A5051753610- Medical Image Segmentation Techniques
- Advanced Radiotherapy Techniques
- Medical Imaging Techniques and Applications
- Medical Imaging and Analysis
- Acute Ischemic Stroke Management
- Functional Brain Connectivity Studies
- AI in cancer detection
- Generative Adversarial Networks and Image Synthesis
- Machine Learning in Healthcare
- Radiomics and Machine Learning in Medical Imaging
- Brain Tumor Detection and Classification
- Advanced Neuroimaging Techniques and Applications
- Cerebrovascular and Carotid Artery Diseases
- Lung Cancer Diagnosis and Treatment
- Surgical Simulation and Training
- Radiation Therapy and Dosimetry
- Face recognition and analysis
- Parkinson's Disease Mechanisms and Treatments
- Artificial Intelligence in Healthcare and Education
- Image Retrieval and Classification Techniques
- COVID-19 diagnosis using AI
- Advanced Neural Network Applications
- Health, Environment, Cognitive Aging
- Explainable Artificial Intelligence (XAI)
- Advanced X-ray and CT Imaging
Alberta Children's Hospital
2020-2025
University of Calgary
2020-2025
Alberta Children's Hospital Research Institute
2022-2024
Ontario Brain Institute
2024
Pediatrics and Genetics
2024
Marien Hospital Wesel
2020
Marienhospital Bottrop
2020
University of Lübeck
2011-2018
Parkinson's disease (PD) is a severe neurodegenerative that affects millions of people. Early diagnosis important to facilitate prompt interventions slow down progression. However, accurate PD can be challenging, especially in the early stages. The aim this work was develop and evaluate robust explainable deep learning model for classification trained from one largest collections T1-weighted magnetic resonance imaging datasets.A total 2,041 MRI datasets 13 different studies were collected,...
Abstract Biological brain age predicted using machine learning models based on high‐resolution imaging data has been suggested as a potential biomarker for neurological and cerebrovascular diseases. In this work, we aimed to develop deep predict the biological structural magnetic resonance angiography datasets from large database of 2074 adults (21–81 years). Since different modalities can provide complementary information, combining them might allow identify more complex aging patterns,...
Parkinson's disease (PD) is the second most common neurodegenerative disease. Accurate PD diagnosis crucial for effective treatment and prognosis but can be challenging, especially at early stages. This study aimed to develop evaluate an explainable deep learning model classification from multimodal neuroimaging data. The was trained using one of largest collections T1-weighted diffusion-tensor magnetic resonance imaging (MRI) datasets. A total 1264 datasets eight different studies were...
Purpose: Explainability and fairness are two key factors for the effective ethical clinical implementation of deep learning-based machine learning models in healthcare settings. However, there has been limited work on investigating how unfair performance manifests explainable artificial intelligence (XAI) methods, XAI can be used to investigate potential reasons unfairness. Thus, aim this was analyze effects previously established sociodemographic-related confounders classifier...
Distributed learning is a promising alternative to central for machine (ML) model training, overcoming data-sharing problems in healthcare. Previous studies exploring federated (FL) or the traveling (TM) setup medical image-based disease classification often relied on large databases with limited number of centers simulated artificial centers, raising doubts about real-world applicability. This study develops and evaluates convolution neural network (CNN) Parkinson's using data acquired by...
Abstract Objective Artificial intelligence (AI) models trained using medical images for clinical tasks often exhibit bias in the form of subgroup performance disparities. However, since not all sources real-world imaging data are easily identifiable, it is challenging to comprehensively assess their impacts. In this article, we introduce an analysis framework systematically and objectively investigating impact biases on AI models. Materials Methods Our utilizes synthetic neuroimages with...
Sharing multicenter imaging datasets can be advantageous to increase data diversity and size but may lead spurious correlations between site-related biological non-biological image features target labels, which machine learning (ML) models exploit as shortcuts. To date, studies analyzing how if deep use such effects a shortcut are scarce. Thus, the aim of this work was investigate encoded in feature space an established model designed for Parkinson's disease (PD) classification based on...
Distributed learning enables collaborative machine model training without requiring cross-institutional data sharing, thereby addressing privacy concerns. However, local quality control variability can negatively impact performance while systematic human visual inspection is time-consuming and may violate the goal of keeping inaccessible outside acquisition centers. This work proposes a novel self-supervised method to identify eliminate harmful during distributed fully-automatically. Harmful...
Artificial neural networks (ANNs) were originally modeled after their biological counterparts, but have since conceptually diverged in many ways. The resulting network architectures are not well understood, and furthermore, we lack the quantitative tools to characterize structures. Network science provides an ideal mathematical framework with which systems of interacting components, has transformed our understanding across domains, including mammalian brain. Yet, little been done bring ANNs....
Objective: Intra-interventional respiratory motion estimation is becoming a vital component in modern radiation therapy delivery or high intensity focused ultrasound systems. The treatment quality could tremendously benefit from more accurate dose using real-time tracking based on magnetic-resonance (MR) (US) imaging techniques. However, current practice often relies indirect measurements of external breathing indicators, which has an inherently limited accuracy. In this work, we present new...
3D facial landmarks are known to be diagnostically relevant biometrics for many genetic syndromes. The objective of this study was extend a state-of-the-art image-based 2D landmarking algorithm the challenging task landmark identification on subjects with syndromes, who often have moderate severe dysmorphia. automatic presented here uses detection and models identify 12 surface scans. evaluated using test set 444 scans ground truth identified by two different human observers. Three hundred...
Background Lesion-symptom mapping (LSM) is a statistical technique to investigate the population-specific relationship between structural integrity and post-stroke clinical outcome. In practice, patients are commonly evaluated using National Institutes of Health Stroke Scale (NIHSS), an 11-domain score quantitate neurological deficits due stroke. So far, LSM studies have mostly used total NIHSS for analysis, which might not uncover subtle structure–function relationships associated with...
Many machine learning tasks in neuroimaging aim at modeling complex relationships between a brain's morphology as seen structural MR images and clinical scores variables of interest. A frequently modeled process is healthy brain aging for which many image-based age estimation or age-conditioned template generation approaches exist. While regression task, related to generative modeling. Both can be inverse directions the same relationship age. However, this view rarely exploited most existing...
This work investigates if deep learning (DL) models can classify originating site locations directly from magnetic resonance imaging (MRI) scans with and without correction for intensity differences.
Automatic segmentation of ischemic stroke lesions in magnetic resonance (MR) images is important clinical practice and for neuroscientific trials. The key problem to detect largely inhomogeneous regions varying sizes, shapes locations. We present a lesion method based on local features extracted from multi-spectral MR data that are selected model human observer's discrimination criteria. A support vector machine classifier trained expert-segmented examples then used classify formerly unseen...
Breathing-induced location uncertainties of internal structures are still a relevant issue in the radiation therapy thoracic and abdominal tumours. Motion compensation approaches like gating or tumour tracking usually driven by low-dimensional breathing signals, which acquired real-time during treatment. These signals only surrogates motion target organs at risk, and, consequently, appropriate models needed to establish correspondence between sought patterns. In this work, we present...