Disease progression models (DPMs) are valuable tools for characterizing early cognitive decline in Alzheimer's Disease (AD) and supporting clinical decision-making. This study aimed to (1) evaluate the diagnostic and prognostic performance of parametric DPMs, (2) identify optimal subsets of neuropsychological markers for DPM construction, and (3) benchmark three parametric DPM frameworks in early detection tasks. We analyzed longitudinal neuropsychological data from 1163 participants classified as cognitively unimpaired (CU) or with mild cognitive impairment (MCI). Three DPM approaches (Leaspy, RPDPM, and GRACE) were trained on selected marker subsets and evaluated using metrics related to diagnostic accuracy, time to conversion estimation, and robustness to missing data. Model performance was assessed via detection rates, area under the curve (AUC), mean absolute error (MAE), and Pearson correlation between estimated/observed onset ages. Leaspy achieved the highest diagnostic accuracy with an AUC of 0.96 and strong correlation with observed conversion time (r = 0.78). RPDPM showed superior robustness to missing data and maintained accurate predictions even with up to 40% data loss. GRACE offered the best trajectory fit (lowest error) but lower sensitivity to clinical transitions. A compact combination of neuropsychological tests, particularly CDRSB, ADAS13, and MMSE, was sufficient for reliable model training. Prognostic evaluation demonstrated that Leaspy provided the most consistent identification of individuals who converted to mild cognitive impairment within five years. Parametric DPMs based solely on neuropsychological measures can effectively support early detection and prognosis of cognitive decline. Leaspy showed the best overall performance, while RPDPM proved more resilient to missing data. These models enable individualized disease timelines and can inform clinical decision-making and patient stratification. All code and data used are publicly available, facilitating reproducibility and clinical translation.

Artificial Intelligence (AI) models for electrocardiogram (ECG) interpretation rely on large, diverse datasets, but existing clinical datasets are often skewed, overrepresenting normal rhythms and lacking rare pathologies. This limits model performance and generalizability. While generative AI has shown promise in other domains, its application to biosignals has so far been restricted to short segments, reducing clinical utility. Additionally, potentially promising tools like inpainting for artefact removal have not yet been explored in the biosignal domain. To overcome these limitations, we propose ECGEN, a latent diffusion model (LDM) designed to synthesize long-duration, realistic ECGs for data augmentation, rhythm-specific generation, and signal restoration. We developed ECGEN in three configurations: (1) ECGEN-Small, a 30-second conditional generator for sinus rhythm and atrial fibrillation (AFib); (2) ECGEN-Medium, a 90-second model for inpainting tasks; and (3) ECGEN-Large, an unconditional model for generating 320-second ECGs. Models were trained on real clinical ECGs from stroke patients using a vector quantized-variational autoencoder (VQ-VAE) for encoding and a denoising diffusion implicit model (DDIM) as the latent space backbone. We evaluated model outputs using heart rate (HR), heart rate variability (HRV), and morphological coherence. ECGEN-Small achieved an AUC of 0.98 for the AFib vs. sinus rhythm classification. ECGEN-Medium successfully inpainted missing segments with plausible HR dynamics. ECGEN-Large generated long ECGs with consistent morphology but showed distributional shifts in HRV (e.g., inflated standard deviation of normal-to-normal intervals, SdNN), indicating incomplete modeling of global temporal dependencies. Network depth was found to significantly influence output quality and training stability. ECGEN demonstrates the feasibility of long-duration ECG generation using LDMs, offering applications in dataset augmentation and signal restoration. However, distributional mismatches and occasional artefacts highlight challenges in unsupervised biosignal synthesis. Future work should address long-range temporal modeling and refine realism through conditioning or adversarial training.