Jussi Tohka

This study uses unsupervised deep learning to analyze rat brain histology, revealing tissue organization and detecting pathological changes without manual annotations.

This paper introduces a surface-intrinsic framework for hippocampal unfolding that improves alignment accuracy and morphology preservation compared to volumetric methods.

This study introduces frequency-dependent diffusion tensor distribution imaging (ωDTD) combined with machine learning to better characterize ischemic tissue damage in rats, showing improved representation of microstructural changes compared to conventional DTI.

The paper develops a protocol to characterize epileptogenic tissue using multimodal data from surgical samples, aiming to improve diagnosis and treatment of drug-resistant epilepsy.

The paper investigates how (y)-dependent MNAR data affects neurodegeneration classification, showing biases in performance metrics and the robustness of mean/median imputation over tree-based methods.

This paper predicts future amyloid beta positivity in amyloid-negative individuals using machine learning with non-invasive measures, finding better performance with CSF-based definitions.

This study shows chronic stress and neuronal activity remodel nodes of Ranvier in mice, linking structural changes to anxiety-like behavior.

The paper develops a CNN-based method for automated hippocampus segmentation in rat MRI images post-TBI, showing superior performance compared to registration-based methods.

The paper predicts future cognition using dynamic cortical anatomy changes from longitudinal neuroimaging data in children.

DeepACSON segments myelinated axons, cell nuclei, and mitochondria in low-resolution 3D-EM datasets of rat brain tissue, capturing morphological changes post-injury.

The paper introduces MU-Net, a convolutional neural network for automated skull-stripping and segmentation in mouse brain MRI, achieving high accuracy and speed.

This paper predicts Alzheimer's disease progression using multi-modal data and predictive learning models, showing better performance than single-modal approaches.

The paper shows that astrocytes from schizophrenia patients exhibit transcriptional and functional changes, including inflammation and synaptic pathway alterations, which are partially reversible with clozapine treatment.

The paper identifies reduced grey matter density in the left entorhinal cortex as a biomarker for predicting mild cognitive impairment in cognitively healthy elderly individuals.

The paper introduces a feature selection and extraction method for multiclass brain imaging problems, reducing feature dimensions while improving classification performance.

This study investigates the role of the meningeal lymphatic system in migraine pain by analyzing inflammatory and nociceptive responses in mice lacking this system.

The paper validates statistical methods for comparing inter-subject correlations between fMRI groups using permutation tests and recommends subject-wise permutations over element-wise ones.

The paper develops an automated 3D segmentation pipeline for axonal morphology analysis in rat white matter and finds reduced axonal diameter in chronic brain injury.

The paper shows that combining cortical and subcortical white/gray contrast improves age prediction and reveals cognitive performance differences through residual analysis.

This paper introduces sign-consistency bagging (SCB) for variable importance in brain imaging, showing improved reproducibility and classification accuracy compared to existing methods.

This study compares MRI-based feature representations for predicting MCI-to-AD conversion, finding regional features perform best but without significant statistical differences.

The paper predicts chronological age and cognitive performance using white/gray contrast metrics in brain imaging, showing better correlation with IQ than cortical thickness.

This paper introduces FuSeISC, a method for functional brain segmentation using inter-subject correlation in fMRI data, identifying both consistent and variable processing areas.

The paper develops a machine learning method to predict symptom severity in ASD using cortical thickness data from multi-site ABIDE data, achieving better performance than existing approaches.