Andrea Corradini
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Andrea Corradini has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Bernadectus Yudi Dwiandiyanta
Informatics Department, Faculty of Industrial Technology, Universitas Atma Jaya Yogyakarta, Daerah Istimewa Yogyakarta
1 shared publication - Joseph Passant1 shared publication
- Martinus Maslim
Universitas Atma Jaya Yogyakarta, Sleman, Indonesia
1 shared publication - Mary Idamkina
University of Liverpool
1 shared publication - Ronald Andrew Stockwell1 shared publication
- Tuhin Banerjee
Canon EMEA
1 shared publication
Latest Publications
- Open Access
- 26 May 2026
- Published by AccScience Publishing in Artificial Intelligence in Health
- Vol. 3 (4) , 026030003
A clinically weighted loss function for risk-sensitive blood glucose forecasting
Blood glucose forecasting models have the potential to improve glycemic control and reduce the burden of diabetes management. However, most prior work has focused on optimizing statistical accuracy, which does not necessarily align with clinical safety and efficacy. This study introduces a clinically weighted root mean squared error (CW-RMSE) loss function that incorporates clinical risk into model optimization by weighting prediction errors according to continuous glucose error grid analysis (CG-EGA) classification and glycemic range to reduce clinically harmful errors. WinFormer, an Informer-based prediction model employing windowed attention, was trained using the REPLACE-BG dataset and either root mean squared error (RMSE) or CW-RMSE loss functions to predict blood glucose values over a 120-min prediction horizon. Compared to RMSE loss, CW-RMSE loss produced significant improvements in CG-EGA classification in hypoglycemia (accurate predictions: 83.2% to 88.0%; erroneous predictions: 13.2% to 8.1%) while maintaining performance for predictions of values in euglycemic (accurate predictions: 92.6% to 92.4%; erroneous predictions: 2.7% to 2.4%) and hyperglycemic ranges (accurate predictions: 91.2% to 90.6%; erroneous predictions: 4.1% to 4.1%). In comparison to RMSE loss, CW-RMSE loss also led to significant improvements in RMSE and mean absolute error (MAE) in hypoglycemia (RMSE: 39.8 mg/dL to 29.8 mg/dL; MAE: 26.6 mg/dL to 16.4 mg/dL), but worse RMSE and MAE in the euglycemia range (RMSE: 31.8 mg/dL to 33.8 mg/dL; MAE: 23.1 mg/dL to 25.1 mg/dL) and the hyperglycemia range (RMSE: 44.9 mg/dL to 52.3 mg/dL; MAE: 30.0 mg/dL to 34.8 mg/dL). These findings demonstrate that cost-sensitive learning can effectively guide blood glucose forecasting models toward more clinically meaningful and risk-aware optimization.
- 26 May 2026
- Published by AccScience Publishing in Artificial Intelligence in Health
- Vol. 3 (4) , 026030003
Open AccessA clinically weighted loss function for risk-sensitive blood glucose forecasting
Blood glucose forecasting models have the potential to improve glycemic control and reduce the burden of diabetes management. However, most prior work has focused on optimizing statistical accuracy, which does not necessarily align with clinical safety and efficacy. This study introduces a clinically weighted root mean squared error (CW-RMSE) loss function that incorporates clinical risk into model optimization by weighting prediction errors according to continuous glucose error grid analysis (CG-EGA) classification and glycemic range to reduce clinically harmful errors. WinFormer, an Informer-based prediction model employing windowed attention, was trained using the REPLACE-BG dataset and either root mean squared error (RMSE) or CW-RMSE loss functions to predict blood glucose values over a 120-min prediction horizon. Compared to RMSE loss, CW-RMSE loss produced significant improvements in CG-EGA classification in hypoglycemia (accurate predictions: 83.2% to 88.0%; erroneous predictions: 13.2% to 8.1%) while maintaining performance for predictions of values in euglycemic (accurate predictions: 92.6% to 92.4%; erroneous predictions: 2.7% to 2.4%) and hyperglycemic ranges (accurate predictions: 91.2% to 90.6%; erroneous predictions: 4.1% to 4.1%). In comparison to RMSE loss, CW-RMSE loss also led to significant improvements in RMSE and mean absolute error (MAE) in hypoglycemia (RMSE: 39.8 mg/dL to 29.8 mg/dL; MAE: 26.6 mg/dL to 16.4 mg/dL), but worse RMSE and MAE in the euglycemia range (RMSE: 31.8 mg/dL to 33.8 mg/dL; MAE: 23.1 mg/dL to 25.1 mg/dL) and the hyperglycemia range (RMSE: 44.9 mg/dL to 52.3 mg/dL; MAE: 30.0 mg/dL to 34.8 mg/dL). These findings demonstrate that cost-sensitive learning can effectively guide blood glucose forecasting models toward more clinically meaningful and risk-aware optimization.
- Open Access
- 28 October 2025
- Published by Universitas Airlangga in Journal of Information Systems Engineering and Business Intelligence
- Vol. 11 (3) , 500-513
Diabetic Retinopathy Fundus Image Classification Using Self-Organizing Maps
Background: Diabetic retinopathy (DR) is a condition that impairs the blood vessels in the retina, resulting in vision loss ranging from temporary to permanent blindness. It commonly affects individuals diagnosed with diabetes mellitus (DM). Fundoscopy is a technique used to identify DR by examining the fundus of the eye during an eye examination. This process is time-consuming and can be expensive. Objective: This study aimed to examine the identification of DR using digital image processing methods. Methods: The self-organizing map (SOM) artificial neural network was employed. Diabetic retinopathy will be categorized according to its severity, including normal, mild, moderate, or severe. This classification considers the quantity of exudates and microaneurysms and the blood vessel structure in the fundus image. The dataset used in this investigation comprised 1000 fundus images acquired from the MESSIDOR ophthalmology database. Results: The findings indicate that the SOM approach achieves a training accuracy of 72% and a testing accuracy of 62%. Conclusion: The DR severity classification system can effectively extract DR-related features by segmenting exudates, blood vessels, and microaneurysms from funduscopic images during training, testing, and evaluation. Keywords: Diabetic Retinopathy, Self-Organizing Map, Fundus Image Classification, Digital Image Processing
- 28 October 2025
- Published by Universitas Airlangga in Journal of Information Systems Engineering and Business Intelligence
- Vol. 11 (3) , 500-513
Open AccessDiabetic Retinopathy Fundus Image Classification Using Self-Organizing Maps
Background: Diabetic retinopathy (DR) is a condition that impairs the blood vessels in the retina, resulting in vision loss ranging from temporary to permanent blindness. It commonly affects individuals diagnosed with diabetes mellitus (DM). Fundoscopy is a technique used to identify DR by examining the fundus of the eye during an eye examination. This process is time-consuming and can be expensive. Objective: This study aimed to examine the identification of DR using digital image processing methods. Methods: The self-organizing map (SOM) artificial neural network was employed. Diabetic retinopathy will be categorized according to its severity, including normal, mild, moderate, or severe. This classification considers the quantity of exudates and microaneurysms and the blood vessel structure in the fundus image. The dataset used in this investigation comprised 1000 fundus images acquired from the MESSIDOR ophthalmology database. Results: The findings indicate that the SOM approach achieves a training accuracy of 72% and a testing accuracy of 62%. Conclusion: The DR severity classification system can effectively extract DR-related features by segmenting exudates, blood vessels, and microaneurysms from funduscopic images during training, testing, and evaluation. Keywords: Diabetic Retinopathy, Self-Organizing Map, Fundus Image Classification, Digital Image Processing
- 14 November 2024
- Published by Springer Nature
- p. 349-360
A Deep-Learning Approach for the Identification of New Subtypes of Lung Cancer
- 14 November 2024
- Published by Springer Nature
- p. 349-360
A Deep-Learning Approach for the Identification of New Subtypes of Lung Cancer
- 15 November 2023
- Published by Springer Nature
- p. 368-380
A Machine Learning Approach to Predict Cyclists’ Functional Threshold Power
- 15 November 2023
- Published by Springer Nature
- p. 368-380
A Machine Learning Approach to Predict Cyclists’ Functional Threshold Power