Pedro Romero-Aroca
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
Pedro Romero-Aroca has a total of 276 co-authors. Here are the top co-authors by number of shared publications:
- Marc Baget-Bernaldiz
Ophthalmology Service, University Hospital Sant Joan, Institut de Investigacio Sanitaria Pere Virgili (IISPV), Universitat Rovira & Virgili, 43204 Reus, Spain
26 shared publications - Aida Valls
Universitat Rovira i Virgili
25 shared publications - Antonio Moreno
Universitat Rovira i Virgili
16 shared publications - Ramon Sagarra-Alamo
Ophthalmology Service, University Hospital Sant Joan, Institut de Investigacio Sanitaria Pere Virgili (IISPV), 43204 Reus, Spain
13 shared publications - Eugeni Garcia-Curto
Institut d’Investigació Sanitària Pere Virgili, Tarragona, Catalonia, Spain
9 shared publications - Jordi Pascual-Fontanilles
ITAKA, Dept. Enginyeria Informàtica i Matemàtiques, Universitat Rovira i Virgili, Tarragona, Catalonia, Spain
8 shared publications
Latest Publications
- Open Access
CALIBER: Conditional Adversarial Learning for Intelligent Retinal Vessel Segmentation and Quantitative Biomarker Extraction
The examination of the retinal vasculature through fundus images provides a direct insight into vessel morphology, which is crucial in the diagnosis of the leading causes of blindness, such as glaucoma, diabetic retinopathy (DR), and age-related macular degeneration (AMD). However, the objective and quantitative interpretation of the retinal vasculature is not possible without a precise vessel segmentation. This paper introduces Conditional Adversarial Learning for Intelligent Retinal Vessel Segmentation and Quantitative Biomarker Extraction (CALIBER), a new deep-learning-based framework that unifies high-precision retinal vessel segmentation with automated vascular biomarker extraction and quantification. The segmentation module adapts a conditional generative adversarial network and is validated across six public datasets, achieving accuracy rates between 0.925 and 0.977, and continuous ROC-AUC scores ranging from 0.765 to 0.979. Using the segmented vasculature, we extract eight quantitative biomarkers and demonstrate significant morphological differences between healthy subjects and individuals with glaucoma, DR, and AMD. We establish exploratory out-of-sample diagnostic decision thresholds, specifically identifying fractal dimensions of 1.08 with branching angles of <87 for advanced DR, confirming that quantitative vascular phenotyping is viable through automated segmentation-based analysis. Understanding these early changes holds promise for the development of effective treatment and blindness prevention protocols. Accordingly, CALIBER merges as a valuable tool in ophthalmic diagnostics by jointly enabling a precise vessel segmentation and a biomarker extraction for the early detection of vascular changes.Open AccessCALIBER: Conditional Adversarial Learning for Intelligent Retinal Vessel Segmentation and Quantitative Biomarker Extraction
The examination of the retinal vasculature through fundus images provides a direct insight into vessel morphology, which is crucial in the diagnosis of the leading causes of blindness, such as glaucoma, diabetic retinopathy (DR), and age-related macular degeneration (AMD). However, the objective and quantitative interpretation of the retinal vasculature is not possible without a precise vessel segmentation. This paper introduces Conditional Adversarial Learning for Intelligent Retinal Vessel Segmentation and Quantitative Biomarker Extraction (CALIBER), a new deep-learning-based framework that unifies high-precision retinal vessel segmentation with automated vascular biomarker extraction and quantification. The segmentation module adapts a conditional generative adversarial network and is validated across six public datasets, achieving accuracy rates between 0.925 and 0.977, and continuous ROC-AUC scores ranging from 0.765 to 0.979. Using the segmented vasculature, we extract eight quantitative biomarkers and demonstrate significant morphological differences between healthy subjects and individuals with glaucoma, DR, and AMD. We establish exploratory out-of-sample diagnostic decision thresholds, specifically identifying fractal dimensions of 1.08 with branching angles of <87 for advanced DR, confirming that quantitative vascular phenotyping is viable through automated segmentation-based analysis. Understanding these early changes holds promise for the development of effective treatment and blindness prevention protocols. Accordingly, CALIBER merges as a valuable tool in ophthalmic diagnostics by jointly enabling a precise vessel segmentation and a biomarker extraction for the early detection of vascular changes. - Open Access
- 12 March 2026
- Published by Springer Nature in Annals of Biomedical Engineering
- Vol. 54 (9) , 1-21
DDFU-Net: A Deep Decoder-Focused U-Net Model for Retinal Lesion Segmentation
Early detection of retinal lesions helps to avoid visual loss or blindness. The main lesions associated with eye diseases include soft exudates, hard exudates, microaneurysms, and hemorrhages. However, the segmentation of these four kinds of lesions is difficult and time-consuming due to their uncertainty in size, contrast, and high inter-class similarity. To address these issues, this study presents Deep Decoder-Focused U-Net (DDFU-Net), an asymmetric dense U-Net model for automatic and accurate multi-lesion segmentation using fundus images. Our approach simultaneously segments all four kinds of retinal lesions after proving that multi-task learning yields better results than single-task learning. DDFU-Net incorporates an asymmetric design with five dense blocks in the encoder and seven dense blocks in the decoder. This design enhances feature extraction while ensuring a more refined reconstruction of lesion boundaries, particularly for small and complex structures. By allocating more layers to the decoder, the model improves segmentation accuracy by gradually restoring spatial details lost during down-sampling, mitigating over-compression, and enhancing fine-grained feature preservation. Comprehensive experiments on IDRiD and DDR datasets well demonstrate the superiority of our approach, which outperforms state-of-the-art segmentation methods. Specifically, DDFU-Net achieved a mean Area Under the Precision-Recall Curve of 54.86%, a mean Intersection Over Union of 36.96%, and mean Dice scores of 52.24% on the DDR test set. On the IDRiD test set, it achieved 66.69%, 57.31%, and 69.93%, respectively. The asymmetric structure outperforms traditional symmetric U-Nets by capturing more detailed features during encoding while reducing complexity during decoding. The proposed method can be useful to aid in the diagnosis of eye diseases, reducing the workload of specialists and improving the attention to patients.
- 12 March 2026
- Published by Springer Nature in Annals of Biomedical Engineering
- Vol. 54 (9) , 1-21
Open AccessDDFU-Net: A Deep Decoder-Focused U-Net Model for Retinal Lesion Segmentation
Early detection of retinal lesions helps to avoid visual loss or blindness. The main lesions associated with eye diseases include soft exudates, hard exudates, microaneurysms, and hemorrhages. However, the segmentation of these four kinds of lesions is difficult and time-consuming due to their uncertainty in size, contrast, and high inter-class similarity. To address these issues, this study presents Deep Decoder-Focused U-Net (DDFU-Net), an asymmetric dense U-Net model for automatic and accurate multi-lesion segmentation using fundus images. Our approach simultaneously segments all four kinds of retinal lesions after proving that multi-task learning yields better results than single-task learning. DDFU-Net incorporates an asymmetric design with five dense blocks in the encoder and seven dense blocks in the decoder. This design enhances feature extraction while ensuring a more refined reconstruction of lesion boundaries, particularly for small and complex structures. By allocating more layers to the decoder, the model improves segmentation accuracy by gradually restoring spatial details lost during down-sampling, mitigating over-compression, and enhancing fine-grained feature preservation. Comprehensive experiments on IDRiD and DDR datasets well demonstrate the superiority of our approach, which outperforms state-of-the-art segmentation methods. Specifically, DDFU-Net achieved a mean Area Under the Precision-Recall Curve of 54.86%, a mean Intersection Over Union of 36.96%, and mean Dice scores of 52.24% on the DDR test set. On the IDRiD test set, it achieved 66.69%, 57.31%, and 69.93%, respectively. The asymmetric structure outperforms traditional symmetric U-Nets by capturing more detailed features during encoding while reducing complexity during decoding. The proposed method can be useful to aid in the diagnosis of eye diseases, reducing the workload of specialists and improving the attention to patients.
- Open Access
- 20 October 2025
- Published by MDPI AG in Diagnostics
- Vol. 15 (20) , 2644
An Ensemble Model for Fundus Images to Aid in Age-Related Macular Degeneration Grading
Background: Age-related macular degeneration (AMD) is a leading cause of visual impairment in the elderly population. Periodic examinations through fundus image analysis are paramount for early diagnosis and adequate treatment. Automatic artificial intelligence algorithms have proven useful for AMD grading, with the ensemble strategies recently gaining special attention. Methods: This study presents an ensemble model that combines 2 individual models of a different nature. The first model was based on the ResNetRS architecture and supervised learning. The second model, known as RETFound, was based on a visual transformer architecture and self-supervised learning. Results: Our experiments were conducted using 149,819 fundus images from the Age-Related Eye Disease Study (AREDS) public dataset. An additional private dataset of 1679 images was used to validate our approach. The results on AREDS achieved a quadratic weighted kappa of 0.7364 and an accuracy of 66.03%, which outperforms the previous methods in the literature. Conclusions: The ensemble strategy presented in this study could be useful for the screening of AMD in a clinical setting. Consequently, eye care for AMD patients would be improved while clinical costs and workload would be reduced.
- 20 October 2025
- Published by MDPI AG in Diagnostics
- Vol. 15 (20) , 2644
Open AccessAn Ensemble Model for Fundus Images to Aid in Age-Related Macular Degeneration Grading
Background: Age-related macular degeneration (AMD) is a leading cause of visual impairment in the elderly population. Periodic examinations through fundus image analysis are paramount for early diagnosis and adequate treatment. Automatic artificial intelligence algorithms have proven useful for AMD grading, with the ensemble strategies recently gaining special attention. Methods: This study presents an ensemble model that combines 2 individual models of a different nature. The first model was based on the ResNetRS architecture and supervised learning. The second model, known as RETFound, was based on a visual transformer architecture and self-supervised learning. Results: Our experiments were conducted using 149,819 fundus images from the Age-Related Eye Disease Study (AREDS) public dataset. An additional private dataset of 1679 images was used to validate our approach. The results on AREDS achieved a quadratic weighted kappa of 0.7364 and an accuracy of 66.03%, which outperforms the previous methods in the literature. Conclusions: The ensemble strategy presented in this study could be useful for the screening of AMD in a clinical setting. Consequently, eye care for AMD patients would be improved while clinical costs and workload would be reduced.
- Open Access
- 22 September 2025
- Published by IOS Press
- p. 202-211
Constructing Fuzzy Linguistic Variables Using Crisp Decision Trees. An Application to Diabetic Retinopathy
Diabetic retinopathy is a leading cause of vision loss worldwide. Early detection and understanding of its progression are crucial for preventing serious complications. This work is based on assessing the severity of diabetic retinopathy based on features extracted from eye fundus images. We use LezioSeg, a deep learning-based computer vision model, to segment and locate retinopathy lesions. Specifically, we quantify the number of microaneurysms and hemorrhages in the four retinal quadrants. These features are then used to train Retiprogram, a multiclass classification model based on a Fuzzy Random Forest. The goal is to identify the actual level of Diabetic Retinopathy. A key challenge with Fuzzy Random Forests is the definition of appropriate linguistic variables. This paper proposes a novel method for automatically generating these fuzzy sets from decision values identified at constructing crisp decision trees from a labeled set of examples. These automatically generated fuzzy sets serve as linguistic variables for the Retiprogram. We use crisp Random Forests to evaluate the Retiprogram’s performance against both Yuan and Shaw’s fuzzification method. The proposed method outperforms the standard both in accuracy and kappa indicators.
- 22 September 2025
- Published by IOS Press
- p. 202-211
Open AccessConstructing Fuzzy Linguistic Variables Using Crisp Decision Trees. An Application to Diabetic Retinopathy
Diabetic retinopathy is a leading cause of vision loss worldwide. Early detection and understanding of its progression are crucial for preventing serious complications. This work is based on assessing the severity of diabetic retinopathy based on features extracted from eye fundus images. We use LezioSeg, a deep learning-based computer vision model, to segment and locate retinopathy lesions. Specifically, we quantify the number of microaneurysms and hemorrhages in the four retinal quadrants. These features are then used to train Retiprogram, a multiclass classification model based on a Fuzzy Random Forest. The goal is to identify the actual level of Diabetic Retinopathy. A key challenge with Fuzzy Random Forests is the definition of appropriate linguistic variables. This paper proposes a novel method for automatically generating these fuzzy sets from decision values identified at constructing crisp decision trees from a labeled set of examples. These automatically generated fuzzy sets serve as linguistic variables for the Retiprogram. We use crisp Random Forests to evaluate the Retiprogram’s performance against both Yuan and Shaw’s fuzzification method. The proposed method outperforms the standard both in accuracy and kappa indicators.
- Open Access
- 23 November 2024
- Published by MDPI AG in Journal of Clinical Medicine
- Vol. 13 (23) , 7083
Incidence of Diabetic Retinopathy in Individuals with Type 2 Diabetes: A Study Using Real-World Data
Background/Objectives: This study aimed to assess the incidence of diabetic retinopathy (DR) in patients with type 2 diabetes (T2DM) treated in primary-care settings in Catalonia, Spain, and identify key risk factors associated with DR development. Methods: A retrospective cohort study was conducted using the SIDIAP (System for Research and Development in Primary Care) database. Patients aged 30–90 with T2DM who underwent retinal screening between 2010 and 2015 were included. Multivariable Cox regression analysis was used to assess the impact of clinical variables, including HbA1c levels, diabetes duration, and comorbidities, on DR incidence. Results: This study included 146,506 patients, with a mean follow-up time of 6.96 years. During this period, 4.7% of the patients developed DR, resulting in an incidence rate of 6.99 per 1000 person-years. Higher HbA1c levels were strongly associated with an increased DR risk, with patients with HbA1c > 10% having more than four times the risk compared to those with HbA1c levels < 7% (hazard ratio: 4.23; 95% CI: 3.90–4.58). Other significant risk factors for DR included greater diabetes duration, male sex, ex-smoker status, macrovascular disease, and chronic kidney disease. In contrast, obesity appeared to be a protective factor against DR, with an HR of 0.93 (95% CI: 0.89–0.98). Conclusions: In our real-world setting, the incidence rate of DR was 6.99 per 1000 person-years. Poor glycemic control, especially HbA1c > 10%, and prolonged diabetes duration were key risk factors. Effective management of these factors is crucial in preventing DR progression. Regular retinal screenings in primary care play a vital role in early detection and reducing the DR burden for T2DM patients.
- 23 November 2024
- Published by MDPI AG in Journal of Clinical Medicine
- Vol. 13 (23) , 7083
Open AccessIncidence of Diabetic Retinopathy in Individuals with Type 2 Diabetes: A Study Using Real-World Data
Background/Objectives: This study aimed to assess the incidence of diabetic retinopathy (DR) in patients with type 2 diabetes (T2DM) treated in primary-care settings in Catalonia, Spain, and identify key risk factors associated with DR development. Methods: A retrospective cohort study was conducted using the SIDIAP (System for Research and Development in Primary Care) database. Patients aged 30–90 with T2DM who underwent retinal screening between 2010 and 2015 were included. Multivariable Cox regression analysis was used to assess the impact of clinical variables, including HbA1c levels, diabetes duration, and comorbidities, on DR incidence. Results: This study included 146,506 patients, with a mean follow-up time of 6.96 years. During this period, 4.7% of the patients developed DR, resulting in an incidence rate of 6.99 per 1000 person-years. Higher HbA1c levels were strongly associated with an increased DR risk, with patients with HbA1c > 10% having more than four times the risk compared to those with HbA1c levels < 7% (hazard ratio: 4.23; 95% CI: 3.90–4.58). Other significant risk factors for DR included greater diabetes duration, male sex, ex-smoker status, macrovascular disease, and chronic kidney disease. In contrast, obesity appeared to be a protective factor against DR, with an HR of 0.93 (95% CI: 0.89–0.98). Conclusions: In our real-world setting, the incidence rate of DR was 6.99 per 1000 person-years. Poor glycemic control, especially HbA1c > 10%, and prolonged diabetes duration were key risk factors. Effective management of these factors is crucial in preventing DR progression. Regular retinal screenings in primary care play a vital role in early detection and reducing the DR burden for T2DM patients.
- Open Access
- 18 November 2024
- Published by MDPI AG in Journal of Clinical Medicine
- Vol. 13 (22) , 6935
Two Handheld Retinograph Devices Evaluated by Ophthalmologists and an Artificial Intelligence Algorithm
Background/Objectives: Telemedicine in diabetic retinopathy (RD) screening is effective but does not reach the entire diabetes population. The use of portable cameras and artificial intelligence (AI) can help in screening diabetes. Methods: We evaluated the ability of two handheld cameras, one based on a smartphone and the other on a smartscope, to obtain images for comparison with OCT. Evaluation was carried out in two stages: the first by two retina specialists and the second using an artificial intelligence algorithm that we developed. Results: The retina specialists reported that the smartphone images required mydriasis in all cases, compared to 73.05% of the smartscope images and 71.11% of the OCT images. Images were ungradable in 27.98% of the retinographs with the smartphone and in 7.98% with the smartscope. The detection of any DR using the AI algorithm showed that the smartphone obtained lower recall values (0.89) and F1 scores (0.89) than the smartscope, with 0.99. Low results were also obtained using the smartphone to detect mild DR (146 retinographs), compared to using the smartscope (218 retinographs). Conclusions: we consider that the use of handheld devices together with AI algorithms for reading retinographs can be useful for DR screening, although the ease of image acquisition through small pupils with these devices needs to be improved.
- 18 November 2024
- Published by MDPI AG in Journal of Clinical Medicine
- Vol. 13 (22) , 6935
Open AccessTwo Handheld Retinograph Devices Evaluated by Ophthalmologists and an Artificial Intelligence Algorithm
Background/Objectives: Telemedicine in diabetic retinopathy (RD) screening is effective but does not reach the entire diabetes population. The use of portable cameras and artificial intelligence (AI) can help in screening diabetes. Methods: We evaluated the ability of two handheld cameras, one based on a smartphone and the other on a smartscope, to obtain images for comparison with OCT. Evaluation was carried out in two stages: the first by two retina specialists and the second using an artificial intelligence algorithm that we developed. Results: The retina specialists reported that the smartphone images required mydriasis in all cases, compared to 73.05% of the smartscope images and 71.11% of the OCT images. Images were ungradable in 27.98% of the retinographs with the smartphone and in 7.98% with the smartscope. The detection of any DR using the AI algorithm showed that the smartphone obtained lower recall values (0.89) and F1 scores (0.89) than the smartscope, with 0.99. Low results were also obtained using the smartphone to detect mild DR (146 retinographs), compared to using the smartscope (218 retinographs). Conclusions: we consider that the use of handheld devices together with AI algorithms for reading retinographs can be useful for DR screening, although the ease of image acquisition through small pupils with these devices needs to be improved.
- Open Access
- 25 September 2024
- Published by IOS Press
- p. 80-89
Fuzzy Logic-Based Variable Encoding for Improved Diabetic Retinopathy Prediction
Electronic Health Records (EHRs) contain valuable historical information for building clinical decision support systems. In this study, we focus on exploring novel techniques for improving the prediction of the severity degree of Diabetic Retinopathy (DR) in Diabetes Mellitus patients. In a previous paper, we evaluated the behaviour of different classifiers using the patients’ retrospective EHR data to assess their current level of DR, achieving good results. Continuing that work, we now focus on studying different methods for encoding numerical variables, in order to improve the accuracy of these predictions. We propose three normalization methods based on fuzzy sets for encoding numerical data. Because of the inherent uncertainty of medical data, using fuzzy logic to represent the numerical variables can enhance the accuracy of a classifier. The results of the experimental tests, conducted on a dataset of 2108 patients, show that for low-complexity classifiers (such as KNN or CNN) a classical fuzzification technique works the best, while for more complex architectures (like TapNet or ResNet) a fuzzy two-hot encoding gives the best performance. The final aim of the research is to build a clinical decision support system that can make an accurate and personalised prediction of DR evolution.
- 25 September 2024
- Published by IOS Press
- p. 80-89
Open AccessFuzzy Logic-Based Variable Encoding for Improved Diabetic Retinopathy Prediction
Electronic Health Records (EHRs) contain valuable historical information for building clinical decision support systems. In this study, we focus on exploring novel techniques for improving the prediction of the severity degree of Diabetic Retinopathy (DR) in Diabetes Mellitus patients. In a previous paper, we evaluated the behaviour of different classifiers using the patients’ retrospective EHR data to assess their current level of DR, achieving good results. Continuing that work, we now focus on studying different methods for encoding numerical variables, in order to improve the accuracy of these predictions. We propose three normalization methods based on fuzzy sets for encoding numerical data. Because of the inherent uncertainty of medical data, using fuzzy logic to represent the numerical variables can enhance the accuracy of a classifier. The results of the experimental tests, conducted on a dataset of 2108 patients, show that for low-complexity classifiers (such as KNN or CNN) a classical fuzzification technique works the best, while for more complex architectures (like TapNet or ResNet) a fuzzy two-hot encoding gives the best performance. The final aim of the research is to build a clinical decision support system that can make an accurate and personalised prediction of DR evolution.
- Open Access
- 10 September 2024
- Published by MDPI AG in Preprints.org
Real World Analysis of Two Handheld Retinograph, Evaluation by Ophthalmologist and an Artificial Intelligence Algorithm
(1) Telemedicine in diabetic retinopathy (RD) screening is effective but does not reach the entire diabetes population. The use of portable cameras and artificial intelligence (AI) can help diabetes screening. (2) Methods: We evaluated the ability of two handheld cameras, one based on a smartphone and the other on a smartscope, to obtain images comparing OCT. Evaluation was done in two stages, the first by two retina specialists and the second using an artificial intelligence algorithm that we developed. (3) Results: The retina specialists report that the smartphone images must mydriasis in all cases compared to 73.05% of the smartscope images and 71.11% of the OCT images. Images were ungradable in 27.98% of the retinographs with the smartphone and 7.98% with the smartscope. The detection of any-DR using the AI algorithm showed that the smartphone obtained lower recall values (0.89) and F1scores (0.89) than the smartscope, with 0.99. The detection of Mild-DR using the smartphone also obtained low results (146 retinographs) compared to the smartscope (218 retinographs). (4) Conclusions: The use of handheld devices together with AI algorithms for reading retinographs can be useful for DR screening, although these devices need to improve the ease image acquisition through small pupils.
- 10 September 2024
- Published by MDPI AG in Preprints.org
Open AccessReal World Analysis of Two Handheld Retinograph, Evaluation by Ophthalmologist and an Artificial Intelligence Algorithm
(1) Telemedicine in diabetic retinopathy (RD) screening is effective but does not reach the entire diabetes population. The use of portable cameras and artificial intelligence (AI) can help diabetes screening. (2) Methods: We evaluated the ability of two handheld cameras, one based on a smartphone and the other on a smartscope, to obtain images comparing OCT. Evaluation was done in two stages, the first by two retina specialists and the second using an artificial intelligence algorithm that we developed. (3) Results: The retina specialists report that the smartphone images must mydriasis in all cases compared to 73.05% of the smartscope images and 71.11% of the OCT images. Images were ungradable in 27.98% of the retinographs with the smartphone and 7.98% with the smartscope. The detection of any-DR using the AI algorithm showed that the smartphone obtained lower recall values (0.89) and F1scores (0.89) than the smartscope, with 0.99. The detection of Mild-DR using the smartphone also obtained low results (146 retinographs) compared to the smartscope (218 retinographs). (4) Conclusions: The use of handheld devices together with AI algorithms for reading retinographs can be useful for DR screening, although these devices need to improve the ease image acquisition through small pupils.
- Open Access
- 9 September 2024
- Published by MDPI AG in Diagnostics
- Vol. 14 (17) , 1992
Artificial Intelligence-Based Screening System for Diabetic Retinopathy in Primary Care
Background: This study aimed to test an artificial intelligence-based reading system (AIRS) capable of reading retinographies of type 2 diabetic (T2DM) patients and a predictive algorithm (DRPA) that predicts the risk of each patient with T2DM of developing diabetic retinopathy (DR). Methods: We tested the ability of the AIRS to read and classify 15,297 retinal photographs from our database of diabetics and 1200 retinal images taken with Messidor-2 into the different DR categories. We tested the DRPA in a sample of 40,129 T2DM patients. The results obtained by the AIRS and the DRPA were then compared with those provided by four retina specialists regarding sensitivity (S), specificity (SP), positive predictive value (PPV), negative predictive value (NPV), accuracy (ACC), and area under the curve (AUC). Results: The results of testing the AIRS for identifying referral DR (RDR) in our database were ACC = 98.6, S = 96.7, SP = 99.8, PPV = 99.0, NPV = 98.0, and AUC = 0.958, and in Messidor-2 were ACC = 96.78%, S = 94.64%, SP = 99.14%, PPV = 90.54%, NPV = 99.53%, and AUC = 0.918. The results of our DRPA when predicting the presence of any type of DR were ACC = 0.97, S = 0.89, SP = 0.98, PPV = 0.79, NPV = 0.98, and AUC = 0.92. Conclusions: The AIRS performed well when reading and classifying the retinographies of T2DM patients with RDR. The DRPA performed well in predicting the absence of DR based on some clinical variables.
- 9 September 2024
- Published by MDPI AG in Diagnostics
- Vol. 14 (17) , 1992
Open AccessArtificial Intelligence-Based Screening System for Diabetic Retinopathy in Primary Care
Background: This study aimed to test an artificial intelligence-based reading system (AIRS) capable of reading retinographies of type 2 diabetic (T2DM) patients and a predictive algorithm (DRPA) that predicts the risk of each patient with T2DM of developing diabetic retinopathy (DR). Methods: We tested the ability of the AIRS to read and classify 15,297 retinal photographs from our database of diabetics and 1200 retinal images taken with Messidor-2 into the different DR categories. We tested the DRPA in a sample of 40,129 T2DM patients. The results obtained by the AIRS and the DRPA were then compared with those provided by four retina specialists regarding sensitivity (S), specificity (SP), positive predictive value (PPV), negative predictive value (NPV), accuracy (ACC), and area under the curve (AUC). Results: The results of testing the AIRS for identifying referral DR (RDR) in our database were ACC = 98.6, S = 96.7, SP = 99.8, PPV = 99.0, NPV = 98.0, and AUC = 0.958, and in Messidor-2 were ACC = 96.78%, S = 94.64%, SP = 99.14%, PPV = 90.54%, NPV = 99.53%, and AUC = 0.918. The results of our DRPA when predicting the presence of any type of DR were ACC = 0.97, S = 0.89, SP = 0.98, PPV = 0.79, NPV = 0.98, and AUC = 0.92. Conclusions: The AIRS performed well when reading and classifying the retinographies of T2DM patients with RDR. The DRPA performed well in predicting the absence of DR based on some clinical variables.
- Open Access
- 26 July 2024
- Published by MDPI AG in Journal of Clinical Medicine
- Vol. 13 (15) , 4387
Evaluation of Postoperative Pain When Adding a Tibial Nerve Block to the Femoral Nerve Block for Total Knee Arthroplasty
Background: The aim of this study was to compare the postoperative analgesic efficacy when a tibial nerve block was added to the femoral nerve block for total knee arthroplasty (TKA). Methods: A total of 60 patients were randomly assigned to the experimental group (EG) or the control group (CG) in a 1:1 ratio. The thirty patients who formed the CG underwent an ultrasound-guided femoral nerve block together with neuraxial anaesthesia and the administration of opioids and NSAIDs through an intravenous elastomeric pump for the management of the postoperative pain; the other thirty, who formed the EG, underwent neuraxial anaesthesia together with femoral and tibial nerve blocks. The efficacy of the analgesic effect was evaluated based on the numerical pain rating scale (NPRS) and on the need for analgesic rescue at different time intervals within 48 h after surgery. Results: At 24 h, the mean NPRS score in the EG and CG at rest was 1.50 ± 1.19 and 1.63 ± 1.60 [U = 443.5, p = 0.113], respectively. With joint movement, the mean NPRS score was 2.80 ± 1.49 and 3.57 ± 1.79 [U = 345, p = 0.113], respectively. Ten patients in the EG [33.3%] and 24 in the CG [80%] required rescue analgesia [Phi = 0.471, p < 0.001]. At 48 h, the mean NPRS score in the EG and CG at rest was 0.33 ± 0.60 and 0.43 ± 0.72 [U = 428, p = 0.681], respectively. With movement, the mean NPRS score was 1.03 ± 0.99 in the EG and 1.60 ± 1.07 in the CG [U = 315, p = 0.038]. No patient in the EG group required rescue analgesia, while three patients in the CG [10%] did [Phi = 0.229, p = 0.076]. The mean opioid dosage in the CG was 300 mg, whereas in the EG it was 40 mg ± 62.14 [U < 0.05, p < 0.001]. Conclusions: Adding a tibial nerve block to the femoral nerve block in TKA may achieve the same analgesic efficacy within 48 h after surgery and would reduce the systematic use of opioids.
- 26 July 2024
- Published by MDPI AG in Journal of Clinical Medicine
- Vol. 13 (15) , 4387
Open AccessEvaluation of Postoperative Pain When Adding a Tibial Nerve Block to the Femoral Nerve Block for Total Knee Arthroplasty
Background: The aim of this study was to compare the postoperative analgesic efficacy when a tibial nerve block was added to the femoral nerve block for total knee arthroplasty (TKA). Methods: A total of 60 patients were randomly assigned to the experimental group (EG) or the control group (CG) in a 1:1 ratio. The thirty patients who formed the CG underwent an ultrasound-guided femoral nerve block together with neuraxial anaesthesia and the administration of opioids and NSAIDs through an intravenous elastomeric pump for the management of the postoperative pain; the other thirty, who formed the EG, underwent neuraxial anaesthesia together with femoral and tibial nerve blocks. The efficacy of the analgesic effect was evaluated based on the numerical pain rating scale (NPRS) and on the need for analgesic rescue at different time intervals within 48 h after surgery. Results: At 24 h, the mean NPRS score in the EG and CG at rest was 1.50 ± 1.19 and 1.63 ± 1.60 [U = 443.5, p = 0.113], respectively. With joint movement, the mean NPRS score was 2.80 ± 1.49 and 3.57 ± 1.79 [U = 345, p = 0.113], respectively. Ten patients in the EG [33.3%] and 24 in the CG [80%] required rescue analgesia [Phi = 0.471, p < 0.001]. At 48 h, the mean NPRS score in the EG and CG at rest was 0.33 ± 0.60 and 0.43 ± 0.72 [U = 428, p = 0.681], respectively. With movement, the mean NPRS score was 1.03 ± 0.99 in the EG and 1.60 ± 1.07 in the CG [U = 315, p = 0.038]. No patient in the EG group required rescue analgesia, while three patients in the CG [10%] did [Phi = 0.229, p = 0.076]. The mean opioid dosage in the CG was 300 mg, whereas in the EG it was 40 mg ± 62.14 [U < 0.05, p < 0.001]. Conclusions: Adding a tibial nerve block to the femoral nerve block in TKA may achieve the same analgesic efficacy within 48 h after surgery and would reduce the systematic use of opioids.