Biology subjects

Gronlund, C.

Publications and source records attributed to Gronlund, C..

5 recordsLinked to original sources

bioRxiv2026

Evaluation of synthetic data generation for longitudinal 6-year follow-up subclinical cardiovascular disease data using GAN models

Although largely preventable, cardiovascular disease (CVD) is the leading cause of death worldwide. Synthetic data generation (SDG) of health data can facilitate data sharing for research and development of personalised cardiovascular disease prevention tools. Current state-of-the-art SDG methods are limited to cross-sectional data or focused on short-term or in-patient EHR applications lacking benchmarks for generalisability. Since prevention is a long-term longitudinal objective, these limitations must be addressed. We evaluated two tabular generative adversarial networks, TGAN and CTGAN, for privacy and fidelity on a structured 6-year longitudinal trial dataset within CVD prevention. We compared performance with a rudimentary perturbation method. In addition, we performed a variable-wise evaluation and an age-stratified analysis of the higher performing model to assess preservation of clinically relevant characteristics. Overall, CTGAN showed better fidelity than TGAN. The models performed similarly within privacy, showing better results than the perturbation method. CTGAN had lower performance in comparison to the perturbation method within numerical fidelity of continuous variables and overall temporal fidelity. Furthermore, CTGAN's performance was lower for younger age groups than older. CTGAN shows better than or equal performance within fidelity and privacy than TGAN and the perturbation method on a long-term longitudinal dataset within CVD prevention. To the best of our knowledge, this is the first study evaluating CTGAN on a longitudinal dataset and our findings suggest that handling of temporal characteristics and imbalanced subgroups need further refinement.

bioinformatics↗

bioRxiv2026

Evaluating in-context learning with prompting regimes for transformer-based synthetic health tabular data generation

Tabular synthetic data generation (SDG) can facilitate research and development in the health domain, as access to health data is restricted. To accommodate for its structural complexity, such as multimodal distributions and fragmentation, transformer models provides an alternative to conventional SDG methods since they are better equipped to infer global spatial coherence due to textual context of tabular records. However, methods for optimising their adaption to tabular SDG are still largely unexplored. We contribute to the understanding of transformer-based SDG by evaluating fidelity and privacy of three prompting regimes, zero shot, one shot and few shot training, for tabular health SDG using small transformer models. We evaluated llama 3.2 (1B and 3B parameters) and Phi (1.3B and 2.7B parameters), on a real clinical dataset. Our results showed that zero shot learning was less prone to hallucinations and performed overall better for fidelity, but at the loss of privacy in comparison to the other regimes. Our results showed no differences between model sizes and families. Future studies should extend the analysis to other models and datasets, and explore strategies to reduce hallucinations.

bioinformatics↗

bioRxiv2023

Combining high-density electromyography and ultrafast ultrasound to assess individual motor unit properties in vivo

This study aims to compare two methods for the identification of anatomical and mechanical motor unit (MU) properties through the integration of high-density surface electromyography (HDsEMG) and ultrafast ultrasound (UUS). The two approaches rely on a combined analysis of the firing pattern of active MUs, identified from HDsEMG, and tissue velocity sequences of the muscle cross-section, obtained from UUS. The first method is the spike-triggered averaging (STA) of the tissue velocity sequence based on the occurrences of MU firings. The second is a method based on spatio-temporal independent component analysis (STICA) enhanced with the information of single MU firings. We compared the capability of these two approaches to identify the regions where single MU fibers are located within the muscle cross-section (MU displacement area) in vivo. HDsEMG signals and UUS images were detected simultaneously from biceps brachii in ten participants (6 males and 4 females) during low-level isometric elbow flexions. Experimental signals were processed by implementing both STA and STICA approaches. The medio-lateral distance between the estimated MU displacement areas and the centroid of the MU action potential distributions was used to compare the two methods. We found that STICA and STA are able to detect MU displacement areas. However, STICA provides more precise estimations to the detriment of higher computational complexity.

bioengineering↗

bioRxiv2023

Spatially repeatable components from ultrafast ultrasound are associated with motor unit activity in human isometric contractions

ObjectiveUltrafast ultrasound imaging has been used to measure intramuscular mechanical dynamics associated with single motor unit (MU) activations. Detecting MU activity from ultrasound sequences requires decomposing a displacement velocity field into components consisting of spatial maps and temporal displacement signals. These components can be associated with putative MU activity or spurious movements (noise). The differentiation between putative MUs and noise has been accomplished by comparing the temporal displacement signals with MU firings obtained from needle EMG. Here, we examined whether the repeatability of the spatial maps over brief time intervals can serve as a criterion for distinguishing putative MUs from noise in low-force isometric contractions. ApproachIn five healthy subjects, ultrafast ultrasound images and high-density surface EMG (HDsEMG) were recorded simultaneously from biceps brachii. MUs identified through HDsEMG decomposition were used as a reference to assess the outcomes of the ultrasound-based decomposition. For each contraction, displacement velocity sequences from the same eight-second ultrasound recording were separated into consecutive two-second epochs and decomposed. The Jaccard Similarity Coefficient (JSC) was employed to evaluate the repeatability of components spatial maps across epochs. Finally, the association between the ultrasound components and the MUs decomposed from HDsEMG was assessed. Main resultsAll the MU-matched components had JSC > 0.38, indicating they were repeatable and accounted for about one-third of the HDsEMG-detected MUs (1.8 {+/-} 1.6 matches over 4.9 {+/-} 1.8 MUs). The repeatable components (with JSC over the empirical threshold of 0.38) represented 14% of the total components (6.5 {+/-} 3.3 components). These findings align with our hypothesis that intra-sequence repeatability can differentiate putative MUs from spurious components and can be used for data reduction. SignificanceThe results of our study provide the foundation for developing stand-alone methods to identify MU in ultrafast ultrasound sequences and represent a step forward towards real-time imaging of active MU territories. These methods are relevant for studying muscle neuromechanics and designing novel neural interfaces.

bioengineering↗

bioRxiv2022

Optimization and comparison of two methods for spike train estimation in an unfused tetanic contraction of low threshold motor units

BackgroundHuman movement is generated by activating motor units (MUs), i.e., the smallest structures that can be voluntarily controlled. Recent findings have shown imaging of voluntarily activated MUs using ultrafast ultrasound based on displacement velocity images and a decomposition algorithm. Given this, estimates of trains of twitches (unfused tetanic signals) evoked by the neural discharges (spikes) of spinal motor neurons are provided. Based on these signals, a band-pass filter method (BPM) has been used to estimate its spike train. In addition, an improved spike estimation method consisting of a continuous Haar wavelet transform method (HWM) has been suggested. However, the parameters of the two methods have not been optimized, and their performance has not been compared rigorously. MethodHWM and BPM were optimized using simulations. Their performance was evaluated based on simulations and two experimental datasets with 21 unfused tetanic contractions considering their rate of agreement, spike offset, and spike offset variability with respect to the simulated or experimental spikes. ResultsA range of parameter sets that resulted in the highest possible agreement with simulated spikes was provided. Both methods highly agreed with simulated and experimental spikes, but HWM was a better spike estimation method than BPM because it had a higher agreement, less bias, and less variation (p < 0.001). ConclusionsThe optimized HWM will be an important contributor to further developing the identification and analysis of MUs using imaging, providing indirect access to the neural drive of the spinal cord to the muscle by the unfused tetanic signals.

bioengineering↗