Pasar al contenido principal

GEMELOS DIGITALES PARA DETECCION PRECOZ DE SIGNOS DE DETERIORO COGNITIVO Y ASISTENCIA PERSONALIZADA MEDIANTE TECNICAS DE IA JUSTA Y EXPLICABLE

PID2024-161711OB-I00

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal para la Investigación y el Desarrollo Experimental
Subprograma Subprograma Estatal de Generación de Conocimiento Científico-Técnico y Desarrollo Experimental
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2024
Plan Estatal 2024-2027 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSIDAD DE ALICANTE

Resultados relacionados

Resultados totales (Incluyendo duplicados): 7
Encontrada(s) 1 página(s)

RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Multimodal pain assessment with transformers

  • Benavent-Lledó, Manuel
  • López Valle, María Dolores
  • Ortiz Pérez, David
  • Mulero Pérez, David
  • Garcia-Rodriguez, Jose
  • Psarrou, Alexandra
  • Benavent-Lledó, Manuel|||/items/0d1bda20-de83-4080-97c9-1820faf481da
  • Ortiz Pérez, David|||/items/e77e67d5-78fc-4170-a005-b6d3fd3ca79e
  • Mulero Pérez, David|||/items/e3a342d3-f8d6-4010-bf87-a509394c0d36
  • Garcia-Rodriguez, Jose|||/items/d5bc18f8-44db-4723-b291-5849321e9e69
Pain assessment is a critical challenge in healthcare, requiring accurate and objective measurement to enhance patient care. Traditional methods rely on subjective self-reporting, which lacks reliability, particularly for patients with communication difficulties. This work presents PainFusion+, a multimodal transformer architecture for pain assessment that integrates physiological signals and facial expressions to improve pain assessment accuracy. For physiological signals, our approach first employs convolutional neural networks to extract local patterns from short signal fragments, capturing essential pain-related features. These localized representations are then processed by a transformer encoder, which models long-range dependencies to form a comprehensive global representation. For facial video data, we leverage a frozen video transformer to extract expressive features without requiring fine-tuning, significantly reducing computational costs. Finally, both feature spaces are fused using a transformer encoder, allowing effective cross-modal learning. Experiments on publicly available datasets demonstrate that PainFusion+ outperforms existing models. In biomedical signal processing, our method achieves over a 16 % improvement in accuracy. For multimodal pain estimation, it achieves 35.40 % accuracy on the BioVid dataset, setting a new state-of-the-art benchmark.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Protecting the Vulnerable: Leveraging GANs and Depth Maps for Cognitive Decline Scenarios

  • Burgos Miras, José
  • Martinez-Martin, Ester
  • Garcia-Rodriguez, Jose
  • Martinez-Martin, Ester|||/items/74a8fa4f-d7ad-4df1-9b91-355b474a0cd2
  • Garcia-Rodriguez, Jose|||/items/d5bc18f8-44db-4723-b291-5849321e9e69
In an increasingly digitized society, facial recognition has evolved from a convenience to a critical accessibility tool, particularly for vulnerable populations facing cognitive or motor impairments. By eliminating the cognitive load of traditional passwords, these biometric systems facilitate essential digital inclusion. However, this reliance creates a heightened imperative for security; as these users often depend on passive authentication methods, they are disproportionately susceptible to sophisticated adversarial attacks. Consequently, protecting these interfaces against manipulated or synthetically generated identities is not merely a technical challenge, but a fundamental safeguard against the financial and social exploitation of at-risk demographics. This paper aims to develop a novel GAN-based facial identification architecture capable of realistic image synthesis and adversarial detection, thereby ensuring robustness against targeted security breaches. With great emphasis on real-world applications, some experiments under diverse settings have been carried out. The obtained results are very promising contributing toward safe and secure authentication technologies and open new research lines to be explored.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Digital Health Competence and Attitudes Toward AI Among Health Care Professionals: Convergent Mixed Methods Study

  • Jiménez-García, Segundo
  • Domingo Pozo, Manuela
  • Garcia-Rodriguez, Jose
  • Tomás, David
  • Ortiz Pérez, David
  • Mikkonen, Kristina
  • Jarva, Erika
  • Vizcaya-Moreno, M. Flores
  • Jiménez-García, Segundo|||/items/35eb3c1f-fb4b-46bb-a9de-7820d35dfc16
  • Domingo Pozo, Manuela|||/items/c8d1ccfd-0084-416b-9cbd-a646c79fee8e
  • Garcia-Rodriguez, Jose|||/items/d5bc18f8-44db-4723-b291-5849321e9e69
  • Tomás, David|||/items/cf031fc3-3c30-454b-a6df-3762c29cc296
  • Ortiz Pérez, David|||/items/e77e67d5-78fc-4170-a005-b6d3fd3ca79e
  • Vizcaya-Moreno, M. Flores|||/items/47df53a0-4d4d-48fe-8f29-471e90417a09
Background: Digital transformation is reshaping health care systems and requires health care professionals to develop advanced digital health competencies. The integration of AI into clinical practice introduces new demands related to critical evaluation, human oversight, ethical responsibility, and regulatory compliance. However, evidence linking validated measures of digital health competence with health care professionals’ attitudes toward AI remains limited. Objective: This study aimed to (1) assess digital health competence among health care professionals in Spain, (2) examine organizational conditions supporting competence development, (3) explore perceptions about AI use in the workplace, and (4) examine the association between digital health competence and attitudes toward AI. Methods: A national cross-sectional convergent mixed methods study was conducted between November 2023 and January 2024 with a voluntary convenience sample of 229 health care professionals. Digital health competence was assessed using DigiHealthCom (digital health competence instrument; 42 items, 5 domains) and DigiComInf (the aspects associated with digital health competence instrument; 15 items, 3 domains). Confirmatory factor analysis evaluated structural validity. Open-ended responses regarding AI perceptions were explored using inductive qualitative content analysis. AI attitudes were classified into 4 categories (positive, negative, ambivalent, and uncertain), and their association with digital health competence was examined using multinomial logistic regression adjusted for age and sex. Results: Participants were predominantly female (169/229, 73.8%) and nurses (123/229, 53.7%), with a mean age of 45.9 (SD 10) years. Overall digital health competence was moderate (mean 3.01, SD 0.51), with the highest scores in information and communication technology competence (3.34, SD 0.63) and the lowest in competence related to evaluating and implementing digital solutions (2.83, SD 0.63). Organizational and educational support for competence development was also moderate (mean 2.51, SD 0.57), with organizational planning receiving the lowest DigiComInf scores (2.21, SD 0.76). Confirmatory factor analysis supported the proposed factor structures for both instruments (DigiHealthCom: comparative fit index=0.95, root-mean-square error of approximation=0.049; DigiComInf: comparative fit index=0.95, root-mean-square error of approximation=0.084). Most participants expressed positive attitudes toward AI in the workplace (134/228, 58.77%). Qualitative findings revealed a pattern of conditional optimism, with expected benefits for efficiency and patient care balanced by concerns regarding training, governance, regulation, and human oversight. Conclusions: Digital health competence was positively associated with health care professionals’ attitudes toward AI. However, the cross-sectional design does not allow conclusions regarding the direction of this association. Deficiencies in higher-order competencies in evaluation and implementation, together with limited organizational support, highlight areas that may benefit from targeted educational and organizational strategies to promote the safe and responsible use of AI in health care.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Deep insights into cognitive decline: A survey of leveraging non-intrusive modalities with deep learning techniques

  • Ortiz Pérez, David
  • Benavent-Lledó, Manuel
  • Garcia-Rodriguez, Jose
  • Tomás, David
  • Vizcaya-Moreno, M. Flores
  • Tomás, David|||/items/cf031fc3-3c30-454b-a6df-3762c29cc296
  • Garcia-Rodriguez, Jose|||/items/d5bc18f8-44db-4723-b291-5849321e9e69
  • Benavent-Lledó, Manuel|||/items/0d1bda20-de83-4080-97c9-1820faf481da
  • Ortiz Pérez, David|||/items/e77e67d5-78fc-4170-a005-b6d3fd3ca79e
  • Vizcaya-Moreno, M. Flores|||/items/47df53a0-4d4d-48fe-8f29-471e90417a09
Cognitive decline is a natural part of aging. However, under some circumstances, this decline is more pronounced than expected, typically due to disorders such as Alzheimer’s disease. Early detection of an anomalous decline is crucial, as it can facilitate timely professional intervention. While medical data can help, it often involves invasive procedures. An alternative approach is to employ non-intrusive techniques such as speech or handwriting analysis, which do not disturb daily activities. This survey reviews the most relevant non-intrusive methodologies that use deep learning techniques to automate the cognitive decline detection task, including audio, text, and visual processing. We discuss the key features and advantages of each modality and methodology, including state-of-the-art approaches like Transformer architecture and foundation models. In addition, we present studies that integrate different modalities to develop multimodal models. We also highlight the most significant datasets and the quantitative results from studies using these resources. From this review, several conclusions emerge. In most cases, text-based approaches consistently outperform other modalities. Furthermore, combining various approaches from individual modalities into a multimodal model consistently enhances performance across nearly all scenarios.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Enhancing action recognition by leveraging the hierarchical structure of actions and textual context

  • Benavent-Lledó, Manuel
  • Mulero Pérez, David
  • Ortiz Pérez, David
  • Garcia-Rodriguez, Jose
  • Argyros, Antonis
  • Benavent-Lledó, Manuel|||/items/0d1bda20-de83-4080-97c9-1820faf481da
  • Mulero Pérez, David|||/items/e3a342d3-f8d6-4010-bf87-a509394c0d36
  • Ortiz Pérez, David|||/items/e77e67d5-78fc-4170-a005-b6d3fd3ca79e
  • Garcia-Rodriguez, Jose|||/items/d5bc18f8-44db-4723-b291-5849321e9e69
We propose a novel approach to improve action recognition by exploiting the hierarchical organization of actions and by incorporating contextualized textual information, including location and previous actions, to reflect the action’s temporal context. To achieve this, we introduce a transformer architecture tailored for action recognition that employs both visual and textual features. Visual features are obtained from RGB and optical flow data, while text embeddings represent contextual information. Furthermore, we define a joint loss function to simultaneously train the model for both coarse- and fine-grained action recognition, effectively exploiting the hierarchical nature of actions. To demonstrate the effectiveness of our method, we extend the Toyota Smarthome Untrimmed (TSU) dataset by incorporating action hierarchies, resulting in the Hierarchical TSU dataset, a hierarchical dataset designed for monitoring activities of the elderly in home environments. An ablation study assesses the performance impact of different strategies for integrating contextual and hierarchical data. Experimental results demonstrate that the proposed method consistently outperforms SOTA methods on the Hierarchical TSU dataset, Assembly101 and IkeaASM, achieving over a 17% improvement in top-1 accuracy.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

CogniAlign: Word-level multimodal speech alignment with gated cross-attention for Alzheimer’s detection

  • Ortiz Pérez, David
  • Benavent-Lledó, Manuel
  • Rodríguez Juan, Javier
  • Garcia-Rodriguez, Jose
  • Tomás, David
  • Ortiz Pérez, David|||/items/e77e67d5-78fc-4170-a005-b6d3fd3ca79e
  • Benavent-Lledó, Manuel|||/items/0d1bda20-de83-4080-97c9-1820faf481da
  • Rodríguez Juan, Javier|||/items/4f464549-7f4d-4a47-9602-a42f7e1b5975
  • Garcia-Rodriguez, Jose|||/items/d5bc18f8-44db-4723-b291-5849321e9e69
  • Tomás, David|||/items/cf031fc3-3c30-454b-a6df-3762c29cc296
Early detection of cognitive disorders such as Alzheimer’s disease is critical for enabling timely clinical intervention and improving patient outcomes. In this work, we introduce CogniAlign, a multimodal architecture for Alzheimer’s detection that integrates audio and textual modalities, two non-intrusive sources of information that offer complementary insights into cognitive health. Unlike prior approaches that fuse modalities at a coarse level, CogniAlign leverages a word-level temporal alignment strategy that synchronizes audio embeddings with corresponding textual tokens based on transcription timestamps. This alignment supports the development of token-level fusion techniques, enabling more precise cross-modal interactions. To fully exploit this alignment, we propose a Gated Cross-Attention Fusion mechanism, where audio features attend over textual representations, guided by the superior unimodal performance of the text modality. In addition, we incorporate prosodic cues, specifically interword pauses, by inserting pause tokens into the text and generating audio embeddings for silent intervals, further enriching both streams. We evaluate CogniAlign on the ADReSSo dataset, where it achieves an accuracy of 87.35 % over a Leave-One-Subject-Out setup and of 90.36 % over a 5 fold Cross-Validation, outperforming existing state-of-the-art methods. A detailed ablation study confirms the advantages of our alignment strategy, attention-based fusion, and prosodic modeling. Finally, we perform a corpus analysis to assess the impact of the proposed prosodic features and apply Integrated Gradients to identify the most influential input segments used by the model in predicting cognitive health outcomes.
RUA. Repositorio Institucional de la Universidad de Alicante
Publicación Otros (Other)

Sign4all: a Spanish Sign Language dataset

  • Morillas-Espejo, Francisco
  • Martinez-Martin, Ester
  • Morillas-Espejo, Francisco|||/items/b5013d27-cf4b-4462-97b3-65b4109fefad
  • Martinez-Martin, Ester|||/items/74a8fa4f-d7ad-4df1-9b91-355b474a0cd2
Sign Language Recognition (SLR) is a critical component of human-machine interaction, enabling more inclusive technologies for the deaf and hard-of-hearing community. However, current datasets often suffer from data sparsity and a bias toward right-handed signs. To support this effort, we present Sign4all, a dataset for Spanish Sign Language (LSE), specifically designed for Isolated Sign Language Recognition (ISLR). The dataset is composed of 7,756 high-resolution RGB video recordings and their corresponding skeletal keypoints, covering 24 signs related to daily activities, more specifically a vocabulary centered in the catering field. Unlike sparse lexicons, Sign4all adopts a high-density approach, providing an average of 323 samples per sign to facilitate data-intensive deep learning models. Moreover, the dataset provides a handedness balance, with equal representation of left- and right-handed signs for every sign to support handedness invariance. Each sample was manually segmented, temporally normalized and preprocessed through spatial normalization to guarantee consistency and compatibility with different deep learning pipelines. Technical validation using Transformer and skeletal models demonstrates the dataset’s integrity and the need of providing pre-computed augmentation splits. All data is formatted in widely supported file types (AVI for video, HDF5 for keypoints), enabling direct use in machine learning frameworks such as TensorFlow or PyTorch.