Marlen Perez-Diaz
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Marlen Perez-Diaz has a total of 56 co-authors. Here are the top co-authors by number of shared publications:
- Rubén Orozco-Morales
Universidad Central “Marta Abreu” de Las Villas
9 shared publications - José Daniel López-Cabrera
Universidad Central “Marta Abreu” de Las Villas
4 shared publications - Yakdiel Rodriguez-Gallo
Faculty of Engineering, Don Bosco University, Soyapango, El Salvador
3 shared publications - Yusely Ruiz Gonzalez
Universidad Central “Marta Abreu” de Las Villas
3 shared publications - Jhon A. Fernández Castro
Automatic Department, Universidad Central “Marta Abreu” de las Villas, Santa Clara, Cuba
2 shared publications - Jorge Armando Portal-Díaz
Departamento de Control Automático, Facultad de Ingeniería Eléctrica, Universidad Central “Marta Abreu” de Las Villas, Santa Clara, Cuba
2 shared publications
Latest Publications
- Open Access
- 10 July 2026
- Published by MDPI AG in Preprints.org
Radiomic Characterization of Breast Tissue from Breast CT Images Obtained with Synchrotron Beams
Backgroung / Objective: Radiomics offers a powerful, non-invasive approach for extracting quantitative features to predict lesion phenotypes. This study performs a radiomic characterization of breast tissue using synchrotron radiation breast computed tomography (SR-bCT). Method: Four mastectomy samples were imaged at the Australian Synchrotron (ANSTO) using five energies (25, 32, 35, 40, and 60 keV). From 10 slices per energy for each sample, 1546 regions of interest (ROIs) were extracted across four tissue subtypes: fatty, glandular, fibrous, and microcalcified. Using Pyradiomics, 93 features were initially calculated and then reduced to 34, eliminating highly correlated variables to reduce redundancies. Seven linear regression models and a discriminant analysis evaluated subtype tissue separation and radiomic characterization across individuals and combined samples and energies with a good fit (0.81≤ R ≤ 0.98). Results: The study identified 6 robust possible biomarkers independent of sample variability and energy levels, whose mean values are significantly different among the 4 tissue subtypes (clusters, p< 0.001). The selected biomarkers were: 90th Percentile, Kurtosis, Skewness, GLCM IDM, GLDM LDE, and NGTDM Coarseness. These metrics successfully differentiated all tissue pairs (microcalcifications/fat, microcalcifications/gland, microcalcifications/fiber, fat/gland, fat/fiber and gland/fiber), with p < 0.05 inside the most general regression model and p< 0.0001 with the linear discriminant separation in clusters. Conclusion: Findings indicate that some first-order and second-order texture metrics reflecting global dependencies remain stable across experimental conditions. Conversely, fine-texture metrics are highly sensitive to sample energy changes, limiting their generalizability. These results align with successful biomarkers in mammography and validate the potential of radiomics in SR-bCT characterization.
- 10 July 2026
- Published by MDPI AG in Preprints.org
Open AccessRadiomic Characterization of Breast Tissue from Breast CT Images Obtained with Synchrotron Beams
Backgroung / Objective: Radiomics offers a powerful, non-invasive approach for extracting quantitative features to predict lesion phenotypes. This study performs a radiomic characterization of breast tissue using synchrotron radiation breast computed tomography (SR-bCT). Method: Four mastectomy samples were imaged at the Australian Synchrotron (ANSTO) using five energies (25, 32, 35, 40, and 60 keV). From 10 slices per energy for each sample, 1546 regions of interest (ROIs) were extracted across four tissue subtypes: fatty, glandular, fibrous, and microcalcified. Using Pyradiomics, 93 features were initially calculated and then reduced to 34, eliminating highly correlated variables to reduce redundancies. Seven linear regression models and a discriminant analysis evaluated subtype tissue separation and radiomic characterization across individuals and combined samples and energies with a good fit (0.81≤ R ≤ 0.98). Results: The study identified 6 robust possible biomarkers independent of sample variability and energy levels, whose mean values are significantly different among the 4 tissue subtypes (clusters, p< 0.001). The selected biomarkers were: 90th Percentile, Kurtosis, Skewness, GLCM IDM, GLDM LDE, and NGTDM Coarseness. These metrics successfully differentiated all tissue pairs (microcalcifications/fat, microcalcifications/gland, microcalcifications/fiber, fat/gland, fat/fiber and gland/fiber), with p < 0.05 inside the most general regression model and p< 0.0001 with the linear discriminant separation in clusters. Conclusion: Findings indicate that some first-order and second-order texture metrics reflecting global dependencies remain stable across experimental conditions. Conversely, fine-texture metrics are highly sensitive to sample energy changes, limiting their generalizability. These results align with successful biomarkers in mammography and validate the potential of radiomics in SR-bCT characterization.
- 7 November 2025
- Published by Springer Nature in Health and Technology
- Vol. 16 (1) , 127-135
Deep learning vs. machine learning in the detection of pulmonary nodules with chest X-rays
- 7 November 2025
- Published by Springer Nature in Health and Technology
- Vol. 16 (1) , 127-135
Deep learning vs. machine learning in the detection of pulmonary nodules with chest X-rays
- 30 April 2025
- Published by Springer Nature
- p. 340-353
Detection of Lung Nodules with Chest X-Ray and Machine Learning
- 30 April 2025
- Published by Springer Nature
- p. 340-353
Detection of Lung Nodules with Chest X-Ray and Machine Learning
- Open Access
- 26 March 2024
- Published by Frontiers Media SA in Frontiers in Medical Technology
- Vol. 6, 1362688
Use of fractals in determining the malignancy degree of lung nodules
Introduction: A Computer-Assisted Detection (CAD) System for classification into malignant-benign classes using CT images is proposed.Methods: Two methods that use the fractal dimension (FD) as a measure of the lung nodule contour irregularities (Box counting and Power spectrum) were implemented. The LIDC-IDRI database was used for this study. Of these, 100 slices belonging to 100 patients were analyzed with both methods.Results: The performance between both methods was similar with an accuracy higher than 90%. Little overlap was obtained between FD ranges for the different malignancy grades with both methods, being slightly better in Power spectrum. Box counting had one more false positive than Power spectrum.Discussion: Both methods are able to establish a boundary between the high and low malignancy degree. To further validate these results and enhance the performance of the CAD system, additional studies will be necessary.
- 26 March 2024
- Published by Frontiers Media SA in Frontiers in Medical Technology
- Vol. 6, 1362688
Open AccessUse of fractals in determining the malignancy degree of lung nodules
Introduction: A Computer-Assisted Detection (CAD) System for classification into malignant-benign classes using CT images is proposed.Methods: Two methods that use the fractal dimension (FD) as a measure of the lung nodule contour irregularities (Box counting and Power spectrum) were implemented. The LIDC-IDRI database was used for this study. Of these, 100 slices belonging to 100 patients were analyzed with both methods.Results: The performance between both methods was similar with an accuracy higher than 90%. Little overlap was obtained between FD ranges for the different malignancy grades with both methods, being slightly better in Power spectrum. Box counting had one more false positive than Power spectrum.Discussion: Both methods are able to establish a boundary between the high and low malignancy degree. To further validate these results and enhance the performance of the CAD system, additional studies will be necessary.
- 4 January 2024
- Published by Springer Nature
- p. 290-301
Methods for Beam Hardening Artifacts Reduction in CT
- 4 January 2024
- Published by Springer Nature
- p. 290-301
Methods for Beam Hardening Artifacts Reduction in CT
- Open Access
- 1 January 2024
- Published by GN1 Sistemas e Publicacoes Ltd. in ITEGAM- Journal of Engineering and Technology for Industrial Applications (ITEGAM-JETIA)
- Vol. 10 (48) , 69-74
DETECTION OF LUNG NODULES USING SUPPORT VECTOR MACHINE
Lung cancer is a disease of high mortality worldwide. Therefore, early diagnosis and treatment can save lives. Lung cancer appears as a solitary nodule on chest x-ray, which is sometimes very difficult to detect for the human eye. Therefore, developing a computer-aided diagnosis (CAD) system for the detection of lung nodules, using machine learning (ML) could have a significant impact on patient prognosis. The proposed algorithm begins by pre-processing the images to improve their quality. The lung area is then segmented by thresholding. In the next step, nodule candidates are determined using a sliding band filter and segmented by applying a threshold algorithm, based on adaptive distance (ADT). Next, the suspicious areas are processed by a support vector machine (SVM), based on 15 shape and texture characteristics. Three SVM models were trained and validated with images from a public JSRT database. The best result was obtained with the radial base model (87 % sensitivity). This performance is valued as favorable with respect to human performance. El cáncer de pulmón es una enfermedad de elevada mortalidad en todo el mundo. Por ello, su diagnóstico y tratamiento precoces pueden salvar vidas. El cáncer de pulmón aparece como un nódulo solitario en la radiografía de tórax, que a veces es muy difícil de detectar para el ojo humano. Por lo tanto, el desarrollo de un sistema de diagnóstico asistido por ordenador (CAD) para la detección de nódulos pulmonares, utilizando el aprendizaje automático (ML) podría tener un impacto significativo en el pronóstico de los pacientes. El algoritmo propuesto comienza preprocesando las imágenes para mejorar su calidad. A continuación, se segmenta el área pulmonar mediante umbralización. En el siguiente paso, se determinan los nódulos candidatos mediante un filtro de banda deslizante y se segmentan aplicando un algoritmo de umbral, basado en la distancia adaptativa (ADT). A continuación, las áreas sospechosas se procesan mediante una máquina de vectores de soporte (SVM), basándose en 15 características de forma y textura. Se entrenaron y validaron tres modelos SVM con imágenes de una base de datos pública de JSRT. El mejor resultado se obtuvo con el modelo de base radial (87 % de sensibilidad). Este rendimiento se valora como favorable con respecto al rendimiento humano. Traducción realizada con la versión gratuita del traductor DeepL.com O cancro do pulmão é uma doença de elevada mortalidade em todo o mundo. Por conseguinte, o diagnóstico e o tratamento precoces podem salvar vidas. O cancro do pulmão aparece como um nódulo solitário na radiografia do tórax, que é por vezes muito difícil de detetar pelo olho humano. Por conseguinte, o desenvolvimento de um sistema de diagnóstico assistido por computador (CAD) para a deteção de nódulos pulmonares, utilizando a aprendizagem automática (ML), pode ter um impacto significativo no prognóstico dos doentes. O algoritmo proposto começa por pré-processar as imagens para melhorar a sua qualidade. A área do pulmão é então segmentada por limiarização. No passo seguinte, os candidatos a nódulos são determinados utilizando um filtro de banda deslizante e segmentados através da aplicação de um algoritmo de limiar, baseado na distância adaptativa (ADT). Em seguida, as áreas suspeitas são processadas por uma máquina de vectores de apoio (SVM), com base em 15 características de forma e textura. Foram treinados e validados três modelos SVM com imagens de uma base de dados pública do JSRT. O melhor resultado foi obtido com o modelo de base radial (87 % de sensibilidade). Este desempenho é considerado favorável no que respeita ao desempenho humano. Traduzido com a versão gratuita do tradutor - DeepL.com
- 1 January 2024
- Published by GN1 Sistemas e Publicacoes Ltd. in ITEGAM- Journal of Engineering and Technology for Industrial Applications (ITEGAM-JETIA)
- Vol. 10 (48) , 69-74
Open AccessDETECTION OF LUNG NODULES USING SUPPORT VECTOR MACHINE
Lung cancer is a disease of high mortality worldwide. Therefore, early diagnosis and treatment can save lives. Lung cancer appears as a solitary nodule on chest x-ray, which is sometimes very difficult to detect for the human eye. Therefore, developing a computer-aided diagnosis (CAD) system for the detection of lung nodules, using machine learning (ML) could have a significant impact on patient prognosis. The proposed algorithm begins by pre-processing the images to improve their quality. The lung area is then segmented by thresholding. In the next step, nodule candidates are determined using a sliding band filter and segmented by applying a threshold algorithm, based on adaptive distance (ADT). Next, the suspicious areas are processed by a support vector machine (SVM), based on 15 shape and texture characteristics. Three SVM models were trained and validated with images from a public JSRT database. The best result was obtained with the radial base model (87 % sensitivity). This performance is valued as favorable with respect to human performance. El cáncer de pulmón es una enfermedad de elevada mortalidad en todo el mundo. Por ello, su diagnóstico y tratamiento precoces pueden salvar vidas. El cáncer de pulmón aparece como un nódulo solitario en la radiografía de tórax, que a veces es muy difícil de detectar para el ojo humano. Por lo tanto, el desarrollo de un sistema de diagnóstico asistido por ordenador (CAD) para la detección de nódulos pulmonares, utilizando el aprendizaje automático (ML) podría tener un impacto significativo en el pronóstico de los pacientes. El algoritmo propuesto comienza preprocesando las imágenes para mejorar su calidad. A continuación, se segmenta el área pulmonar mediante umbralización. En el siguiente paso, se determinan los nódulos candidatos mediante un filtro de banda deslizante y se segmentan aplicando un algoritmo de umbral, basado en la distancia adaptativa (ADT). A continuación, las áreas sospechosas se procesan mediante una máquina de vectores de soporte (SVM), basándose en 15 características de forma y textura. Se entrenaron y validaron tres modelos SVM con imágenes de una base de datos pública de JSRT. El mejor resultado se obtuvo con el modelo de base radial (87 % de sensibilidad). Este rendimiento se valora como favorable con respecto al rendimiento humano. Traducción realizada con la versión gratuita del traductor DeepL.com O cancro do pulmão é uma doença de elevada mortalidade em todo o mundo. Por conseguinte, o diagnóstico e o tratamento precoces podem salvar vidas. O cancro do pulmão aparece como um nódulo solitário na radiografia do tórax, que é por vezes muito difícil de detetar pelo olho humano. Por conseguinte, o desenvolvimento de um sistema de diagnóstico assistido por computador (CAD) para a deteção de nódulos pulmonares, utilizando a aprendizagem automática (ML), pode ter um impacto significativo no prognóstico dos doentes. O algoritmo proposto começa por pré-processar as imagens para melhorar a sua qualidade. A área do pulmão é então segmentada por limiarização. No passo seguinte, os candidatos a nódulos são determinados utilizando um filtro de banda deslizante e segmentados através da aplicação de um algoritmo de limiar, baseado na distância adaptativa (ADT). Em seguida, as áreas suspeitas são processadas por uma máquina de vectores de apoio (SVM), com base em 15 características de forma e textura. Foram treinados e validados três modelos SVM com imagens de uma base de dados pública do JSRT. O melhor resultado foi obtido com o modelo de base radial (87 % de sensibilidade). Este desempenho é considerado favorável no que respeita ao desempenho humano. Traduzido com a versão gratuita do tradutor - DeepL.com
- 16 October 2023
- Published by Springer Nature in Health and Technology
- Vol. 13 (6) , 915-923
Automatic detection of breast masses using deep learning with YOLO approach
- 16 October 2023
- Published by Springer Nature in Health and Technology
- Vol. 13 (6) , 915-923
Automatic detection of breast masses using deep learning with YOLO approach
- 26 May 2023
- Published by Springer Nature in European Radiology
- Vol. 33 (7) , 5058-5059
Use of systems with deep learning and machine learning for the detection and classification of malignant vs. benign spinal fractures with MRI: can deep/machine learning help us further for detection and characterization?
- 26 May 2023
- Published by Springer Nature in European Radiology
- Vol. 33 (7) , 5058-5059
Use of systems with deep learning and machine learning for the detection and classification of malignant vs. benign spinal fractures with MRI: can deep/machine learning help us further for detection and characterization?
- Open Access
- 10 November 2022
- Published by Springer Nature in Health and Technology
- Vol. 12 (6) , 1117-1132
New patch-based strategy for COVID-19 automatic identification using chest x-ray images
Purpose The development of a robust model for automatic identification of COVID-19 based on chest x-rays has been a widely addressed topic over the last couple of years; however, the scarcity of good quality images sets, and their limited size, have proven to be an important obstacle to obtain reliable models. In fact, models proposed so far have suffered from over-fitting erroneous features instead of learning lung features, a phenomenon known as shortcut learning. In this research, a new image classification methodology is proposed that attempts to mitigate this problem. Methods To this end, annotation by expert radiologists of a set of images was performed. The lung region was then segmented and a new classification strategy based on a patch partitioning that improves the resolution of the convolution neural network is proposed. In addition, a set of native images, used as an external evaluation set, is released. Results The best results were obtained for the 6-patch splitting variant with 0.887 accuracy, 0.85 recall and 0.848 F1score on the external validation set. Conclusion The results show that the proposed new strategy maintains similar values between internal and external validation, which gives our model generalization power, making it available for use in hospital settings.
- 10 November 2022
- Published by Springer Nature in Health and Technology
- Vol. 12 (6) , 1117-1132
Open AccessNew patch-based strategy for COVID-19 automatic identification using chest x-ray images
Purpose The development of a robust model for automatic identification of COVID-19 based on chest x-rays has been a widely addressed topic over the last couple of years; however, the scarcity of good quality images sets, and their limited size, have proven to be an important obstacle to obtain reliable models. In fact, models proposed so far have suffered from over-fitting erroneous features instead of learning lung features, a phenomenon known as shortcut learning. In this research, a new image classification methodology is proposed that attempts to mitigate this problem. Methods To this end, annotation by expert radiologists of a set of images was performed. The lung region was then segmented and a new classification strategy based on a patch partitioning that improves the resolution of the convolution neural network is proposed. In addition, a set of native images, used as an external evaluation set, is released. Results The best results were obtained for the 6-patch splitting variant with 0.887 accuracy, 0.85 recall and 0.848 F1score on the external validation set. Conclusion The results show that the proposed new strategy maintains similar values between internal and external validation, which gives our model generalization power, making it available for use in hospital settings.
- 28 December 2021
- Published by Walter de Gruyter GmbH in Biomedical Engineering / Biomedizinische Technik
- Vol. 67 (1) , 1-9
Analysis of objective quality metrics in computed tomography images affected by metal artifacts
Image quality (IQ) assessment plays an important role in the medical world. New methods to evaluate image quality have been developed, but their application in the context of computer tomography is yet limited. In this paper the performance of fifteen well-known full reference (FR) IQ metrics is compared with human judgment using images affected by metal artifacts and processed with metal artifact reduction methods from a phantom. Five region of interest with different sizes were selected. IQ was evaluated by seven experienced radiologists completely blinded to the information. To measure the correlation between FR-IQ, and the score assigned by radiologists non-parametric Spearman rank-order correlation coefficient and Kendall’s Rank-order Correlation coefficient were used; so as root mean square error and the mean absolute error to measure the prediction accuracy. Cohen’s kappa was employed with the purpose of assessing inter-observer agreement. The metrics GMSD, IWMSE, IWPSNR, WSNR and OSS-PSNR were the best ranked. Inter-observer agreement was between 0.596 and 0.954, with p<0.001 in all study. The objective scores predicted by these methods correlate consistently with the subjective evaluations. The application of this metrics will make possible a better evaluation of metal artifact reduction algorithms in future works.
- 28 December 2021
- Published by Walter de Gruyter GmbH in Biomedical Engineering / Biomedizinische Technik
- Vol. 67 (1) , 1-9
Analysis of objective quality metrics in computed tomography images affected by metal artifacts
Image quality (IQ) assessment plays an important role in the medical world. New methods to evaluate image quality have been developed, but their application in the context of computer tomography is yet limited. In this paper the performance of fifteen well-known full reference (FR) IQ metrics is compared with human judgment using images affected by metal artifacts and processed with metal artifact reduction methods from a phantom. Five region of interest with different sizes were selected. IQ was evaluated by seven experienced radiologists completely blinded to the information. To measure the correlation between FR-IQ, and the score assigned by radiologists non-parametric Spearman rank-order correlation coefficient and Kendall’s Rank-order Correlation coefficient were used; so as root mean square error and the mean absolute error to measure the prediction accuracy. Cohen’s kappa was employed with the purpose of assessing inter-observer agreement. The metrics GMSD, IWMSE, IWPSNR, WSNR and OSS-PSNR were the best ranked. Inter-observer agreement was between 0.596 and 0.954, with p<0.001 in all study. The objective scores predicted by these methods correlate consistently with the subjective evaluations. The application of this metrics will make possible a better evaluation of metal artifact reduction algorithms in future works.