Biology subjects

Tohka, J.

Publications and source records attributed to Tohka, J..

At least 19 recordsLinked to original sources

bioRxiv2026

A surface-intrinsic framework for topology-preserving hippocampal alignment and precision morphometry

Accurate alignment of hippocampal anatomy across individuals remains challenging due to complex and highly variable folding patterns that are not well captured by conventional volumetric approaches. HippUnfold introduced a surface-based representation of the hippocampus, but key components--including coordinate estimation and inter-subject correspondence--were defined in the volumetric domain, making them susceptible to topological errors and interpolation artifacts. Here, we introduce a surface-intrinsic formulation of hippocampal unfolding in which geometry, intrinsic coordinates, and correspondence are defined directly on subject-specific surface manifolds. Intrinsic anterior-posterior and proximal-distal coordinates are computed by solving Laplace equations on the surface, and correspondence is established through surface-based resampling in unfolded space, replacing inverse volumetric warping. Relative to the original HippUnfold approach, this formulation improves test-retest consistency, subject identifiability, and mesh quality, while better preserving subject-specific gyral and sulcal morphology. Surface representations show reduced distortion between folded and unfolded spaces and eliminate misplaced or outlier vertices associated with volumetric warping. These improvements translate to enhanced sensitivity in a clinical application, improving lateralization of temporal lobe epilepsy. These results demonstrate that a surface-intrinsic formulation provides a principled and robust foundation for hippocampal unfolding, enabling topology-preserving alignment and more accurate characterization of inter-individual variability in health and disease.

neuroscience↗

bioRxiv2026

Autoencoders for unsupervised analysis of rat myeloarchitecture

Quantitative assessment of brain histology is often constrained by predefined feature sets and labor-intensive manual annotations. To overcome these limitations, we employed unsupervised deep learning to automatically extract and quantify tissue organizational patterns from myelin-stained rat brain sections without the need for prior labeling. We evaluated nonlinear convolutional autoencoders (AEs) against linear principal component analysis (PCA) for feature representation, followed by clustering with Gaussian mixture models. Compared to PCA, AEs better preserved fine axonal architecture and produced more consistent and interpretable tissue clusters across hierarchical levels. The resulting clusters revealed anatomically meaningful tissue organization, including different white matter densities and grey matter subregions. When applied to tissue from sham and mild traumatic brain injury animals, AE-derived features also captured pathology-related alterations, such as white matter loss and injury-specific microstructural changes. These findings demonstrate that unsupervised deep learning can automatically characterize tissue organization at multiple scales and detect pathological changes, offering a scalable, unbiased approach to quantitative neuropathology.

neuroscience↗

bioRxiv2025

Frequency-dependent diffusion tensor distribution imaging in the evaluation of ischemic stroke

Non-invasive MRI is widely used to assess and monitor ischemic stroke, yet conventional approaches often lack sensitivity to subtle microstructural changes and struggle to evaluate tissue viability across lesion, penumbra, and distal regions. In this study, frequency dependent diffusion tensor distribution imaging ({omega}DTD) was combined with clustering of diffusion tensor distributions D({omega}) and multivariate regression modeling to characterize ischemic tissue alterations in a whole brain section. Ex vivo {omega}DTD and histology were performed in rats subjected to middle cerebral artery occlusion (MCAO) or sham surgery (P = 17) 24 hours after reperfusion. Lesions showed cell loss and an increased presence of smaller, likely glial, cells. A random forest (RF) model was used to explain and predict histological parameters from diffusion tensor imaging (DTI), manually bin resolved {omega}DTD features, and cluster resolved {omega}DTD parameters. Model performance was evaluated using leave one animal out cross validation (LOO CV). {omega}DTD features better represented cell number than DTI metrics ({omega}DTD R2 = 0.73 vs. DTI R2 = 0.49), with similar advantages for nuclear area and circularity ({omega}DTD R2 = 0.64 and 0.61 vs. DTI R2 = 0.40 and 0.35). The RF model further proved beneficial in capturing complex, nonlinear relationships between MRI parameters and tissue characteristics. Overall, these results indicate that {omega}DTD provides richer microstructural information than standard DTI, and that combining {omega}DTD with advanced machine learning methods enhances interpretation of ischemic tissue damage.

neuroscience↗

bioRxiv2025

Multimodal approach to characterize surgically removed epileptogenic zone from patients with focal drug-resistant epilepsy: from operating room to wet lab

ObjectiveWe have established a comprehensive sample handling protocol designed for the multiscale assessment of epileptogenic tissue. This protocol aims to identify novel therapeutic targets and enhance the diagnosis and stratification of patients with drug-resistant epilepsy, thereby optimizing their treatment with anti-seizure medications and surgical interventions. MethodsPatients with drug-resistant focal epilepsy, recommended for surgical treatment, are recruited after detailed multidisciplinary preoperative evaluation at the Epilepsy Center at Kuopio University Hospital in Finland. A day before the resective surgery, patients undergo magnetic resonance imaging (MRI) including advanced methodologies. During the surgery, each piece of resected tissue is placed under oxygenation on ice-cold artificial cerebral spinal fluid-solution. The pieces are then immediately transported to the laboratory, assessed by a neuropathologist, and sliced for both clinical diagnosis and research. Two adjacent slices are provided for research and are sent to the University of Eastern Finland. ResultsThe developed sample handling protocol provides the opportunity for detailed characterization of the tissue from the same patient using emerging imaging, electrophysiology, and molecular biology technologies. We have optimized the conditions for preserving the resected tissue alive for electrophysiological measurements and simultaneously making possible ex vivo studies including multi-omics acquisition, electron microscopy, histology, and MRI. Our protocol enables the mapping of functional readouts to structural and molecular alterations in human tissue. Our goal is to integrate multimodal data and co-register the resected tissues within the whole brains in vivo MRI space. This approach aims to enhance the characterization and localization of epileptogenic zones and refine surgical treatment targets by identifying abnormalities in global connectivity and structural patterns. SignificanceWe have successfully developed a systematic protocol for the collection and analysis of multimodal data. This protocol aims to elucidate the structural, functional, and molecular characteristics that render tissue epileptogenic, thereby enhancing the diagnosis and subsequent care of patients with epilepsy. Key pointsO_LIIntroducing a systematic sample handling protocol to assess tissue epileptogenicity at structural, functional and molecular levels. C_LIO_LIA multimodal approach integrating advanced technology with detailed characterization of epileptogenic tissue properties to obtain patient-specific data. C_LIO_LICorrelation of data from preoperative MRI and resected tissue to predict tissue pathology from clinical MRI. C_LI

neuroscience↗

bioRxiv2025

Impact of missing data correlated with labels to be predicted in neurodegeneration classification tasks

We introduce a new subtype of Missing Not at Random (MNAR) data, where the missingness is correlated with the labels (y) to be predicted, termed (y)-dependent MNAR. We demonstrate that this subtype can significantly bias the estimation of performance metrics in typical machine learning tasks. Unbiased error estimation is crucial in predictive modeling to accurately assess model performance, identify potential biases, and ensure generalizability to new, unseen data. We explore the effects of imputing this new subtype of MNAR and compare it with general missing types, namely Missing at Random (MAR) and Missing Completely at Random (MCAR). Our comparison analysis employs both synthetic and clinical datasets, including the Alzheimers Disease Neuroimaging Initiative (ADNI) dataset, the Parkinsons Progression Markers Initiative (PPMI) dataset, and the Anti-Amyloid Treatment in Asymptomatic Alzheimers Disease (A4) dataset. After introducing missingness into the datasets, we trained different classifiers paired with various imputation methods and measured repeated cross-validation test metrics. Our findings reveal that datasets with non-ignorable missing types (MNAR) exhibit a strong bias compared to ignorable types (MAR and MCAR) in downstream analysis. Non-linear classifiers tend to exploit patterns from imputed data, particularly when the imputed values correlate with the target label (y), which can lead to unreliable estimation of the generalization error. Mean and median imputations proved to be more robust than tree-based or gradient boosting methods.

bioinformatics↗

bioRxiv2023

Machine learning prediction of future amyloid beta positivity in amyloid-negative individuals

INTRODUCTIONThe pathophysiology of Alzheimers disease (AD) involves {beta}-amyloid (A{beta}) accumulation. Early identification of individuals with abnormal {beta}-amyloid levels is crucial, but A{beta} quantification with positron emission tomography (PET) and cerebrospinal fluid (CSF) is invasive and expensive. METHODSWe propose a machine learning framework using standard non-invasive (MRI, demographics, APOE, neuropsychology) measures to predict future A{beta}-positivity in A{beta}-negative individuals. We separately study A{beta}-positivity defined by PET and CSF. RESULTS: Cross-validated AUC for 4-year A{beta} conversion prediction was 0.78 for the CSF-based and 0.68 for the PET-based A{beta} definitions. Although not trained for the clinical status-change prediction, the CSF-based model excelled in predicting future mild cognitive impairment (MCI)/dementia conversion in cognitively normal/MCI individuals (AUCs, respectively, 0.76 and 0.89 with a separate dataset). DISCUSSIONStandard measures have potential in detecting future A{beta}-positivity and assessing conversion risk, even in cognitively normal individuals. The CSF-based definition led to better predictions than the PET-based definition.

neuroscience↗

bioRxiv2022

Mind the gap: Nodes of Ranvier are remodeled by chronic psychosocial stress and neuronal activity

Differential expression of myelin-related genes and changes in myelin thickness have been demonstrated in mice after chronic psychosocial stress, a risk factor for anxiety disorders. To determine whether and how stress affects structural remodeling of nodes of Ranvier, another form of myelin plasticity, we developed a 3D reconstruction analysis of node morphology in C57BL/6NCrl and DBA/2NCrl mice. We identified strain-dependent effects of chronic stress on node morphology, including elongation of paranodes in the medial prefrontal cortex (mPFC) in DBA/2NCrl mice. Furthermore, chronic chemogenetic activation of the ventral hippocampus-to-mPFC pathway resulted in increased risk assessment behavior and shortened paranodes specifically in stimulated axons, providing a direct link between anxiety-like behavior and remodeling of the nodes. Altogether, our data demonstrate genetic regulation of nodal remodeling in stress and suggest an activity-dependent regulation of paranodes in anxiety-related circuits. Nodal remodeling may thus contribute to the aberrant circuit function associated with anxiety disorders.

neuroscience↗

bioRxiv2021

Automatic segmentation of the rat brain hippocampus in MRI after traumatic brain injury

AO_SCPLOWBSTRACTC_SCPLOWRegistration-based methods are commonly used in the anatomical segmentation of magnetic resonance (MR) brain images. However, they are sensitive to the presence of deforming brain pathologies that may interfere with the alignment of the atlas image with the target image. Our goal was to develop an algorithm for automated segmentation of the normal and injured rat hippocampus. We implemented automated segmentation using a U-Net-like Convolutional Neural Network (CNN). of sham-operated experimental controls and rats with lateral-fluid-percussion induced traumatic brain injury (TBI) on MR images and trained ensembles of CNNs. Their performance was compared to three registration-based methods: single-atlas, multi-atlas based on majority voting and Similarity and Truth Estimation for Propagated Segmentations (STEPS). Then, the automatic segmentations were quantitatively evaluated using six metrics: Dice score, Hausdorff distance, precision, recall, volume similarity and compactness using cross-validation. Our CNN and multi-atlas -based segmentations provided excellent results (Dice scores > 0.90) despite the presence of brain lesions, atrophy and ventricular enlargement. In contrast, the performance of singe-atlas registration was poor (Dice scores < 0.85). Unlike registration-based methods, which performed better in segmenting the contralateral than the ipsilateral hippocampus, our CNN-based method performed equally well bilaterally. Finally, we assessed the progression of hippocampal damage after TBI by applying our automated segmentation tool. Our data show that the presence of TBI, time after TBI, and whether the location of the hippocampus was ipsilateral or contralateral to the injury explained hippocampal volume (p = 0.029, p < 0.001, and p < 0.001 respectively).

bioinformatics↗

bioRxiv2021

Individualized prediction of future cognition using baseline developmental changes in cortical anatomy

Predictive modeling studies have started to reveal brain measures underlying cognition; however, most studies are based on cross-sectional data (static final brain measures acquired at one time point). Since brain development comprises of continuously ongoing events leading to cognitive development, predictive modeling studies need to consider dynamic as opposed to static final brain measures. Using longitudinal neuroimaging and cognitive data (global executive composite score, an index of executive function) from 82 individuals (aged 5-14 years, scanned 3 times), we built highly accurate prediction models (r=0.61, p=1.6e-09) of future cognition (assessed at visit 3) based on baseline developmental changes in cortical anatomy (from visit 1 to 2). More importantly, dynamic brain measures (change in cortical anatomy from visit 1 to 2) and not static brain measures (cortical anatomy at visit 1 and 2) were critical for predicting future cognition, suggesting the need for considering dynamic brain measures in predicting cognitive outcomes.

neuroscience↗

bioRxiv2020

Patient iPSC-astrocytes show transcriptional and functional dysregulation in schizophrenia

Human astrocytes are multifunctional brain cells and may contribute to the pathophysiology of schizophrenia (SCZ). We differentiated astrocytes from induced pluripotent stem cells of monozygotic twins discordant for SCZ, and found sex-specific gene expression and signaling pathway alterations related particularly to inflammation and synaptic functions. While Ingenuity Pathway Analysis identified SCZ disease and synaptic transmission pathway changes in SCZ astrocytes, the most consistent findings were related to collagen and cell adhesion associated pathways. Neuronal responses to glutamate and GABA differed between astrocytes from control persons, affected twins, and their unaffected co-twins, and were normalized by clozapine treatment. SCZ astrocyte cell transplantation to the mouse forebrain caused gene expression changes in demyelination, synaptic dysfunction and inflammation pathways of mouse brain cells and resulted in behavioral changes in cognitive and olfactory functions. Altogether, our results show that astrocytes contribute to both familial risk and clinical manifestation of SCZ in a sex-specific manner.

neuroscience↗

bioRxiv2020

Predicting the future development of mild cognitive impairment in the cognitively healthy elderly

Identifying measures that predict future cognitive impairment in healthy individuals is necessary to inform treatment strategies for candidate dementia-preventative and modifying interventions. Here, we derive such measures by studying "converters" who transitioned from cognitively normal at baseline to mild-cognitive impairment (MCI) in a longitudinal study of 1213 elderly participants. We first establish reduced grey matter density (GMD) in left entorhinal cortex (EC) as a biomarker for impending cognitive decline in healthy individuals, employing a matched sampling control for several dementia risk-factors, thereby mitigating the potential effects of bias on our statistical tests. Next, we determine the predictive performance of baseline demographic, genetic, neuropsychological and MRI measures by entering these variables into an elastic net-regularized classifier. Our trained statistical model classified converters and controls with validation Area-Under-the-Curve>0.9, identifying only delayed verbal memory and left EC GMD as relevant predictors for classification. This performance was maintained on test classification of out-of-sample converters and controls. Our results suggest a parsimonious but powerful predictive model for MCI development in the cognitively healthy elderly.

neuroscience↗

bioRxiv2020

Quantitative longitudinal predictions of Alzheimer's disease by multi-modal predictive learning

BackgroundQuantitatively predicting the progression of Alzheimers disease (AD) in an individual on a continuous scale, such as AD assessment scale-cognitive (ADAS-cog) scores, is informative for a personalized approach as opposed to qualitatively classifying the individual into a broad disease category. We hypothesize that multi-modal data and predictive learning models can be employed for longitudinally predicting ADAS-cog scores. MethodsMultivariate regression techniques were employed to model baseline multi-modal data (demographics, neuroimaging, and cerebrospinal fluid based markers, and genetic factors) and future ADAS-cog scores. Prediction models were subjected to repeated cross-validation and the resulting mean absolute error and cross-validated correlation of the model assessed. ResultsPrediction models on multi-modal data outperformed single modal data up to 36 months. Incorporating baseline ADAS-cog scores to prediction models marginally improved predictive performance. ConclusionsFuture ADAS-cog scores were successfully estimated via predictive learning aiding clinicians in identifying those at greater risk of decline and apply interventions at an earlier disease stage and inform likely future disease progression in individuals enrolled in AD clinical trials.

bioinformatics↗

bioRxiv2020

The role of the meningeal lymphatic system in local inflammation and trigeminal nociception implicated in migraine pain

BackgroundA system of lymphatic vessels has been recently characterized in the meninges, with a postulated role in cleaning the brain via cerebral fluid drainage. As meninges are the origin site of migraine pain, we hypothesized that malfunctioning of the lymphatic system should affect the local trigeminal nociception. To test this hypothesis, we studied nociceptive and inflammatory mechanisms in the meninges of K14-VEGFR3-Ig mice lacking the meningeal lymphatic system. MethodsWe recorded the spiking activity of meningeal afferents and estimated the local mast cells infiltration, calcitonin gene-related peptide (CGRP) and cytokine levels (basal and stimulated), as well as the dural trigeminal innervation in freshly-isolated hemiskull preparations from K14-VEGFR3-Ig (K14) or wild type C57BL/6 mice (WT). ResultsWe found that the meningeal level of CGRP and of the pro-inflammatory cytokines IL12-p70 and TNF (implicated in migraine) were reduced in the meninges of K14 mice. On the contrary, in the meninges of K14 mice, we found an increased level of the mast cell activator MCP-1 and, consistently, a larger number of dural mast cells. The other migraine-related pro-inflammatory cytokines did not differ between the two genotypes. The patterns of trigeminal innervation in meninges remained unchanged and we did not observe alterations in basal or ATP-induced nociceptive firing in the meningeal afferents. ConclusionsIn summary, the lack of meningeal lymphatic system does not induce migraine-like nociceptive state per se, but leads to a new balance between pro- and antiinflammatory factors implicated in migraine mechanisms.

neuroscience↗

bioRxiv2020

Automated skull-stripping and segmentation with Multi-Task U-Net in large mouse brain MRI databases

Skull-stripping and region segmentation are fundamental steps in preclinical magnetic resonance imaging (MRI) studies, and these common procedures are usually performed manually. We present Multi-task U-Net (MU-Net), a convolutional neural network designed to accomplish both tasks simultaneously. MU-Net achieved higher segmentation accuracy than state-of-the-art multi-atlas segmentation methods with an inference time of 0.35 seconds and no pre-processing requirements. We evaluated the performance of our network in the presence of skip connections and recently proposed framing connections, finding the simplest network to be the most effective. We tested MU-Net with an unusually large dataset combining several independent studies consisting of 1,782 mouse brain MRI volumes of both healthy and Huntington animals, and measured average Dice scores of 0.906 (striati), 0.937 (cortex), and 0.978 (brain mask). These high evaluation scores demonstrate that MU-Net is a powerful tool for segmentation and skull-stripping, decreasing inter and intra-rater variability of manual segmentation. The MU-Net code and the trained model are publicly available at https://github.com/Hierakonpolis/MU-Net.

bioinformatics↗

bioRxiv2019

DeepACSON: Automated White Matter Segmentation

Automated segmentation techniques are essential to tracing the entirety of ultrastructures in large three-dimensional electron microscopy (3D-EM) images of the brain tissue. Current automated techniques use deep convolutional neural networks (DCNNs) and rely on high-contrast cellular membranes to trace a small number of neuronal processes in very high-resolution EM datasets. We developed DeepACSON to segment large field-of-view, low-resolution 3D-EM datasets of white matter where tens of thousands of myelinated axons traverse the tissue. DeepACSON performs DCNN-based semantic segmentation and shape decomposition-based instance segmentation. With its top-down design, DeepACSON manages to account for severe membrane discontinuities inescapable with the low-resolution imaging. In particular, the instance segmentation of DeepACSON uses the tubularity of myelinated axons, decomposing an under-segmented myelinated axon into its constituent axons. We applied DeepACSON to ten serial block-face scanning electron microscopy datasets of rats after sham-operation or traumatic brain injury, segmenting hundreds of thousands of long-span myelinated axons, thousands of cell nuclei, and millions of mitochondria with excellent evaluation scores. DeepACSON quantified the morphology and spatial aspects of white matter ultrastructures, capturing nanoscopic morphological alterations five months after the injury.

neuroscience↗

bioRxiv2019

Regularized Bagged Canonical Component Analysis for Multiclass Learning in Brain Imaging

A fundamental problem of supervised learning algorithms for brain imaging applications is that the number of features far exceeds the number of subjects. In this paper, we propose a combined feature selection and extraction approach for multiclass problems. This method starts with a bagging procedure which calculates the sign consistency of the multivariate analysis (MVA) projection matrix feature-wise to determine the relevance of each feature. This relevance measure provides a parsimonious matrix, which is combined with a hypothesis test to automatically determine the number of selected features. Then, a novel MVA regularized with the sign and magnitude consistency of the features is used to generate a reduced set of summary components providing a compact data description.\n\nWe evaluated the proposed method with two multiclass brain imaging problems: 1) the classification of the elderly subjects in four classes (cognitively normal, stable mild cognitive impairment (MCI), MCI converting to AD in 3 years, and Alzheimers disease) based on structural brain imaging data from the ADNI cohort; 2) the classification of children in 3 classes (typically developing, and 2 types of Attention Deficit/Hyperactivity Disorder (ADHD)) based on functional connectivity. Experimental results confirmed that each brain image (defined by 29.852 features in the ADNI database and 61.425 in the ADHD) could be represented with only 30 - 45% of the original features. Furthermore, this information could be redefined into two or three summary components, providing not only a gain of interpretability but also classification rate improvements when compared to state-of-art reference methods.

bioinformatics↗

bioRxiv2019

Cortical and subcortical T1 white/gray contrast, chronological age, and cognitive performance

The maturational schedule of typical brain development is tightly constrained; deviations from it are associated with cognitive atypicalities, and are potentially predictive of developmental disorders. Previously, we have shown that the white/gray contrast at the inner border of the cortex is a good predictor of chronological age, and is sensitive to aspects of brain development that reflect cognitive performance. Here we extend that work to include the contrast at the white/gray border of subcortical structures. We show that cortical and subcortical contrast together yield better age-predictions than any non-kernel-based method based on a single image-type, and that the residuals of the improved predictions provide new insight into unevenness in cognitive performance. We demonstrate the improvement in age predictions in two large datasets: the NIH Pediatric Data, with 831 scans of typically developing individuals between 4 and 22 years of age; and the Pediatric Imaging, Neurocognition, and Genetics data, with 909 scans of individuals in a similar age-range. Assessment of the relation of the residuals of these age predictions to verbal and performance IQ revealed correlations in opposing directions, and a principal component analysis of the residuals of the model that best fit the contrast data produced components related to either performance IQ or verbal IQ. Performance IQ was associated with the first principle component, reflecting increased cortical contrast, broadly, with almost no subcortical presence; verbal IQ was associated with the second principle component, reflecting reduced contrast in the basal ganglia and increased contrast in the bilateral arcuate fasciculi.

neuroscience↗

bioRxiv2018

Comparing fMRI inter-subject correlations between groups

Inter-subject correlation (ISC) based analysis is a conceptually simple approach to analyze functional magnetic resonance imaging (fMRI) data acquired under naturalistic stimuli such as a movie. We describe and validate the statistical approaches for comparing ISCs between two groups of subjects implemented in the ISC toolbox, which is an open source software package for ISC-based analysis of fMRI data. The approaches are based on permutation tests. We validated the approaches using five different data sets from the ICBM functional reference battery tasks. First, we created five null datasets (one for each task) by dividing the subjects into two matched groups and assumed that no group difference exists. Second, based on one null dataset, we created datasets with simulated ISC differences of varying size between the two groups. Based on the experiments with these two types of data, we recommend the use of subject-wise permutations, instead of element-wise permutations. The tests based on subject-wise permutations led to correct false positive rates. We observed that the null-distributions should be voxel-specific and not based on pooling all voxels across the brain as is typical in fMRI. This was the case even if studentized permutation tests were used. Additionally, we experimented with an fMRI dataset acquired using a dance movie stimulus for comparison of a group of adult males on the autism spectrum to a matched typically developed group. The experiment confirmed the differences between voxel-based permutation tests and global model based permutation tests.

neuroscience↗