Ermidio Rezzonico
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ermidio Rezzonico has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Francesco Magoga
Clinic of Radiology, Imaging Institute of Southern Switzerland (IIMSI), Ente Ospedaliero Cantonale (EOC), Lugano, Switzerland
5 shared publications - Stefania Rizzo
Department of Radiology, Institute of Integrated Diagnostic of Southern Switzerland (IDISI), Ente Ospedaliero Cantonale (EOC), Lugano, Switzerland
5 shared publications - Fabio Mattiussi
Imaging Institute of Southern Switzerland (IIMSI), Ente Ospedaliero Cantonale (EOC), 6900 Lugano, Switzerland
4 shared publications - Filippo Del Grande
Integrated Diagnostics Institute of Southern Switzerland (IDISI), Ente Ospedaliero Cantonale (EOC), Lugano, Switzerland
3 shared publications - Alcide Alessandro Azzena
AAST GOM Niguarda
1 shared publication - Alcide Alessandro Azzena
Imaging Institute of Southern Switzerland (IIMSI), Ente Ospedaliero Cantonale (EOC), 6900 Lugano, Switzerland
1 shared publication
Latest Publications
- 17 September 2026
- Published by Springer Nature in La radiologia medica
- Vol. 131 (10) , 1792-1803
IgeaNet: a deep learning model for opportunistic osteoporosis screening from chest X-ray images
- 17 September 2026
- Published by Springer Nature in La radiologia medica
- Vol. 131 (10) , 1792-1803
IgeaNet: a deep learning model for opportunistic osteoporosis screening from chest X-ray images
- 10 May 2026
- Published by Elsevier in Forensic Imaging
- Vol. 45
Fish&CheckNet: A deep learning model for identifying bodies from facial CT scans – A feasibility study
- 10 May 2026
- Published by Elsevier in Forensic Imaging
- Vol. 45
Fish&CheckNet: A deep learning model for identifying bodies from facial CT scans – A feasibility study
- Open Access
- 19 September 2025
- Published by MDPI AG in Bioengineering
- Vol. 12 (9) , 997
ErisNet: A Deep Learning Model for Noise Reduction in CT Images
Background: ErisNet, a novel AI model to reduce noise in Computed Tomography images. Methods: We trained ErisNet on 23 post-mortem whole-body CT scans. We assessed the objective performance with mean square error (MSE), peak signal-to-noise ratio (PSNR), structural similarity index (SSIM) measure, visual information fidelity (VIF), edge preservation index (EPI) and noise variance (NV). We assessed the qualitative performance by six radiologists. To support the visual assessment, we placed circular regions of interest (ROI) in the vitreous body, brain, liver and spleen parenchyma and paravertebral muscle. Results: ErisNet achieved MSE 64.07 ± 46.81, PSNR 31.32 ± 3.69 dB, SSIM 0.93 ± 0.06, VIF 0.49 ± 0.09, EPI 0.97 ± 0.01 and NV 64.69 ± 46.80. The ROI analysis showed a reduction in noise: the SD of the HU decreased by 8% in the vitreous body (from 17.6 to 16.2 HU), by 18% in the brain parenchyma (from 18.85 to 15.40 HU) and by 15–19% in the liver, spleen and paravertebral muscle. The six radiologists confirmed these results by assigning high scores (scale from one to five): overall quality 4.5 ± 0.6, noise suppression/detail preservation 4.7 ± 0.5 and diagnostic confidence 4.8 ± 0.4 (p < 0.01). Conclusions: ErisNet improves the quality of CT images and shows strong potential for processing low-dose scans.
- 19 September 2025
- Published by MDPI AG in Bioengineering
- Vol. 12 (9) , 997
Open AccessErisNet: A Deep Learning Model for Noise Reduction in CT Images
Background: ErisNet, a novel AI model to reduce noise in Computed Tomography images. Methods: We trained ErisNet on 23 post-mortem whole-body CT scans. We assessed the objective performance with mean square error (MSE), peak signal-to-noise ratio (PSNR), structural similarity index (SSIM) measure, visual information fidelity (VIF), edge preservation index (EPI) and noise variance (NV). We assessed the qualitative performance by six radiologists. To support the visual assessment, we placed circular regions of interest (ROI) in the vitreous body, brain, liver and spleen parenchyma and paravertebral muscle. Results: ErisNet achieved MSE 64.07 ± 46.81, PSNR 31.32 ± 3.69 dB, SSIM 0.93 ± 0.06, VIF 0.49 ± 0.09, EPI 0.97 ± 0.01 and NV 64.69 ± 46.80. The ROI analysis showed a reduction in noise: the SD of the HU decreased by 8% in the vitreous body (from 17.6 to 16.2 HU), by 18% in the brain parenchyma (from 18.85 to 15.40 HU) and by 15–19% in the liver, spleen and paravertebral muscle. The six radiologists confirmed these results by assigning high scores (scale from one to five): overall quality 4.5 ± 0.6, noise suppression/detail preservation 4.7 ± 0.5 and diagnostic confidence 4.8 ± 0.4 (p < 0.01). Conclusions: ErisNet improves the quality of CT images and shows strong potential for processing low-dose scans.
- Open Access
Radiation Dose Reduction in Cancer Imaging with New-Model CT Scanners: A Quality of Care Evaluation
Background/Objectives: The primary aim of this study was to evaluate whether the replacement of roughly one-decade-old computed tomography (CT) scanners with new-model CT scanners were associated with an additional reduction in the radiation dose delivered to oncological patients, in a radiological setting where the optimization of protocols had already reached very low radiation doses. An exploratory secondary objective was to evaluate the potential differences in the objective image quality between the CT scans obtained before and after the installation of the new-generation CT scanners. Methods: Chest and abdominal CT examinations conducted for oncologic purposes were retrospectively selected from two time periods—prior to scanner replacement (2022) and following an upgrade (2024)—after five CT systems in our radiology department were replaced. We extracted and compared the CT dose index (CTDI) and dose length product (DLP) for each CT phase. For the objective image quality evaluation, we calculated the signal-to-noise ratio (SNR) and the contrast-to-noise ratio (CNR) at the center of the liver and the aorta. An appropriate statistical analysis was performed and a p-value < 0.05 was considered significant. Results: We included 14,601 CT acquisitions, of which 9013 (61.7%) were performed before and 5588 (38.3%) after the replacement of the CT scanners. There were significantly lower values for the CTDI and DLP with the new CT scanners compared to the old ones. The CTDI with the new CT scanners was significantly lower in all phases (p-value = 0.002 for unenhanced phase, and p < 0.001 for arterial, portal venous, and delayed phases). The DLP using the new CT scanners was significantly lower in the arterial, portal venous, and delayed phases (p < 0.001), and it was not significantly different in the unenhanced phase (p = 0.36). There was no significant difference in the SNR at the liver level (p = 0.72) or at the aorta level (p = 0.51). There was no significant difference in the CNR at the liver level (p = 0.24), whereas the CNR was higher with the new CT scanners at the aorta level (p = 0.03). Conclusions: The transition to new-model CT scanners resulted in a significant reduction in the radiation dose delivered by chest and abdomen CT scans, without compromising the objective image quality.
Open AccessRadiation Dose Reduction in Cancer Imaging with New-Model CT Scanners: A Quality of Care Evaluation
Background/Objectives: The primary aim of this study was to evaluate whether the replacement of roughly one-decade-old computed tomography (CT) scanners with new-model CT scanners were associated with an additional reduction in the radiation dose delivered to oncological patients, in a radiological setting where the optimization of protocols had already reached very low radiation doses. An exploratory secondary objective was to evaluate the potential differences in the objective image quality between the CT scans obtained before and after the installation of the new-generation CT scanners. Methods: Chest and abdominal CT examinations conducted for oncologic purposes were retrospectively selected from two time periods—prior to scanner replacement (2022) and following an upgrade (2024)—after five CT systems in our radiology department were replaced. We extracted and compared the CT dose index (CTDI) and dose length product (DLP) for each CT phase. For the objective image quality evaluation, we calculated the signal-to-noise ratio (SNR) and the contrast-to-noise ratio (CNR) at the center of the liver and the aorta. An appropriate statistical analysis was performed and a p-value < 0.05 was considered significant. Results: We included 14,601 CT acquisitions, of which 9013 (61.7%) were performed before and 5588 (38.3%) after the replacement of the CT scanners. There were significantly lower values for the CTDI and DLP with the new CT scanners compared to the old ones. The CTDI with the new CT scanners was significantly lower in all phases (p-value = 0.002 for unenhanced phase, and p < 0.001 for arterial, portal venous, and delayed phases). The DLP using the new CT scanners was significantly lower in the arterial, portal venous, and delayed phases (p < 0.001), and it was not significantly different in the unenhanced phase (p = 0.36). There was no significant difference in the SNR at the liver level (p = 0.72) or at the aorta level (p = 0.51). There was no significant difference in the CNR at the liver level (p = 0.24), whereas the CNR was higher with the new CT scanners at the aorta level (p = 0.03). Conclusions: The transition to new-model CT scanners resulted in a significant reduction in the radiation dose delivered by chest and abdomen CT scans, without compromising the objective image quality.
- Open Access
- 16 April 2025
- Published by MDPI AG in Tomography
- Vol. 11 (4) , 47
Use of Open-Source Large Language Models for Automatic Synthesis of the Entire Imaging Medical Records of Patients: A Feasibility Study
Background/Objectives: Reviewing the entire history of imaging exams of a single patient’s records is an essential step in clinical practice, but it is time and resource consuming, with potential negative effects on workflow and on the quality of medical decisions. The main objective of this study was to evaluate the applicability of three open-source large language models (LLMs) for the automatic generation of concise summaries of patient’s imaging records. Secondary objectives were to assess correlations among the LLMs and to evaluate the length reduction provided by each model. Methods: Three state-of-the-art open-source large language models were selected: Llama 3.2 11B, Mistral 7B, and Falcon 7B. Each model was given a set of radiology reports. The summaries produced by the models were evaluated by two experienced radiologists and one experienced clinical physician using standardized metrics. Results: A variable number of radiological reports (n = 12–56) from four patients were selected and evaluated. The summaries generated by the three LLM showed a good level of accuracy compared with the information contained in the original reports, with positive ratings on both clinical relevance and ease of reference. According to the experts’ evaluations, the use of the summaries generated by LLMs could help to reduce the time spent on reviewing the previous imaging examinations performed, preserving the quality of clinical data. Conclusions: Our results suggest that LLMs are able to generate summaries of the imaging history of patients, and these summaries could improve radiology workflow making it easier to manage large volumes of reports.
- 16 April 2025
- Published by MDPI AG in Tomography
- Vol. 11 (4) , 47
Open AccessUse of Open-Source Large Language Models for Automatic Synthesis of the Entire Imaging Medical Records of Patients: A Feasibility Study
Background/Objectives: Reviewing the entire history of imaging exams of a single patient’s records is an essential step in clinical practice, but it is time and resource consuming, with potential negative effects on workflow and on the quality of medical decisions. The main objective of this study was to evaluate the applicability of three open-source large language models (LLMs) for the automatic generation of concise summaries of patient’s imaging records. Secondary objectives were to assess correlations among the LLMs and to evaluate the length reduction provided by each model. Methods: Three state-of-the-art open-source large language models were selected: Llama 3.2 11B, Mistral 7B, and Falcon 7B. Each model was given a set of radiology reports. The summaries produced by the models were evaluated by two experienced radiologists and one experienced clinical physician using standardized metrics. Results: A variable number of radiological reports (n = 12–56) from four patients were selected and evaluated. The summaries generated by the three LLM showed a good level of accuracy compared with the information contained in the original reports, with positive ratings on both clinical relevance and ease of reference. According to the experts’ evaluations, the use of the summaries generated by LLMs could help to reduce the time spent on reviewing the previous imaging examinations performed, preserving the quality of clinical data. Conclusions: Our results suggest that LLMs are able to generate summaries of the imaging history of patients, and these summaries could improve radiology workflow making it easier to manage large volumes of reports.