Academic interests
- Modeling and prediction of changes in the brain and cognition
- Machine learning: deep learning, Bayesian inference, stochastic processes
- Medical image registration
Background
Cybernetics and Robotics – Master’s program (sivilingeniør), 5-year – NTNU
Publications
-
Roe, James Michael; J., Jagust, William; M., Landau, Susan; M., Harrison, Theresa; Grydeland, Håkon & Slivka, Maksim
[Show all 19 contributors for this article]
(2026).
Cortical thickness changes precede high levels of amyloid by at least 7 years | Nature Neuroscience.
Nature Neuroscience.
ISSN 1097-6256.
doi:
10.1038/s41593-026-02363-4.
Full text in Research Archive
Show summary
Alzheimer’s disease is now defined by underlying pathology, with elevated amyloid-beta (Aβ) sufficient for diagnosis in the absence of cognitive symptoms. We combined longitudinal magnetic resonance imaging and Aβ positron emission tomography data from three cognitively healthy cohorts to examine cortical thickness trajectories in individuals who later converted to Aβ-positive status, using magnetic resonance images acquired years before conversion. Individuals who subsequently developed elevated Aβ showed a thicker cortex and reduced cortical thinning, detectable up to 7 years before conversion. Many effects persisted after accounting for quantitative Aβ levels, suggesting some cortical thickness changes may be partly independent of Aβ. Differences in cortical thickness and its change showed moderate spatial correspondence with Aβ deposition patterns, and the timing of thickness changes tracked the progression of Aβ accumulation. These findings indicate that cortical thickness alterations can precede positron emission tomography-detectable amyloid positivity by several years, suggesting high amyloid burden may not represent the earliest imaging marker of Alzheimer’s disease. Longitudinal MRI reveals that the cortex becomes relatively thicker years before amyloid positivity is detected on PET, providing insight into the earliest brain changes and potential consequences of amyloid accumulation in Alzheimer’s disease.
-
-
-
-
Leonardsen, Esten Høyland; Persson, Karin Ester Torun; Grødem, Edvard Olaf Sætersdal; Dinsdale, Nicola; SCHELLHORN, TILL & Roe, James Michael
[Show all 16 contributors for this article]
(2024).
Constructing personalized characterizations of structural brain aberrations in patients with dementia using explainable artificial intelligence.
npj Digital Medicine.
7(1).
doi:
10.1038/s41746-024-01123-7.
Full text in Research Archive
Show summary
Deep learning approaches for clinical predictions based on magnetic resonance imaging data have shown great promise as a translational technology for diagnosis and prognosis in neurological disorders, but its clinical impact has been limited. This is partially attributed to the opaqueness of deep learning models, causing insufficient understanding of what underlies their decisions. To overcome this, we trained convolutional neural networks on structural brain scans to differentiate dementia patients from healthy controls, and applied layerwise relevance propagation to procure individual-level explanations of the model predictions. Through extensive validations we demonstrate that deviations recognized by the model corroborate existing knowledge of structural brain aberrations in dementia. By employing the explainable dementia classifier in a longitudinal dataset of patients with mild cognitive impairment, we show that the spatially rich explanations complement the model prediction when forecasting transition to dementia and help characterize the biological manifestation of disease in the individual brain. Overall, our work exemplifies the clinical potential of explainable artificial intelligence in precision medicine.
-
Grødem, Edvard Olaf Sætersdal; Sweeney, Kieran & McClean, Megan N
(2020).
Automated calibration of optoPlate LEDs to reduce light dose variation in optogenetic experiments.
BioTechniques.
ISSN 0736-6205.
69(4),
p. 313–316.
doi:
10.2144/btn-2020-0077.
Full text in Research Archive
Show summary
Optogenetic systems use light to precisely control and investigate cellular processes. Until recently, there had been few instruments available for applying controlled light doses to cultures of cells. The optoPlate, a programmable array of 192 LEDs, was developed to meet this need. However, LED performance varies, and without calibration there are substantial brightness differences between LEDs on an optoPlate. Here we present a method for calibrating an optoPlate that uses a programmable microscope stage and optical power meter to automatically measure all 192 LEDs of an optoPlate. The resulting brightness measurements are used to calculate calibration values that tune the electrical current supplied to each optoPlate LED to reduce brightness variation in optogenetic experiments.
View all works in NVA
-
-
-
Grødem, Edvard Olaf Sætersdal; Fjell, Anders Martin; Bjørnerud, Atle; MacIntosh, Bradley & Edin, Nina Frederike Jeppesen
(2025).
Characterizing Brain MRI with Machine Learning Methods with Applications to Alzheimer’s Disease.
Universitetet i Oslo.
Full text in Research Archive
Show summary
The human brain undergoes lifelong structural changes that reflect both normal aging and pathological processes. Alzheimer’s disease (AD) is neurodegenerative disorder that causes pronounced structural changes in the brain in the form of atrophy. In AD, patients’ cognitive function is permanently reduced, and treatment options are limited. Magnetic resonance imaging (MRI) makes it possible to observe these structural changes in vivo. Over the past twenty years, large open imaging databases (e.g., the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and UK Biobank) have enabled studies involving tens of thousands of participants. At the same time, there have been major advances in machine-learning methods, for example in deep learning, Bayesian statistics and image processing. Together, large databases and new machine-learning techniques open up deeper insights into the brain’s evolving anatomy.
Nevertheless, challenges remain. There is considerable inter-individual anatomical differences, and distinguishing natural variation from pathology can be difficult. The MRI data themselves present additional hurdles: the mages are large three-dimensional scans that must be processed to obtain measurable quantities such as regional volume and cortical thickness. Although the imaging databases are extensive, scans often originate from many different scanners, leading to contrast differences that complicate direct comparisons. In longitudinal studies, the morphological changes of interest are often smaller than the measurement noise. Consequently, new methods are needed to analyze changes in brain MRI reliably.
In this thesis, the aim is to develop and improve machine-learning methods for analyzing brain MRI. These methods are intended to quantify changes in brain structures, to characterize the brain’s Natural development across the lifespan and to diagnose AD. The work comprises three papers.
In Paper I we use deep learning to diagnose Alzheimer’s disease from brain MRI. We investigate how the architecture of the deep convolutional network influences the classification outcome. MRI scans of individuals with AD and healthy controls from ADNI serve as material. We find that more complex and deeper architectures such as EfficientNet do not outperform simpler and shallower architectures such as the Simple Fully Convolutional Network (SFCN). Further analyses show that SFCN can be scaled down to very small sizes while still providing strong results. These findings indicate that straightforward preprocessing combined with simple convolutional networks yields classification performance comparable to the more elaborate methods reported in the literature.
In Paper II we study volume changes of various brain structures across the lifespan. We aim to determine when there are individual changes in the brain due to normal aging and to what extent these can be estimated at the individual level. A large dataset of longitudinal brain MRI collected from 14 studies is employed. We model the volume of the brain structures as a stochastic dynamical system and fit the model to the data using a Markov chain Monte Carlo approach. The results show that the brain changes throughout life, but up to the age of 50 there are only small individual differences in the rate of change. After 50 years we observe greater individual variation. From the model we calculate the explained variance from the structure volume at age 18 to the volume later in life, finding that up to 60 years almost all variance is explained by the young-adult volume. The lateral ventricles stand out: here, individual variation in change persists throughout life. We further estimate how accurately individual change can be inferred for single subjects. For individuals under 50 years there is virtually no correlation between an estimated change and the actual individualchange; after 50 years this correlation increases with age. Increasing the interval between volume measurements effectively increases this correlation, whereas adding more measurements within a given interval has a smaller effect.
Paper III examines medical image-registration algorithms, in which a deformation field that aligns two images is estimated so that anatomically identical regions overlap. This can be used both to compare images across individuals and to study anatomical changes over time. Registration is often performed by minimising the image-similarity loss Local Normalised Cross-Correlation (LNCC). We show that using LNCC can lead to erroneous deformation estimates in homogeneous regions where noise dominates the signal. Two new variants are proposed: Regularized-LNCC (R-LNCC) and Weighted-LNCC (W-LNCC). We use real MRI from a wide range of scanners and generate synthetic deformations to test how well the similarity metrics recover them. LNCC produces incorrect deformation estimates in low-variance regions, whereas R-LNCC and W-LNCC resolve this issue. We also find that R-LNCC generally yields significantly better deformation estimates than LNCC, even in more heterogeneous regions.
Together, Papers I, II and III constitute a framework for studying changes in the brain. Our aim is that these tools will be utilized to better understand Alzheimer’s disease and promote healthy aging.
View all works in NVA
Published
Jan. 14, 2025 11:58 AM
- Last modified
Nov. 18, 2025 2:02 PM
