exist automatic methods, they suffer from insufficient accuracy. In this work, we propose two-step segmentation-based deep learning architecture automate for using X-Ray images....
Purpose: Biopsy-based assessment of H3 K27 M status helps in predicting survival, but biopsy is usually limited to unusual presentations and clinical trials. We aimed evaluate whether radiomics can serve as prognostic marker stratify diffuse intrinsic pontine glioma (DIPG) subsets. Methods: In this retrospective study, diagnostic brain MRIs children with DIPG were analyzed. Radiomic features extracted from tumor segmentations data split into training/testing sets (80:20). A conditional...
<h3>BACKGROUND AND PURPOSE:</h3> No qualitative imaging feature currently predicts molecular alterations of pediatric low-grade gliomas with high sensitivity or specificity. The T2-FLAIR mismatch sign <i>IDH</i>-mutated 1p19q noncodeleted adult We aimed to assess the significance in gliomas. <h3>MATERIALS METHODS:</h3> Pretreatment MR images acquired between January 2001 and August 2018 patients were retrospectively identified. Inclusion criteria following: 1) 0–18 years age, 2) availability...
Purpose: Scoliosis is a complex spine deformity with direct functional and cosmetic impacts on the individual. The reference standard for assessing scoliosis severity Cobb angle which measured radiographs by human specialists, carrying interobserver variability inaccuracy of measurements. These limitations may result in lack timely referral management at time scoliotic progression can be saved from surgery. We aimed to create machine learning (ML) model automatic calculation angles 3-foot...
Purpose: Pediatric low-grade gliomas (pLGG) are the most common brain tumour in children, and molecular diagnosis of pLGG enables targeted treatment. We use MRI-based Convolutional Neural Networks (CNNs) for subtype identification augment models using location probability maps. Materials Methods: MRI FLAIR sequences 214 patients (110 male, mean age 8.54 years, 143 BRAF fused 71 V600E mutated tumours) from January 2000 to December 2018 were included this retrospective REB-approved study....
In this work, we examine magnetic resonance imaging (MRI) and ultrasound (US) appointments at the Diagnostic Imaging (DI) department of a pediatric hospital to discover possible relationships between selected patient features no-show or long waiting room time endpoints. The chosen include age, sex, income, distance from hospital, percentage non-English speakers in postal code, single caregivers appointment slot (morning, afternoon, evening), day week (Monday Sunday). We trained univariate...
<h3>BACKGROUND AND PURPOSE:</h3> Molecular biomarker identification increasingly influences the treatment planning of pediatric low-grade neuroepithelial tumors (PLGNTs). We aimed to develop and validate a radiomics-based ADC signature predictive molecular status PLGNTs. <h3>MATERIALS METHODS:</h3> In this retrospective bi-institutional study, we searched PACS for baseline brain MRIs from children with Semiautomated tumor segmentation on maps was performed using semiautomated level tracing...
The use of targeted agents in the treatment pediatric low-grade gliomas (pLGGs) relies on determination molecular status. It has been shown that genetic alterations pLGG can be identified non-invasively using MRI-based radiomic features or convolutional neural networks (CNNs). We aimed to build and assess a combined radiomics CNN non-invasive status identification model. This retrospective study used tumor regions, manually segmented from T2-FLAIR MR images, 336 patients treated for between...
Abstract Purpose Training machine learning models to segment tumors and other anomalies in medical images is an important step for developing diagnostic tools but generally requires manually annotated ground truth segmentations, which necessitates significant time resources. We aim develop a pipeline that can be trained using readily accessible binary image-level classification labels, effectively regions of interest without requiring annotations. Methods This work proposes the use deep...
As deep learning is widely used in the radiology field, explainability of Artificial Intelligence (AI) models becoming increasingly essential to gain clinicians' trust when using for diagnosis. In this research, three experiment sets were conducted with a U-Net architecture improve disease classification performance while enhancing heatmaps corresponding model's focus through incorporating heatmap generators during training. All experiments dataset that contained chest radiographs,...
Convolutional Neural Networks (CNNs) have been used for automated detection of prostate cancer where Area Under Receiver Operating Characteristic (ROC) curve (AUC) is usually as the performance metric. Given that AUC not differentiable, common practice to train CNN using a loss functions based on other metrics such cross entropy and monitoring select best model. In this work, we propose fine-tune trained Genetic Algorithm achieve higher AUC. Our dataset contained 6-channel Diffusion-Weighted...
Abstract Machine learning (ML) approaches can predict BRAF status of pediatric low-grade gliomas (pLGG) on pre-therapeutic brain MRI. The impact training data sample size and type ML model is not established. In this bi-institutional retrospective study, 251 pLGG FLAIR MRI datasets from 2 children’s hospitals were included. Radiomics features extracted tumor segmentations five models (Random Forest, XGBoost, Neural Network (NN) 1 (100:20:2), NN2 (50:10:2), NN3 (50:20:10:2)) tested to...