Biology subjects

Thielemann, C.

Publications and source records attributed to Thielemann, C..

4 recordsLinked to original sources

bioRxiv2025

Classifying Calcium Imaging Dynamics with Deep Learning: Multi-Frequency Analysis through Quantile-Based Time-Series Network Representations

To address the limitations of calcium imaging data, we propose a segmentation-agnostic deep learning framework that integrates Quantile-Based Time-Series Network (QTN) representations with convolutional neural networks to classify neuronal dynamics across multiple spatial resolutions and acquisition frequencies. By transforming fluorescence traces into compact, fixed-size matrices derived from quantile transitions, the method standardizes inputs across recordings while markedly reducing dimensionality and computational cost. Several QTN variants were systematically evaluated, demonstrating strong and consistent classification performance across both whole-image and grid-based preprocessing strategies. Notably, the framework maintained high accuracy under reduced temporal resolution and controlled noise perturbations, confirming that discrimination arises from meaningful temporal patterns rather than artifacts. This study establishes a robust, scalable, and generalizable approach for analyzing calcium imaging dynamics, paving the way for efficient, segmentation-independent characterization of neuronal activity in pharmacological and systems neuroscience applications.

cell biology↗

bioRxiv2025

Network Rerouting Under Ayahuasca: Temporally and Hemisphere-Resolved EEG Connectomics

Ayahuasca profoundly alters conscious experience, yet robust, time-resolved EEG markers of its network-level effects remain limited. We combined machine learning with complex-network analysis to quantify how functional connectivity reorganizes across time and hemispheres in resting-state EEG from a randomized, double-blind, placebo-controlled trial including three 5-min sessions: pre-dose (T1), 2 h post-dose (T2), and 4 h post-dose (T3). The cohort consisted of naive ayahuasca users, a population known to exhibit attenuated or more stable acute responses, making the detection of network-level changes particularly challenging. Connectivity was estimated using multiple metrics and sliding windows (10-120 s), and network features were computed and averaged to ensure statistical validity. A representation-selection step identified Spearman correlation and an intermediate temporal scale as optimal, with classification performance peaking at 60-70 s (independent-test AUC and accuracy = 0.93). Linear mixed models revealed a bilateral decrease in eigenvector centrality (weaker hub influence), increased degree heterogeneity in the right hemisphere, and reduced global efficiency in the left. Edge-level analyses localized these effects: Posterior-left connections weakened acutely (lowest at T2), whereas right temporal-central coupling transiently strengthened (highest at T2). Together, these convergent results support a mechanistic summary: as hub-centric short-cuts weaken, communication is increasingly routed through alternative, more distributed--and less efficient--pathways, with a right-lateralized expression at a later time. Methodologically, the window-optimized, hemisphere-resolved, and edge-validated pipeline extends prior EEG work and highlights temporal scale (approximately 60 s) as a biologically meaningful parameter for detecting psychedelic-induced network reorganization.

neuroscience↗

bioRxiv2025

Human cerebral organoids model tumor infiltration and migration supported by astrocytes in an autologous setting

Efforts to achieve precise and efficient tumor targeting of highly malignant brain tumors are constrained by the dearth of appropriate models to study the effects and potential side effects of radiation, chemotherapy, and immunotherapy on the most complex human organ, the brain. We established a cerebral organoid model of brain tumorigenesis in an autologous setting by overexpressing c-MYC as one of the most common oncogenes in brain tumors. GFP+/c-MYChigh cells were isolated from tumor organoids and used in two different culture approaches: assembloids comprising of a normal cerebral organoid with a GFP+/c-MYChigh tumor sphere and co-culture of cerebral organoid slices at air-liquid interface with GFP+/c-MYChigh cells. GFP+/c-MYChigh cells used in both approaches exhibited tumor-like properties, including overexpression of the c-MYC oncogene, high proliferative and invasive potential, and an immature phenotype as evidenced by increased expression of Ki-67, VIM, and CD133. Organoids and organoid slices served as suitable scaffolds for infiltrating tumor-like cells. Using our highly reproducible and powerful model system that allows long-term culture, we demonstrated that the migratory and infiltrative potential of tumor-like cells is shaped by the environment in which glia cells provide support to tumor-like cells.

cell biology↗

bioRxiv2020

Measuring Neuronal Signals with Microelectrode Arrays: A Finite Element Analysis

ObjectiveMeasuring neuronal cell activity using microelectrode arrays reveals a great variety of derived signal shapes within extracellular recordings. However, possible mechanisms responsible for this variety have not yet been entirely determined, which might hamper any subsequent analysis of the recorded neuronal data. For an investigation of this issue, we propose a computational model based on the finite element method describing the electrical coupling between an electrically active neuron and an extracellular recording electrode in detail. This allows for a systematic study of possible parameters that may play an essential role in defining or altering the shape of the measured electrode potential. Our results indicate that neuronal geometry and neurite structure, as well as the actual pathways of input potentials that evoke action potential generation, have a significant impact on the shape of the resulting extracellular electrode recording and explain most of the known signal shape variety.

bioengineering↗