Browsing by Author "Forkert, Nils D."

Now showing 1 - 7 of 7
  • Thumbnail Image
    ItemOpen Access
    Age-dependent analysis of cerebral structures and arteries in a large database
    (2022-06-30) Mouches, Pauline; Forkert, Nils D.; Goodyear, Bradley G.; Josephson, Colin B.
    Aging of the population is expected to lead to a rapid increase of neurological diseases. Such diseases can progress quickly and detrimentally affect the daily life of patients. Prognosis improves with early diagnosis, but early detection is diffcult. It is crucial to be able to differentiate early stage pathological alteration from normal age-related changes. Thus, there is a need for a better understanding of brain aging and reliable biomarkers. Within that context, the overarching aim of this work is to study normal aging patterns in brain tissues and arteries using a large database of magnetic resonance imaging and angiography data, as well as cardiovascular risk factors from the whole adult life span. To do so, the objectives of this thesis are: (1) to quantify artery morphology variability among adults and identify the impact of age, sex and cardiovascular risk factors on cerebrovascular structures; (2) to combine brain tissue and artery information for biological brain age prediction; (3) to explore the impact of cardiovascular risk factors on the brain age gap, which is a biomarker representing the difference between the biological brain age and chronological age. To achieve these objectives, first, a statistical cerebrovascular atlas is generated from multi-centre adult data. Image analyses and multivariate regression methods are then employed to find associations between brain artery morphology and aging. Second, multi-modal explainable deep learning models are used to accurately estimate the biological brain age and identify predictive brain regions. Third, an exploratory causal analysis is performed to isolate the effects of individual factors on the brain age gap. The results of this work offer a novel insight on brain tissue and artery aging patterns. An in-depth analysis of the brain age gap biomarker is carried out. Novel approaches are proposed to improve brain age prediction models in terms of accuracy and explainability. Finally, innovative methods are used to study cause and effects relationships between brain aging and cardiovascular risk factors. This work aims to uncover clinically relevant findings and represents valuable methodological advancements that could be used in other neuroimaging clinical applications, for instance, to ameliorate predictive models for decision-support.
  • Thumbnail Image
    ItemOpen Access
    Automatic quantification of osteoarthritis features in MRI using deep learning methods
    (2023-03-20) Felfeliyan, Banafshe; Ronsky, Janet L.; Jaremko, Jacob L.; Lebel, Marc R.; Li, Matthew D.; Wong, Andy K.; Far, Behrouz; Forkert, Nils D.
    Osteoarthritis (OA) is a progressive irreversible disease that affects the whole joint involving various tissues including cartilage, bone, and synovium. Accurate OA diagnosis and management requires a thorough understanding of its pathogenesis, along with precise monitoring of its trajectories and changes associated with treatment. Magnetic resonance imaging (MRI) is an excellent tool for understanding OA progression and diagnosis. However, its assessment relies on reader-dependent semiquantitative scores, which are subjective, laborious, and insensitive to small changes. An automated OA assessment pipeline using deep learning (DL) can overcome limitations but faces challenges, including low generalizability, label scarcity, and noisy clinical labels with high variability. Therefore, this thesis’s objective is to overcome challenges associated with automated OA assessment by developing DL methods for the quantitative assessment of predominant OA biomarkers (cartilage, effusion-synovitis, and bone marrow lesion (BML)). This thesis proposed ImprovedMaskRCNN (iMaskRCNN), an accurate DL algorithm for multiscale segmentation of tissues involved in OA (femur, tibia, femoral and tibial cartilages, and hip effusion), which improves Dice score by 1-7%. Algorithm outputs are used to quantify cartilage thickness and hip effusion volume. An unsupervised domain adaptation pipeline using iMaskRCNN and CycleGAN was proposed to overcome DL’s low generalizability across MRI sequences. Therefore, a weakly supervised training strategy was proposed for BML segmentation, using soft labels obtained from clinical scoring and a noise-robust loss, which increased the Dice score and recall by 8% and 22%, respectively. Effusion segmentation was achieved using a self-supervised learning approach to pretrain the iMaskRCNN algorithm for training with a few-labeled dataset and improved the Dice by more than 8%. Following validation of the developed methods, results were compared with patient outcomes indicating strong associations of the extracted BML and effusion features (Δpain vs. ΔDL-effusion r =0.59, and Δpain vs. Δtibial BML r=0.49). This research provides an efficient integrated tool for quantifying OA features from MRI to accurately monitor OA progression and structural changes. This tool enables large population analyses and the extraction of more sensitive measures for clinicians to understand OA. The proposed methods have strong potential to apply to other joints, tissues, and medical image analysis problems.
  • Thumbnail Image
    ItemOpen Access
    Computer-Assisted Diagnosis of Genetic Syndromes Using 3D Facial Surface Scans
    (2023-03-09) Bannister, Jordan J.; Forkert, Nils D.; Hallgrímsson, Benedikt; Lebel, Catherine; Bernier, Francois Paul J.
    Due to the complexity and rarity of genetic syndromes, one of the primary difficulties in treating afflicted patients is diagnosing their condition. Gene technologies have been a key tool to improve diagnosis rates, but genetic testing remains inaccurate, inaccessible, or expensive for many people. Computer-assisted facial phenotyping is a complementary strategy that makes use of inexpensive and widely available technologies. Many genetic syndromes are known to be associated with altered facial morphology, and clinical geneticists often make use of facial phenotype to inform diagnoses. The overarching objective of this research was to develop clinically useful image processing algorithms and machine learning models to improve computer-assisted facial phenotyping and syndrome diagnosis systems based on 3D facial surface images. First, a fully automated 3D facial landmarking algorithm was developed to prepare 3D facial surface scans for analysis without manual labor. Next, analyses comparing different 2D and 3D facial representations were performed to determine an optimal facial image acquisition strategy. Machine learning models of 3D facial morphology were then developed to identify abnormal and characteristically syndromic faces. Additionally, an analysis of non-syndromic facial morphology was performed to present quantitative information about facial sex differences to facial surgeons. The main contributions of this thesis are the automated 3D scan processing methods and normalizing flow framework for 3D facial shape modelling that provide the computational methods needed to create a complete and highly interpretable 3D face-based computer aided diagnosis system. Additionally, results from the subject-matched analysis of 2D and 3D facial representations are the first to empirically suggest that using 3D facial imaging instead of 2D photography improves the performance of face-based syndrome diagnosis systems. Finally, the analysis performed for facial surgeons demonstrates that the methods developed in this thesis are applicable to medical domains other than computer-assisted diagnosis.
  • Thumbnail Image
    ItemOpen Access
    Dementia risk prediction in individuals with mild cognitive impairment: a comparison of Cox regression and machine learning models
    (2022-11-02) Wang, Meng; Greenberg, Matthew; Forkert, Nils D.; Chekouo, Thierry; Afriyie, Gabriel; Ismail, Zahinoor; Smith, Eric E.; Sajobi, Tolulope T.
    Abstract Background Cox proportional hazards regression models and machine learning models are widely used for predicting the risk of dementia. Existing comparisons of these models have mostly been based on empirical datasets and have yielded mixed results. This study examines the accuracy of various machine learning and of the Cox regression models for predicting time-to-event outcomes using Monte Carlo simulation in people with mild cognitive impairment (MCI). Methods The predictive accuracy of nine time-to-event regression and machine learning models were investigated. These models include Cox regression, penalized Cox regression (with Ridge, LASSO, and elastic net penalties), survival trees, random survival forests, survival support vector machines, artificial neural networks, and extreme gradient boosting. Simulation data were generated using study design and data characteristics of a clinical registry and a large community-based registry of patients with MCI. The predictive performance of these models was evaluated based on three-fold cross-validation via Harrell’s concordance index (c-index), integrated calibration index (ICI), and integrated brier score (IBS). Results Cox regression and machine learning model had comparable predictive accuracy across three different performance metrics and data-analytic conditions. The estimated c-index values for Cox regression, random survival forests, and extreme gradient boosting were 0.70, 0.69 and 0.70, respectively, when the data were generated from a Cox regression model in a large sample-size conditions. In contrast, the estimated c-index values for these models were 0.64, 0.64, and 0.65 when the data were generated from a random survival forest in a large sample size conditions. Both Cox regression and random survival forest had the lowest ICI values (0.12 for a large sample size and 0.18 for a small sample size) among all the investigated models regardless of sample size and data generating model. Conclusion Cox regression models have comparable, and sometimes better predictive performance, than more complex machine learning models. We recommend that the choice among these models should be guided by important considerations for research hypotheses, model interpretability, and type of data.
  • Thumbnail Image
    ItemOpen Access
    Distinct phenotypes of multisystem inflammatory syndrome in children: a cohort study
    (2023-04-12) Renson, Thomas; Forkert, Nils D.; Amador, Kimberly; Miettunen, Paivi; Parsons, Simon J.; Dhalla, Muhammed; Johnson, Nicole A.; Luca, Nadia; Schmeling, Heinrike; Stevenson, Rebeka; Twilt, Marinka; Hamiwka, Lorraine; Benseler, Susanne
    Abstract Background Multisystem inflammatory syndrome in children (MIS-C) is a severe disease with an unpredictable course and a substantial risk of cardiogenic shock. Our objectives were to (a) compare MIS-C phenotypes across the COVID-19 pandemic, (b) identify features associated with intensive care need and treatment with biologic agents. Methods Youth aged 0–18 years, fulfilling the World Health Organization case definition of MIS-C, and admitted to the Alberta Children’s Hospital during the first four waves of the COVID-19 pandemic (May 2020-December 2021) were included in this cohort study. Demographic, clinical, biochemical, imaging, and treatment data were captured. Results Fifty-seven MIS-C patients (median age 6 years, range 0–17) were included. Thirty patients (53%) required intensive care. Patients in the third or fourth wave (indicated as phase 2 of the pandemic) presented with higher peak ferritin (µg/l, median (IQR) = 1134 (409–1806) vs. 370 (249–629), P = 0.001), NT-proBNP (ng/l, median (IQR) = 12,217 (3013–27,161) vs. 3213 (1216–8483), P = 0.02) and D-dimer (mg/l, median (IQR) = 4.81 (2.24–5.37) vs. 2.01 (1.27–3.34), P = 0.004) levels, and higher prevalence of liver enzyme abnormalities (n(%) = 17 (68) vs. 11 (34), P = 0.02), hypoalbuminemia (n(%) = 24 (100) vs. 25 (81), P = 0.03) and thrombocytopenia (n(%) 18 (72) vs. 11 (34), P = 0.007) compared to patients in the first two waves (phase 1). These patients had a higher need of non-invasive/mechanical ventilation (n(%) 4 (16) vs. 0 (0), P = 0.03). Unsupervised clustering analyses classified 47% of the patients in the correct wave and 74% in the correct phase of the pandemic. NT-proBNP was the only significant contributor to the need for intensive care in all applied multivariate regression models. Treatment with biologic agents was significantly associated with peak CRP (mg/l (median, IQR = 240.9 (132.9-319.4) vs. 155.8 (101.0-200.7), P = 0.02) and ferritin levels (µg/l, median (IQR) = 1380 (509–1753) vs. 473 (280–296)). Conclusions MIS-C patients in a later stage of the pandemic displayed a more severe phenotype, reflecting the impact of distinct SARS-CoV-2 variants. NT-proBNP emerged as the most crucial feature associated with intensive care need, underscoring the importance of monitoring.
  • Thumbnail Image
    ItemOpen Access
    Machine learning using multimodal clinical, electroencephalographic, and magnetic resonance imaging data can predict incident depression in adults with epilepsy: A pilot study
    (Wiley, 2023-07-08) Delgado-García, Guillermo; Engbers, Jordan D. T.; Wiebe, Samuel; Mouches, Pauline; Amador, Kimberly; Forkert, Nils D.; White, James; Sajobi, Tolulope; Klein, Karl Martin; Josephson, Colin B.; Calgary Comprehensive Epilepsy Program Collaborators
    Objective: To develop a multi-modal machine learning (ML) approach for predicting incident depression in adults with epilepsy. Methods: We randomly selected 200 patients from the Calgary Comprehensive Epilepsy Program registry and linked their registry-based clinical data to their first-available clinical electroencephalogram (EEG) and magnetic resonance imaging (MRI) study. We excluded patients with a clinical or Neurological Disorders Depression Inventory for Epilepsy (NDDI-E)-based diagnosis of major depression at baseline. The NDDI-E was used to detect incident depression over a median of 2.4 years follow-up (IQR 1.5-3.3 years). A ReliefF algorithm was applied to clinical as well as quantitative EEG and MRI parameters for feature selection. Six ML algorithms were trained and tested using stratified three-fold cross-validation. Multiple metrics were used to assess model performances. Results: Of 200 patients, 150 had EEG and MRI data of sufficient quality for ML, of which 59 were excluded due to prevalent depression. Therefore, 91 patients (41 women) were included with a median age of 29 (IQR 22-44) years. A total of 42 features were selected by ReliefF, none of which were quantitative MRI or EEG variables. All models had a sensitivity >80% and 5 of 6 had an F1 score ≥0.72. Multilayer perceptron model had the highest F1 score (median 0.74; interquartile range [IQR] 0.71-0.78) and sensitivity (84.3%). Median area under the receiver operating characteristic curve and normalized Matthews correlation coefficient were 0.70 (IQR 0.64-0.78) and 0.57 (IQR 0.50-0.65), respectively. Significance: Multimodal machine learning using baseline features can predict incident depression in this population. Our pilot models demonstrated high accuracy for depression prediction. However, overall performance and calibration can be improved. This model has promise for identifying those at risk for incident depression during follow-up, though efforts to refine it in larger populations along with external validation are required.
  • Thumbnail Image
    ItemOpen Access
    Using MRI and Atlas-based Volumetrics to Investigate Grey Matter Atrophy in Animal Models of Multiple Sclerosis
    (2023-06) Hamilton, Andrew Max; Dunn, Jeff F.; Yong, V. Wee; Forkert, Nils D.
    Loss of brain volume known as atrophy, occurs at an accelerated rate in Multiple Sclerosis (MS) compared to healthy adults and is closely associated with clinical disability and disease progression. To better understand and treat atrophy, mouse models that feature atrophy along with other aspects of MS pathology are needed. The objective of this thesis was to utilize magnetic resonance imaging (MRI) and atlas-based volumetrics to investigate animal models of MS for atrophy and determine if they are suitable for studying atrophy in MS. First, we investigated the experimental autoimmune encephalomyelitis (EAE) model commonly used to study neuroinflammation in MS. EAE mice had lower brain volumes at chronic long-term disease duration with atrophy identified in both white (WM) and grey matter (GM) regions including the cerebral cortex, cerebellum, hippocampus, corpus callosum, basal forebrain, midbrain, optic tract, and colliculus. Like MS, atrophy was associated with neurodegeneration and long-term physical disability. Next, to validate these results, we measured atrophy in EAE along with cerebral blood flow (CBF), another aspect of MS pathology possibly associated with atrophy. Using continuous arterial spin labelling MRI, we observed a reduction in cortical CBF at peak clinical disease and long-term disease in EAE. Reduced CBF at peak clinical disease appears to be related to systemic inflammation and blood brain barrier breakdown as reduced CBF was also seen in the systemic inflammation controls. Reduced CBF at long-term was instead correlated with atrophy. Finally, we investigated the cuprizone model, commonly used to study demyelination. Following chronic demyelination, atrophy was seen in WM and subcortical GM regions including the corpus callosum, internal capsule, stria terminalis, striatum, thalamus, and globus pallidus. Atrophy coincided with demyelination, and WM axonal injury, but not neuronal or axonal loss. This lack of neuroaxonal loss suggests demyelination alone is not sufficient to cause neurodegeneration and the inflammatory infiltration seen in EAE and MS may be necessary. Overall, this thesis provides a foundation for future studies into atrophy in animal models of MS and demonstrates the utility of our methods for assessing neuroprotective therapies or investigating the pathophysiology behind atrophy in MS.