Tudor Nedelcu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Tudor Nedelcu has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Peter Corcoran
Department of Electronic Engineering, College of Science and Engineering, University of Galway, Galway, Ireland
5 shared publications - Shejin Thavalengal
National University of Ireland Galway, Galway, Ireland
4 shared publications Claudia Costache
National University of Ireland
2 shared publications- Rafaela Carvalho
Fraunhofer Portugal AICOS
2 shared publications - Shabab Bazrafkan
University of Antwerp
2 shared publications - Ana C. Morgado
Fraunhofer Portugal AICOS, Rua Alfredo Allen, 4200-135 Porto, Portugal
1 shared publication
Latest Publications
- Open Access
- 24 December 2021
- Published by MDPI AG in Diagnostics
- Vol. 12 (1) , 36
Integrating Domain Knowledge into Deep Learning for Skin Lesion Risk Prioritization to Assist Teledermatology Referral
Teledermatology has developed rapidly in recent years and is nowadays an essential tool for early diagnosis. In this work, we aim to improve existing Teledermatology processes for skin lesion diagnosis by developing a deep learning approach for risk prioritization with a dataset of retrospective data from referral requests of the Portuguese National Health System. Given the high complexity of this task, we propose a new prioritization pipeline guided and inspired by domain knowledge. We explored automatic lesion segmentation and tested different learning schemes, namely hierarchical classification and curriculum learning approaches, optionally including additional patient metadata. The final priority level prediction can then be obtained by combining predicted diagnosis and a baseline priority level accounting for explicit expert knowledge. In both the differential diagnosis and prioritization branches, lesion segmentation with 30% tolerance for contextual information was shown to improve classification when compared with a flat baseline model trained on original images; furthermore, the addition of patient information was not beneficial for most experiments. Curriculum learning delivered better results than a flat or hierarchical approach. The combination of diagnosis information and a knowledge map, created in collaboration with dermatologists, together with the priority achieved interesting results (best macro F1 of 43.93% for a validated test set), paving the way for new data-centric and knowledge-driven approaches.
- 24 December 2021
- Published by MDPI AG in Diagnostics
- Vol. 12 (1) , 36
Open AccessIntegrating Domain Knowledge into Deep Learning for Skin Lesion Risk Prioritization to Assist Teledermatology Referral
Teledermatology has developed rapidly in recent years and is nowadays an essential tool for early diagnosis. In this work, we aim to improve existing Teledermatology processes for skin lesion diagnosis by developing a deep learning approach for risk prioritization with a dataset of retrospective data from referral requests of the Portuguese National Health System. Given the high complexity of this task, we propose a new prioritization pipeline guided and inspired by domain knowledge. We explored automatic lesion segmentation and tested different learning schemes, namely hierarchical classification and curriculum learning approaches, optionally including additional patient metadata. The final priority level prediction can then be obtained by combining predicted diagnosis and a baseline priority level accounting for explicit expert knowledge. In both the differential diagnosis and prioritization branches, lesion segmentation with 30% tolerance for contextual information was shown to improve classification when compared with a flat baseline model trained on original images; furthermore, the addition of patient information was not beneficial for most experiments. Curriculum learning delivered better results than a flat or hierarchical approach. The combination of diagnosis information and a knowledge map, created in collaboration with dermatologists, together with the priority achieved interesting results (best macro F1 of 43.93% for a validated test set), paving the way for new data-centric and knowledge-driven approaches.
- Open Access
- 1 January 2021
- Published by Springer Nature
- p. 27-38
Multimodal Multi-tasking for Skin Lesion Classification Using Deep Neural Networks
Skin cancer is one of the most common types of cancer and, with its increasing incidence, accurate early diagnosis is crucial to improve prognosis of patients. In the process of visual inspection, dermatologists follow specific dermoscopic algorithms and identify important features to provide a diagnosis. This process can be automated as such characteristics can be extracted by computer vision techniques. Although deep neural networks can extract useful features from digital images for skin lesion classification, performance can be improved by providing additional information. The extracted pseudo-features can be used as input (multimodal) or output (multi-tasking) to train a robust deep learning model. This work investigates the multimodal and multi-tasking techniques for more efficient training, given the single optimization of several related tasks in the latter, and generation of better diagnosis predictions. Additionally, the role of lesion segmentation is also studied. Results show that multi-tasking improves learning of beneficial features which lead to better predictions, and pseudo-features inspired by the ABCD rule provide readily available helpful information about the skin lesion.
- 1 January 2021
- Published by Springer Nature
- p. 27-38
Open AccessMultimodal Multi-tasking for Skin Lesion Classification Using Deep Neural Networks
Skin cancer is one of the most common types of cancer and, with its increasing incidence, accurate early diagnosis is crucial to improve prognosis of patients. In the process of visual inspection, dermatologists follow specific dermoscopic algorithms and identify important features to provide a diagnosis. This process can be automated as such characteristics can be extracted by computer vision techniques. Although deep neural networks can extract useful features from digital images for skin lesion classification, performance can be improved by providing additional information. The extracted pseudo-features can be used as input (multimodal) or output (multi-tasking) to train a robust deep learning model. This work investigates the multimodal and multi-tasking techniques for more efficient training, given the single optimization of several related tasks in the latter, and generation of better diagnosis predictions. Additionally, the role of lesion segmentation is also studied. Results show that multi-tasking improves learning of beneficial features which lead to better predictions, and pseudo-features inspired by the ABCD rule provide readily available helpful information about the skin lesion.
- 12 May 2017
- Published by Springer Nature in Lecture Notes in Computer Science
- p. 304-312
Pupil Light Reflex Mitigation Using Non-linear Image Warping
- 12 May 2017
- Published by Springer Nature in Lecture Notes in Computer Science
- p. 304-312
Pupil Light Reflex Mitigation Using Non-linear Image Warping
- 1 January 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 44, 217-220
Deep learning for facial expression recognition: A step closer to a smartphone that knows your moods
By growing the capacity and processing power of the handheld devices nowadays, a wide range of capabilities can be implemented in these devices to make them more intelligent and user friendly. Determining the mood of the user can be used in order to provide suitable reactions from the device in different conditions. One of the most studied ways of mood detection is by using facial expressions, which is still one of the challenging fields in pattern recognition and machine learning science. Deep Neural Networks (DNN) have been widely used in order to overcome the difficulties in facial expression classification. In this paper it is shown that the classification accuracy is significantly lower when the network is trained with one database and tested with a different database. A solution for obtaining a general and robust network is given as well.
- 1 January 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 44, 217-220
Deep learning for facial expression recognition: A step closer to a smartphone that knows your moods
By growing the capacity and processing power of the handheld devices nowadays, a wide range of capabilities can be implemented in these devices to make them more intelligent and user friendly. Determining the mood of the user can be used in order to provide suitable reactions from the device in different conditions. One of the most studied ways of mood detection is by using facial expressions, which is still one of the challenging fields in pattern recognition and machine learning science. Deep Neural Networks (DNN) have been widely used in order to overcome the difficulties in facial expression classification. In this paper it is shown that the classification accuracy is significantly lower when the network is trained with one database and tested with a different database. A solution for obtaining a general and robust network is given as well.
- 19 July 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 62 (2) , 95-102
Iris liveness detection for next generation smartphones
This paper presents a novel liveness detection method that exploits the acquisition workflow for iris biometrics on smartphones using a hybrid visible (RGB)/near infra-red (NIR) sensor. These devices are able to capture both RGB and NIR images of the eye and iris region in synchronization. This multi-spectral information is mapped into a discrete feature space. An intermediate classifier which uses a distance metric close to Jenson-Shannon divergence is employed to classify the incoming image. Further, a fast, multi-frame pupil localization technique using one-dimensional processing of the eye region is proposed and evaluated. This is used to analyze the pupil characteristics of the images classified as 'live' in the previous stage. It is shown that such an analysis could detect presentation attacks, even with a 3-D face model made of materials that has properties similar to human skin and the ocular region1.
- 19 July 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 62 (2) , 95-102
Iris liveness detection for next generation smartphones
This paper presents a novel liveness detection method that exploits the acquisition workflow for iris biometrics on smartphones using a hybrid visible (RGB)/near infra-red (NIR) sensor. These devices are able to capture both RGB and NIR images of the eye and iris region in synchronization. This multi-spectral information is mapped into a discrete feature space. An intermediate classifier which uses a distance metric close to Jenson-Shannon divergence is employed to classify the incoming image. Further, a fast, multi-frame pupil localization technique using one-dimensional processing of the eye region is proposed and evaluated. This is used to analyze the pupil characteristics of the images classified as 'live' in the previous stage. It is shown that such an analysis could detect presentation attacks, even with a 3-D face model made of materials that has properties similar to human skin and the ocular region1.
- 1 January 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 512-513
Finger vein biometric: Smartphone footprint prototype with vein map extraction using computational imaging techniques
A new vein structure based biometric approach is introduced in this paper. The idea is to use the finger vein structure in the intermediate phalange to identify or authenticate individuals. The concept of the hardware is to be implementable in handheld devices. Hardware configuration and software implementation are presented. The idea of temporal median filter was used in order to fuse finger images and Gabor filter bank was implemented to extract the finger vein maps.
- 1 January 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 512-513
Finger vein biometric: Smartphone footprint prototype with vein map extraction using computational imaging techniques
A new vein structure based biometric approach is introduced in this paper. The idea is to use the finger vein structure in the intermediate phalange to identify or authenticate individuals. The concept of the hardware is to be implementable in handheld devices. Hardware configuration and software implementation are presented. The idea of temporal median filter was used in order to fuse finger images and Gabor filter bank was implemented to extract the finger vein maps.
- 1 January 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 551-552
Hybrid visible + near infrared color filter array for handheld consumer imaging applications
Joint acquisition of multispectral images raised a lot of interest due to the applications that can be performed using a wide spectrum of information. Images acquired simultaneously from two spectrums (visible and near infrared) can be fused to generate enhanced images (de-hazing, portrait enhancement, low light, beautification, enhancement). The multispectral images can be combined or used separately to perform certain task. For iris recognition, it's desirable to acquire good images from the near infrared spectrum. In this paper we will propose a hybrid color filter array, which is able to acquire a good quality image form near infrared spectrum and satisfactory quality for the visible one.
- 1 January 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 551-552
Hybrid visible + near infrared color filter array for handheld consumer imaging applications
Joint acquisition of multispectral images raised a lot of interest due to the applications that can be performed using a wide spectrum of information. Images acquired simultaneously from two spectrums (visible and near infrared) can be fused to generate enhanced images (de-hazing, portrait enhancement, low light, beautification, enhancement). The multispectral images can be combined or used separately to perform certain task. For iris recognition, it's desirable to acquire good images from the near infrared spectrum. In this paper we will propose a hybrid color filter array, which is able to acquire a good quality image form near infrared spectrum and satisfactory quality for the visible one.
- 1 September 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 252-255
Hybrid sensor visible + near infrared demosaicing
The capability of a device to acquire images from a wide spectrum (visible and near infrared) have raised a lot of interest due to the extended operation that can be performed. There are specific application for devices which are able to acquire simultaneously two images from the wide spectrum (beautification, image enhancement, and de-hazing, low light enhancement). In this paper we will describe a method for demosaicing images with respect to the hybrid color filter array pattern.
- 1 September 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 252-255
Hybrid sensor visible + near infrared demosaicing
The capability of a device to acquire images from a wide spectrum (visible and near infrared) have raised a lot of interest due to the extended operation that can be performed. There are specific application for devices which are able to acquire simultaneously two images from the wide spectrum (beautification, image enhancement, and de-hazing, low light enhancement). In this paper we will describe a method for demosaicing images with respect to the hybrid color filter array pattern.
- 1 August 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1925-1929
Alternate multibit embedding method for reversible watermarking
This paper proposes a new embedding scheme for multibit difference expansion reversible watermarking. The prediction error expansion (PEE) schemes expand n times the difference in order to embed up to log2 n bpp. For natural images, this capacity cannot be achieved because overflow or underflow is generated by the embedding process. The proposed method aims to increase the capacity of the embedded information by using a different embedding procedure when the classical one fail. Although the proposed embedding method introduces larger distortion than classical procedure, the experimental results show that the proposed scheme provide an increase of the embedding capacity and outperforms the classical method regarding the image quality with respect to capacity. Experimental results using the classical and proposed multibit difference expansion based on the MED predictor are provided.
- 1 August 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1925-1929
Alternate multibit embedding method for reversible watermarking
This paper proposes a new embedding scheme for multibit difference expansion reversible watermarking. The prediction error expansion (PEE) schemes expand n times the difference in order to embed up to log2 n bpp. For natural images, this capacity cannot be achieved because overflow or underflow is generated by the embedding process. The proposed method aims to increase the capacity of the embedded information by using a different embedding procedure when the classical one fail. Although the proposed embedding method introduces larger distortion than classical procedure, the experimental results show that the proposed scheme provide an increase of the embedding capacity and outperforms the classical method regarding the image quality with respect to capacity. Experimental results using the classical and proposed multibit difference expansion based on the MED predictor are provided.
- 1 July 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Alternate embedding method for difference expansion reversible watermarking
This paper proposes a new embedding scheme for difference expansion reversible watermarking. The proposed method aims to increase the capacity of the embedded information by using a different embedding procedure when the classical one fails because of overflow or underflow. The alternate embedding introduces larger distortion than classical procedure. Experimental results for the case of two-layer prediction-error expansion based on the rhombus predictor are provided. The experimental results show that the proposed scheme can provide a slight increase of the embedding capacity.
- 1 July 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Alternate embedding method for difference expansion reversible watermarking
This paper proposes a new embedding scheme for difference expansion reversible watermarking. The proposed method aims to increase the capacity of the embedded information by using a different embedding procedure when the classical one fails because of overflow or underflow. The alternate embedding introduces larger distortion than classical procedure. Experimental results for the case of two-layer prediction-error expansion based on the rhombus predictor are provided. The experimental results show that the proposed scheme can provide a slight increase of the embedding capacity.