Iñigo Tellaetxe-Elorriaga

Mr. Iñigo Tellaetxe-Elorriaga

orcid

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 January 2025 in IEEE Journal of Biomedical and Health Informatics
Verified
Journal Article
Open access
JCJorge Garcia Condado
Jorge Garcia Condado
ITIñigo Tellaetxe-Elorriaga
Iñigo Tellaetxe-Elorriaga
JCJesus M. Cortes
Jesus M. Cortes
AEAsier Erramuzpe
Asier Erramuzpe
An approach to age modeling involves the supervised prediction of age using machine learning from subject features. The derived age metrics are used to study the relationship between healthy and pathological aging in multiple body systems, as well as the interactions between them. We lack a standard for this type of age modeling. In this work we developed AgeML, an OpenSource software for age-prediction from any type of tabular clinical data following well-established and tested methodologies. The objective is to set standards for reproducibility and standardization of reporting in supervised age modeling tasks. AgeML does age modeling, calculates age deltas, the difference between predicted and chronological age, measures correlations between age deltas and factors, visualizes differences in age deltas of different clinical populations and classifies clinical populations based on age deltas. With this software we are able to reproduce published work and unveil novel relationships between body organs and polygenetic risk scores. AgeML is age modeling made easy for standardization and reproducibility.

21 October 2025 in NeuroImage
Verified
Journal Article
Open access
JCJorge Garcia Condado
Jorge Garcia Condado
IRInes Verdugo Recuero
Ines Verdugo Recuero
ITIñigo Tellaetxe-Elorriaga
Iñigo Tellaetxe-Elorriaga
CBColin Birkenbihl
Colin Birkenbihl
MCMaria Carrigan
Maria Carrigan
JCJesus M. Cortes
Jesus M. Cortes
BrainAge models estimate the biological age of the brain using neuroimaging or clinical features, making them promising tools for studying neurodegenerative diseases like Alzheimer's disease. However, the reliance of BrainAge models on neuroimaging features such as grey matter volume and hippocampal atrophy, can introduce biases linked to individuals' ages as these features are influenced both by normal aging and Alzheimer's disease progression. The potential presence of such age-biases raises a critical question: can BrainAge models trained to estimate biological brain aging make meaningful contributions to Alzheimer's diagnosis, or does any introduced age-bias conflate aging effects with disease pathology? Understanding how deliberate feature selection impacts this confounding effect is essential for developing reliable age-related biomarkers. We ranked neuroimaging and neuropsychological features based on their mutual information with age and their discriminative power across four clinical groups: cognitively normal, Mild Cognitive Impairment, Alzheimer's Disease, stable Mild Cognitive Impairment and progressive Mild Cognitive Impairment. Iteratively, we trained BrainAge models using different subsets of these features, some optimized for predicting aging and others for discrimination of clinical Alzheimer's disease stages. We assess the error in BrainAge delta, the difference between predicted biological age and chronological age, and evaluate its classification performance across clinical groups. Finally, we compare using deltas for classification with a logistic regression model directly trained on the neuroimaging features used in the BrainAge models. Neuroimaging features are more strongly correlated with aging, while neuropsychological features exhibit greater discriminative power for Alzheimer's disease classification. BrainAge models optimized for age prediction yield deltas that are suboptimal when used for classifying Alzheimer´s disease, whereas models trained to generate deltas optimized to be used for classifying Alzheimer's disease have reduced age prediction accuracy. This trade-off suggests that BrainAge models may not optimally separate aging-related changes from disease-specific alterations. BrainAge models have varying classification accuracy as compared to direct utilization of features in logistic regression. However, BrainAge provides a continuous measure, offering a single output that can be used across clinical stages, in contrast to classification approaches that require explicit labels for each disease stage. Aging significantly affects BrainAge-based classification of Alzheimer's disease. Feature selection plays a critical role in mitigating this effect, as the outputs of models trained to predict age, the deltas, may fail in Alzheimer´s disease classification. These findings underscore the need for task-specific feature selection and model design to ensure that BrainAge models are appropriately applied in neurodegenerative disease research.

21 March 2025 in bioRxiv
Verified
Preprint
Open access
PDF file
BCBorja Camino-Pontes
Borja Camino-Pontes
AJAntonio Jimenez-Marin
Antonio Jimenez-Marin
ITIñigo Tellaetxe-Elorriaga
Iñigo Tellaetxe-Elorriaga
IFIzaro Fernandez-Iriondo
Izaro Fernandez-Iriondo
AEAsier Erramuzpe
Asier Erramuzpe
JCJesus M. Cortes
Jesus M. Cortes
The brain’s modular organization, ranging from microcircuits to large-scale networks, has been extensively studied in terms of its structural and functional properties. Particularly insightful has been the investigation of the coupling between structural connectivity (SC) and functional connectivity (FC), whose analysis has revealed important insights into the brain’s efficiency and adaptability related to various cognitive functions. Interestingly, links in SC are intrinsically pairwise but this is not the case for FC; and while recent work demonstrates the relevance of the brain’s high-order interactions (HOI), the coupling of between SC and functional HOI remains unexplored. To address this gap, this study leverages functional MRI and diffusion weighted imaging to delineate the brain’s modular structure by investigating the coupling between SC and functional HOI. Our results demonstrates that structural networks can be associated with both redundant and synergistic functional interactions. In particular, SC exhibits both positive and negative correlations with redundancy, it shows consistent positive correlations with synergy, indicating that a higher density of structural connections is linked to increased synergistic interactions. These findings advance our understanding of the complex relationship between structural and high-order functional properties, shedding light on the brain’s architecture underlying its modular organization. Significance Statement: Understanding the intricate relationship between the brain structural and functional connectivity is instrumental for enabling a richer characterization of cognitive functions. This challenge is complex, since it involves comparing pairwise links representing white-matter fibers with higher-order functional interactions that can involve multiple brain regions sharing information — indeed, these interactions can be redundant in presence of shared information, or synergistic when novel information arises in the whole subsystem. Here we advance the investigation of the coupling between brain structure and high-order interactions, offering new insights into how structural properties are associated with both redundant and synergistic functional modules. Furthermore, the approach presented here opens new doors to investigate how this coupling between structural connectivity and functional HOI is modulated during development or under diverse brain pathologies.

10 July 2025 in medRxiv
Verified
Preprint
Open access
PDF file
JCJorge Garcia Condado
Jorge Garcia Condado
HKHannah M Klinger
Hannah M Klinger
CBColin Birkenbihl
Colin Birkenbihl
MCMadison Cuppels
Madison Cuppels
ALAnnie Liu
Annie Liu
AIAlzheimer’s Disease Neuroimaging Initiative
Alzheimer’s Disease Neuroimaging Initiative
BrainAge delta, the difference between a person’s predicted brain age and their chronological age, is a promising marker of the accumulation of neurodegeneration that may increase vulnerability to Alzheimer’s disease (AD). In this study we use structural MRI regions vulnerable to AD, the hippocampus, the amygdala and cortical thickness measurements to build BrainAge models. We examined whether BrainAge delta moderates the relationship between AD biomarkers and longitudinal cognitive decline performing a meta-analysis across three cohorts: A4/LEARN, HABS and ADNI (2,279 cognitively unimpaired [CU]; 416 with mild cognitive impairment [MCI]). Higher BrainAge delta was linked to faster decline in CU (β = -0.13 [-0.21, - 0.06], p = 0.018, I 2 =1%) and more strongly in MCI (β = -0.31 [-0.30, -0.24], p < 1x10 -16 ). BrainAge also interacted with Aβ-PET (β = -0.09 [-0.13, -0.05], p = 0.0054, I 2 =12%) and plasma pTau 217 (β = -0.09 [-0.15, -0.03], p = 0.018, I 2 =0.1%), but not Tau-PET, to impact cognitive decline, where synergistically higher BrainAge delta and elevated AD markers resulted in faster cognitive decline. We next tested its utility for clinical trial enrichment. Sequential screening with pTau 217 and BrainAge delta reduced required sample size for prevention trials by 77%, versus 61% using pTau 217 alone. These findings support BrainAge delta as a marker of neurodegeneration and may serve as an enrichment tool for AD prevention trials.

18 November 2025 in Journal of Clinical Medicine
Verified
Journal Article
Open access
PDF file
AEAinara Estanga
Ainara Estanga
ITIñigo Tellaetxe-Elorriaga
Iñigo Tellaetxe-Elorriaga
MEMirian Ecay-Torres
Mirian Ecay-Torres
JCJorge García Condado
Jorge García Condado
MGMaite García-Sebastián
Maite García-Sebastián
Miren AltunaMiren Altuna
Miren Altuna
Background: Chronological age is an imprecise proxy for cognitive aging. The Cognitive Age Delta (CAD)—the difference between predicted cognitive age and chronological age—offers a scalable, individualized marker of functional brain aging. We examined determinants of CAD in cognitively unimpaired (CU) adults stratified by Alzheimer’s disease (AD) and vascular biomarkers. Methods: We analyzed 177 CU participants from the Gipuzkoa Alzheimer Project (Basque Country, Northern Spain) classified as amyloid-negative/vascular-negative (CUA−V−, n = 140), amyloid-positive (CUA+, n = 23), or vascular-positive (CUV+, n = 14) using CSF and MRI criteria; vascular burden was defined as Fazekas ≥ 2 on T2-FLAIR or ≥4 microbleeds on SWI, excluding non-traumatic superficial siderosis and established ischemic lesions. MRI was used solely for vascular classification. Associations with demographic, genetic, lifestyle, and reserve measures were tested with General Linear Models. Results: CAD did not differ across biomarker groups (Kruskal–Wallis H(2) = 0.17, p = 0.91). Median (IQR) CAD values were 0.28 (−4.13, 4.69) for CUA−V−, −0.14 (−3.15, 2.87) for CUA+, and 0.77 (−2.22, 3.76) for CUV+, indicating comparable distributions. Higher vocabulary scores (proxy of cognitive reserve) related to a younger cognitive age in CUA−V− (β = −1.39, p < 0.001) and CUA+ (β = −2.08, p = 0.054). In CUA+, greater sedentary time—particularly computer-based sitting—was also associated with lower CAD (daily sitting β = −2.13, p = 0.009; workday computer sitting β = −2.32, p = 0.015). CAD showed no associations with CSF Aβ42, p-tau or t-tau, APOE ε4 load, or vascular risk factors (all p > 0.05). Conclusions: CAD captures interindividual resilience-related variability beyond classical AD biomarkers. Vocabulary, a marker of lifelong enrichment, emerged as a robust determinant of a younger cognitive age, while amyloid and vascular pathology exerted limited influence at preclinical stages. These findings support CAD as a sensitive, scalable endpoint for identifying protective factors and guiding personalized prevention in early Aging.

17 November 2025 in medRxiv
Verified
Preprint
Open access
PDF file
AJAntonio Jimenez-Marin
Antonio Jimenez-Marin
GSGil De Sousa
Gil De Sousa
ITIñigo Tellaetxe-Elorriaga
Iñigo Tellaetxe-Elorriaga
IEInaki Escudero
Inaki Escudero
ACAlberto Cabrera-Zubizarreta
Alberto Cabrera-Zubizarreta
JCJesus M. Cortes
Jesus M. Cortes
Background: Post-stroke spasticity (PSS) is a common and disabling motor complication whose neuroanatomical underpinnings remain incompletely understood. Beyond focal lesion localization, PSS likely arises from large-scale network disruptions involving both cortical and subcortical systems. Here, we combined functional and structural lesion network mapping (LNM) to identify the brain networks and white-matter tracts whose dysconnection best explains the presence and severity of PSS. Methods: We analyzed 281 patients with hemorrhagic stroke, including 88 with PSS and 193 without PSS (NPSS). After matching for age, sex, and lesion size (N=81 PSS; N=60 NPSS), we computed functional LNM using normative connectomes from 1000 healthy participants of the Human Connectome Project and structural LNM using the HCP1065 tractography template. Group-level voxel-wise and tract-wise analyses identified regions and tracts with significant dysconnectivity differences. Finally, a ridge regression model with leave-one-out cross-validation assessed the relationship between dysconnectivity (both functional and structural) and spasticity severity, quantified by the Modified Ashworth Scale (MAS). Results: Lesion-symptom mapping revealed that PSS was primarily associated with lesions in the posterior putamen. Structural LNM showed significantly higher tract dysconnectivity in PSS, particularly within the corticospinal, corticopontine, corticobulbar, corticostriatal, medial lemniscus, and superior thalamic radiation pathways (all pNPSS) and a Negative Dysconnected Network (PSS<NPSS). Despite these bidirectional effects, overall dysconnection was consistently higher in the PSS group. The most affected networks included the cerebellum, sensorimotor, auditory, dorsal attention, default mode, and medial visual resting-state networks. Ridge regression analysis confirmed a strong association between MAS scores and the degree of dysconnectivity across these networks (R2=0.70). Conclusions: Post-stroke spasticity is not the result of damage to a single locus but rather reflects the dysconnection of a distributed structural and functional network encompassing, not only motor, but also attentional and high order cognitive networks. These findings provide a network-based framework for understanding spasticity and support the use of lesion network mapping to predict and potentially guide targeted neuromodulation in stroke rehabilitation.