Jordi Malé
Jordi Malé has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
- Juan Fortea
Sant Pau Memory Unit, IR SANT PAU Hospital de la Santa Creu i Sant Pau Barcelona Spain
5 shared publications - Neus Martínez-Abadías
Departament de Biologia Evolutiva, Ecologia i Ciències Ambientals (BEECA), Facultat de Biologia, Universitat de Barcelona (UB), Barcelona, Spain
5 shared publications - Xavier Sevillano
Universitat Ramon Llull
5 shared publications - David Badia
La Salle Campus, Ramon Llull University, 08022 Barcelona, Spain
4 shared publications - Joan Lluis Pijoan
Universitat Ramon Llull
4 shared publications - Joaquim Porte
La Salle Campus, Ramon Llull University, 08022 Barcelona, Spain
4 shared publications
Latest Publications
- Open Access
- 21 September 2026
- Published by Springer Nature in Pattern Analysis and Applications
- Vol. 29 (4) , 194
Generative latent representations of 3D brain MRI for multi-task downstream analysis in down syndrome
- 21 September 2026
- Published by Springer Nature in Pattern Analysis and Applications
- Vol. 29 (4) , 194
Open AccessGenerative latent representations of 3D brain MRI for multi-task downstream analysis in down syndrome
- Open Access
Generative Latent Representations of 3D Brain MRI for Multi-Task Downstream Analysis in Down Syndrome
Generative models have emerged as powerful tools in medical imaging, enabling tasks such as segmentation, anomaly detection, and high-quality synthetic data generation. These models typically rely on learning meaningful latent representations, which are particularly valuable given the high-dimensional nature of 3D medical images like brain magnetic resonance imaging (MRI) scans. Despite their potential, latent representations remain underexplored in terms of their structure, information content, and applicability to downstream clinical tasks. Investigating these representations is crucial for advancing the use of generative models in neuroimaging research and clinical decision-making. In this work, we develop multiple variational autoencoders (VAEs) to encode 3D brain MRI scans into compact latent space representations for generative and predictive applications. We systematically evaluate the effectiveness of the learned representations through three key analyses: (i) a quantitative and qualitative assessment of MRI reconstruction quality, (ii) a visualisation of the latent space structure using Principal Component Analysis, and (iii) downstream classification tasks on a proprietary dataset of euploid and Down syndrome individuals brain MRI scans. Our results demonstrate that the VAE successfully captures essential brain features while maintaining high reconstruction fidelity. The latent space exhibits clear clustering patterns, particularly in distinguishing individuals with Down syndrome from euploid controls.
Open AccessGenerative Latent Representations of 3D Brain MRI for Multi-Task Downstream Analysis in Down Syndrome
Generative models have emerged as powerful tools in medical imaging, enabling tasks such as segmentation, anomaly detection, and high-quality synthetic data generation. These models typically rely on learning meaningful latent representations, which are particularly valuable given the high-dimensional nature of 3D medical images like brain magnetic resonance imaging (MRI) scans. Despite their potential, latent representations remain underexplored in terms of their structure, information content, and applicability to downstream clinical tasks. Investigating these representations is crucial for advancing the use of generative models in neuroimaging research and clinical decision-making. In this work, we develop multiple variational autoencoders (VAEs) to encode 3D brain MRI scans into compact latent space representations for generative and predictive applications. We systematically evaluate the effectiveness of the learned representations through three key analyses: (i) a quantitative and qualitative assessment of MRI reconstruction quality, (ii) a visualisation of the latent space structure using Principal Component Analysis, and (iii) downstream classification tasks on a proprietary dataset of euploid and Down syndrome individuals brain MRI scans. Our results demonstrate that the VAE successfully captures essential brain features while maintaining high reconstruction fidelity. The latent space exhibits clear clustering patterns, particularly in distinguishing individuals with Down syndrome from euploid controls.
- 12 May 2025
- Published by Springer Nature
- p. 207-221
Towards the Discovery of Down Syndrome Brain Biomarkers Using Generative Models
- 12 May 2025
- Published by Springer Nature
- p. 207-221
Towards the Discovery of Down Syndrome Brain Biomarkers Using Generative Models
- Open Access
- 19 October 2023
- Published by IOS Press
- p. 94-97
A Deep Learning-Based Pipeline for Computing Brain Biomarkers from MRI
This ongoing work outlines a computer vision and deep learning-based pipeline to identify and detect brain biomarkers of diagnostic potential from magnetic resonance imaging (MRI) scans. In this context, this paper describes and analyses two strategies for brain landmark detection, which is a key step in brain biomarker identification: one based on a single Deep Convolutional Neural Network (DCNN) that detects multiple landmarks, and the other based on an ensemble of DCNNs trained to detect one landmark each. Based on our evaluation using two distinct datasets, our preliminary findings demonstrate that the ensemble of DC-NNs achieves superior accuracy in landmarking. Specifically, it successfully detects 84% of the landmarks within a 3mm proximity to their actual locations, with an average error of less than 2mm. In contrast, a single DCNN exhibits an average error of approximately 3mm and locates only 59% of the landmarks within a 3mm distance from their true positions.
- 19 October 2023
- Published by IOS Press
- p. 94-97
Open AccessA Deep Learning-Based Pipeline for Computing Brain Biomarkers from MRI
This ongoing work outlines a computer vision and deep learning-based pipeline to identify and detect brain biomarkers of diagnostic potential from magnetic resonance imaging (MRI) scans. In this context, this paper describes and analyses two strategies for brain landmark detection, which is a key step in brain biomarker identification: one based on a single Deep Convolutional Neural Network (DCNN) that detects multiple landmarks, and the other based on an ensemble of DCNNs trained to detect one landmark each. Based on our evaluation using two distinct datasets, our preliminary findings demonstrate that the ensemble of DC-NNs achieves superior accuracy in landmarking. Specifically, it successfully detects 84% of the landmarks within a 3mm proximity to their actual locations, with an average error of less than 2mm. In contrast, a single DCNN exhibits an average error of approximately 3mm and locates only 59% of the landmarks within a 3mm distance from their true positions.
- Open Access
- 9 July 2021
- Published by MDPI AG in Electronics
- Vol. 10 (14) , 1636
SC-FDE Layer for Sensor Networks in Remote Areas Using NVIS Communications
Despite high costs and lengthy deployments, satellite communications have traditionally been used to provide coverage in remote areas. However, given the fact that there is no radio infrastructure available in these areas, Near Vertical Incidence Skywave (NVIS) technology has positioned itself as an attractive alternative to communicate with low-power nodes in remote areas. This type of communication works in the HF frequency range complying with STANAG and MIL-STD standards, which define a physical layer for scenarios that differ from NVIS and low-power communication. The purpose of this paper was to present the definition of a new communication physical layer based on single-carrier frequency-domain equalization (SC-FDE) based on these standards but adapted to the ionospheric communication channel. This physical layer was compared to an OFDM-based layer from a previous study. The experiments performed show that this new approach achieves better results than OFDM in terms of a higher signal quality with a higher specific BER probability. Finally, this layer was also used in the theoretical design of an NVIS gateway to link sensor network devices spanning large-scale remote areas in a secure manner in the context of ubiquitous sensor networks (USN).
- 9 July 2021
- Published by MDPI AG in Electronics
- Vol. 10 (14) , 1636
Open AccessSC-FDE Layer for Sensor Networks in Remote Areas Using NVIS Communications
Despite high costs and lengthy deployments, satellite communications have traditionally been used to provide coverage in remote areas. However, given the fact that there is no radio infrastructure available in these areas, Near Vertical Incidence Skywave (NVIS) technology has positioned itself as an attractive alternative to communicate with low-power nodes in remote areas. This type of communication works in the HF frequency range complying with STANAG and MIL-STD standards, which define a physical layer for scenarios that differ from NVIS and low-power communication. The purpose of this paper was to present the definition of a new communication physical layer based on single-carrier frequency-domain equalization (SC-FDE) based on these standards but adapted to the ionospheric communication channel. This physical layer was compared to an OFDM-based layer from a previous study. The experiments performed show that this new approach achieves better results than OFDM in terms of a higher signal quality with a higher specific BER probability. Finally, this layer was also used in the theoretical design of an NVIS gateway to link sensor network devices spanning large-scale remote areas in a secure manner in the context of ubiquitous sensor networks (USN).
- Open Access
Analysis of the Ordinary and Extraordinary Ionospheric Modes for NVIS Digital Communications Channels
Sensor networks have become more popular in recent years, now featuring plenty of options and capabilities. Notwithstanding this, remote locations present many difficulties for their study and monitoring. High-frequency (HF) communications are presented as an alternative to satellite communications, being a low-cost and easy-to-deploy solution. Near vertical incidence skywave (NVIS) technology provides a coverage of approximately 250 km (depending on the frequency being used and the ionospheric conditions) without a line of sight using the ionosphere as a communication channel. This paper centers on the study of the ionosphere and its characteristic waves as two independent channels in order to improve any NVIS link, increasing its robustness or decreasing the size of the node antennas through the appliance of specific techniques. We studied the channel sounding of both the ordinary and extraordinary waves and their respective channels, analyzing parameters such as the delay spread and the channel’s availability for each wave. The frequency instability of the hardware used was also measured. Furthermore, the correlation coefficient of the impulse response between both signals was studied. Finally, we applied polarization diversity and two different combining techniques. These measurements were performed on a single frequency link, tuned to 5.4 MHz. An improvement on the mean bit energy-to-noise power spectral density (Eb/N0) was received and the bit error rate (BER) was achieved. The results obtained showed that the extraordinary mode had a higher availability throughout the day (15% more availability), but a delayed spread (approximately 0.3 ms mean value), similar to those of the ordinary wave. Furthermore, an improvement of up to 4 dB was achieved with the usage of polarization diversity, thus reducing transmission errors.
Open AccessAnalysis of the Ordinary and Extraordinary Ionospheric Modes for NVIS Digital Communications Channels
Sensor networks have become more popular in recent years, now featuring plenty of options and capabilities. Notwithstanding this, remote locations present many difficulties for their study and monitoring. High-frequency (HF) communications are presented as an alternative to satellite communications, being a low-cost and easy-to-deploy solution. Near vertical incidence skywave (NVIS) technology provides a coverage of approximately 250 km (depending on the frequency being used and the ionospheric conditions) without a line of sight using the ionosphere as a communication channel. This paper centers on the study of the ionosphere and its characteristic waves as two independent channels in order to improve any NVIS link, increasing its robustness or decreasing the size of the node antennas through the appliance of specific techniques. We studied the channel sounding of both the ordinary and extraordinary waves and their respective channels, analyzing parameters such as the delay spread and the channel’s availability for each wave. The frequency instability of the hardware used was also measured. Furthermore, the correlation coefficient of the impulse response between both signals was studied. Finally, we applied polarization diversity and two different combining techniques. These measurements were performed on a single frequency link, tuned to 5.4 MHz. An improvement on the mean bit energy-to-noise power spectral density (Eb/N0) was received and the bit error rate (BER) was achieved. The results obtained showed that the extraordinary mode had a higher availability throughout the day (15% more availability), but a delayed spread (approximately 0.3 ms mean value), similar to those of the ordinary wave. Furthermore, an improvement of up to 4 dB was achieved with the usage of polarization diversity, thus reducing transmission errors.
- Open Access
NVIS Multicarrier Modulations for Remote-Sensor Applications
The number of Internet of Things (IoT) devices has experienced a large growth during the last decade, as well as the data volume gathered from remote sensors. Satellites are still a suitable communication method and may be preferable for a remote ubiquitous sensor network (USN), which sometimes are located in places without much communications infrastructure where coverage is the principal drawback. Alternatively, the proposed solution for this article aims at a near-vertical incidence skywave (NVIS) channel for high frequencies (HF) with a low-cost platform, allowing a low-power transmissions coverage area up to 250 km for USN. The HF standards are focused on generic communication channels not being robust for NVIS communications. In this article we study and test an alternative based on orthogonal frequency-division multiplexing (OFDM) modulations to make them more robust and less dependent on the channel NVIS communications. For that purpose, we test the HF standard modulations and a designed OFDM modulation to prove the robustness of each. This study has been tested between Barcelona and Tarragona, using different transmission power levels and modulation orders.
Open AccessNVIS Multicarrier Modulations for Remote-Sensor Applications
The number of Internet of Things (IoT) devices has experienced a large growth during the last decade, as well as the data volume gathered from remote sensors. Satellites are still a suitable communication method and may be preferable for a remote ubiquitous sensor network (USN), which sometimes are located in places without much communications infrastructure where coverage is the principal drawback. Alternatively, the proposed solution for this article aims at a near-vertical incidence skywave (NVIS) channel for high frequencies (HF) with a low-cost platform, allowing a low-power transmissions coverage area up to 250 km for USN. The HF standards are focused on generic communication channels not being robust for NVIS communications. In this article we study and test an alternative based on orthogonal frequency-division multiplexing (OFDM) modulations to make them more robust and less dependent on the channel NVIS communications. For that purpose, we test the HF standard modulations and a designed OFDM modulation to prove the robustness of each. This study has been tested between Barcelona and Tarragona, using different transmission power levels and modulation orders.
- Open Access
- 28 May 2020
- Published by MDPI AG in Applied Sciences
- Vol. 10 (11) , 3730
Ionospheric Polarization Techniques for Robust NVIS Remote Sensing Platforms
Every year more interest is focused on high frequencies (HF) communications for remote sensing platforms due to their capacity to establish links of more than 250 km without a line of sight and due to them being a low-cost alternative to satellite communications. In this article, we study the ionospheric ordinary and extraordinary waves to improve the applications of near vertical incidence skywave (NVIS) on a single input multiple output (SIMO) configuration. To obtain the results, we established a link of 95 km to test the diversity combining of ordinary and extraordinary waves by using selection combining (SC) and equal-gain combining (EGC) on a remote sensing platform. The testbench is based on digital modulation transmissions with power transmission between 3 and 100 W. The results show us the main energy per bit to noise spectral density ratio (Eb/N0) and the bit error rate (BER) differences between ordinary and extraordinary waves, SC, and EGC. To conclude, diversity techniques show us a decrease of the power transmission need, allowing for the use of compact antennas and increasing battery autonomy. Furthermore, we present three different improvement options for NVIS SIMO remote sensing platforms depending on the requirements of bitrate, power consumption, and efficiency of communication.
- 28 May 2020
- Published by MDPI AG in Applied Sciences
- Vol. 10 (11) , 3730
Open AccessIonospheric Polarization Techniques for Robust NVIS Remote Sensing Platforms
Every year more interest is focused on high frequencies (HF) communications for remote sensing platforms due to their capacity to establish links of more than 250 km without a line of sight and due to them being a low-cost alternative to satellite communications. In this article, we study the ionospheric ordinary and extraordinary waves to improve the applications of near vertical incidence skywave (NVIS) on a single input multiple output (SIMO) configuration. To obtain the results, we established a link of 95 km to test the diversity combining of ordinary and extraordinary waves by using selection combining (SC) and equal-gain combining (EGC) on a remote sensing platform. The testbench is based on digital modulation transmissions with power transmission between 3 and 100 W. The results show us the main energy per bit to noise spectral density ratio (Eb/N0) and the bit error rate (BER) differences between ordinary and extraordinary waves, SC, and EGC. To conclude, diversity techniques show us a decrease of the power transmission need, allowing for the use of compact antennas and increasing battery autonomy. Furthermore, we present three different improvement options for NVIS SIMO remote sensing platforms depending on the requirements of bitrate, power consumption, and efficiency of communication.