Margarita Chevalier
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Margarita Chevalier has a total of 75 co-authors. Here are the top co-authors by number of shared publications:
- Adrián Belarra
Departamento de Radiología y Rehabilitación, Facultad de Medicina, Universidad Complutense de Madrid, 28040 Madrid, Spain
5 shared publications - Diego Rosich
Physics Institute of Cantabria (IFCA-CSIC-UC), Av. de los Castros s/n, 39005 Santander, Spain
2 shared publications - Irene Hernandez-Giron
School of Physics, University College of Dublin, D04 P7W1 Dublin, Ireland
2 shared publications - Maria Castillo
Medical Physics Group, Radiology Department, Medical School, Complutense University of Madrid, Pza. Ramón y Cajal, s/n, 28040 Madrid, Spain
2 shared publications - Peter Homolka
Center for Medical Physics and Biomedical Engineering, Medical University of Vienna, 1090 Vienna, Austria
2 shared publications - Tatiana Alieva
Universidad Complutense de Madrid
2 shared publications
Latest Publications
- Open Access
- 7 September 2026
- Published by MDPI AG in Technologies
- Vol. 14 (9) , 555
Design, Fabrication, and Evaluation of a Patient-Derived 3D-Printed Anthropomorphic Breast Phantom for Performance Evaluation of Dual-Energy Subtraction in Contrast-Enhanced Mammography
An anthropomorphic patient-derived 3D-printed breast phantom (4.8 cm compressed thickness, 12.2% volumetric breast density) was evaluated for image-quality assessment in contrast-enhanced mammography (CEM). It comprises four 1.2 cm plates printed via fused filament fabrication (PLA/ABS). Five replicates of one of the plates (target) containing a glandular-tissue surrogate and four cavities for iodinated disks with same iodine area density per plate (0.2–0.5–0.75–1.0–1.5 mg/cm2) were fabricated. As a reference, a homogeneous PET-G phantom was printed. CEM images (low-energy (LE), high-energy (HE), and DES) were acquired in three different systems. Iodine signal linearity (mean pixel value, signal difference, signal-difference-to-noise ratio (SdNR)) and background cancellation (residual signals, texture cancellation coefficient (TCC), parameter β from noise power spectrum (1D-NPS)) were estimated. Evaluation across the three systems demonstrated the phantom provides appropriate measurements of iodine signal linearity (R2 > 0.986, based on mean values across the four inserts) and background cancellation (TCC:0.04–0.11). Fitting 1D-NPS = αf−β within 0.25–1.0 mm−1 yielded β values ≈ 3.1–4.0, compatible with the literature, for patient LE and HE images, whereas DES reduced β to ≈1.1. In the homogeneous phantom, β remained invariant between LE and DES images (β ≈ 0.6). The local glandular ratio at each position of the iodinated disks enabled evaluating the impact of background on iodine signal.
- 7 September 2026
- Published by MDPI AG in Technologies
- Vol. 14 (9) , 555
Open AccessDesign, Fabrication, and Evaluation of a Patient-Derived 3D-Printed Anthropomorphic Breast Phantom for Performance Evaluation of Dual-Energy Subtraction in Contrast-Enhanced Mammography
An anthropomorphic patient-derived 3D-printed breast phantom (4.8 cm compressed thickness, 12.2% volumetric breast density) was evaluated for image-quality assessment in contrast-enhanced mammography (CEM). It comprises four 1.2 cm plates printed via fused filament fabrication (PLA/ABS). Five replicates of one of the plates (target) containing a glandular-tissue surrogate and four cavities for iodinated disks with same iodine area density per plate (0.2–0.5–0.75–1.0–1.5 mg/cm2) were fabricated. As a reference, a homogeneous PET-G phantom was printed. CEM images (low-energy (LE), high-energy (HE), and DES) were acquired in three different systems. Iodine signal linearity (mean pixel value, signal difference, signal-difference-to-noise ratio (SdNR)) and background cancellation (residual signals, texture cancellation coefficient (TCC), parameter β from noise power spectrum (1D-NPS)) were estimated. Evaluation across the three systems demonstrated the phantom provides appropriate measurements of iodine signal linearity (R2 > 0.986, based on mean values across the four inserts) and background cancellation (TCC:0.04–0.11). Fitting 1D-NPS = αf−β within 0.25–1.0 mm−1 yielded β values ≈ 3.1–4.0, compatible with the literature, for patient LE and HE images, whereas DES reduced β to ≈1.1. In the homogeneous phantom, β remained invariant between LE and DES images (β ≈ 0.6). The local glandular ratio at each position of the iodinated disks enabled evaluating the impact of background on iodine signal.
- Open Access
Mimicking Tissues in 3D-Printed Radiology Phantoms: Brand, Product, and Color of Printing Filaments Matter!
Additive manufacturing enables the rapid fabrication of radiographic phantoms for X-ray and CT imaging, supporting applications such as patient simulation, dosimetry, imaging protocol optimization, and quality assurance. Polylactic acid (PLA) and acrylonitrile butadiene styrene (ABS) are among the most widely used printing polymers in phantoms; however, their X-ray attenuation properties can vary substantially among manufacturers, product lines within manufacturers, and even between colors of the same product. Cylindrical samples of 34 PLA filaments from 11 manufacturers and 13 ABS filaments from 9 manufacturers were evaluated for X-ray attenuation and energy dependence between 70 and 140 kV using a clinical CT scanner. Measured mass densities ranged from 1.17 to 1.34 g/cm3 for PLA and 1.03–1.11 g/cm3 for ABS. At 120 kV, Hounsfield unit (HU) values spanned 109 to 424 HU for PLA and −34 to 40 HU for ABS. Energy dependence, quantified as the HU at 70 kV minus HU at 140 kV, ranged from −29 to +172 HU for PLA filaments and −52 to −4 HU for ABS filaments. Identical products differing only in color showed HU variations from 90 HU at 120 kV, with no consistent pattern linking specific colors to highest or lowest attenuation. These findings demonstrate that 3D printing materials require individual characterization, as base polymer designation alone does not predict X-ray behavior accurately. The observed variability, however, enables the design of phantoms with tailored attenuation and energy-dependent contrast. Referring only to base polymers when specifying 3D printing materials for radiographic phantoms or suggesting printing materials as radiographic substitutes to mimic a specified tissue or reference material without naming the actual product, including color, is, thus, insufficient.
Open AccessMimicking Tissues in 3D-Printed Radiology Phantoms: Brand, Product, and Color of Printing Filaments Matter!
Additive manufacturing enables the rapid fabrication of radiographic phantoms for X-ray and CT imaging, supporting applications such as patient simulation, dosimetry, imaging protocol optimization, and quality assurance. Polylactic acid (PLA) and acrylonitrile butadiene styrene (ABS) are among the most widely used printing polymers in phantoms; however, their X-ray attenuation properties can vary substantially among manufacturers, product lines within manufacturers, and even between colors of the same product. Cylindrical samples of 34 PLA filaments from 11 manufacturers and 13 ABS filaments from 9 manufacturers were evaluated for X-ray attenuation and energy dependence between 70 and 140 kV using a clinical CT scanner. Measured mass densities ranged from 1.17 to 1.34 g/cm3 for PLA and 1.03–1.11 g/cm3 for ABS. At 120 kV, Hounsfield unit (HU) values spanned 109 to 424 HU for PLA and −34 to 40 HU for ABS. Energy dependence, quantified as the HU at 70 kV minus HU at 140 kV, ranged from −29 to +172 HU for PLA filaments and −52 to −4 HU for ABS filaments. Identical products differing only in color showed HU variations from 90 HU at 120 kV, with no consistent pattern linking specific colors to highest or lowest attenuation. These findings demonstrate that 3D printing materials require individual characterization, as base polymer designation alone does not predict X-ray behavior accurately. The observed variability, however, enables the design of phantoms with tailored attenuation and energy-dependent contrast. Referring only to base polymers when specifying 3D printing materials for radiographic phantoms or suggesting printing materials as radiographic substitutes to mimic a specified tissue or reference material without naming the actual product, including color, is, thus, insufficient.
- Open Access
- 30 October 2025
- Published by MDPI AG in Journal of Functional Biomaterials
- Vol. 16 (11) , 406
Corrosion-Modulating Effect of Pharmaceutical Agents in a Hybrid Coating System on Pure Magnesium
There is a knowledge gap about the effect of pharmaceutical agents on the biodegradation of Mg-based resorbable implants. The present work investigates how three common antibiotics and three anti-inflammatory drugs affect the corrosion of high-purity Mg, with and without ceramic and hybrid ceramic/polymeric coatings, using electrochemical impedance spectroscopy and hydrogen evolution tests. A Ca-P-Si-based ceramic coating is developed using plasma electrolytic oxidation (PEO), after the AC voltage and frequency parameters are optimized. A hybrid coating included a PEO and a poly(ε-caprolactone) (PCL) top layer formed by dip coating. High-purity Mg exhibited an instantaneous onset of corrosion with a corrosion rate of 90 μm/year after 24 h of immersion in a modified α-MEM. A hybrid PEO/PCL coating prevents the onset of corrosion for at least 5 h and reduces the H2 evolution during the following 90 h by two times by the precipitation of 5–40 μm thick Ca-P surface deposits. Gentamicin, naproxen, streptomycin, ciprofloxacin and paracetamol were found to be corrosion accelerators with respect to bare h.p. Mg, whereas aspirin was found to be an inhibitor. Streptomycin-functionalized PEO/PCL system exhibited an active protection mechanism, triggered upon the release of the coating and substrate cations, associated with the coating defect-blocking action of the insoluble Me(II)-streptomycin chelates.
- 30 October 2025
- Published by MDPI AG in Journal of Functional Biomaterials
- Vol. 16 (11) , 406
Open AccessCorrosion-Modulating Effect of Pharmaceutical Agents in a Hybrid Coating System on Pure Magnesium
There is a knowledge gap about the effect of pharmaceutical agents on the biodegradation of Mg-based resorbable implants. The present work investigates how three common antibiotics and three anti-inflammatory drugs affect the corrosion of high-purity Mg, with and without ceramic and hybrid ceramic/polymeric coatings, using electrochemical impedance spectroscopy and hydrogen evolution tests. A Ca-P-Si-based ceramic coating is developed using plasma electrolytic oxidation (PEO), after the AC voltage and frequency parameters are optimized. A hybrid coating included a PEO and a poly(ε-caprolactone) (PCL) top layer formed by dip coating. High-purity Mg exhibited an instantaneous onset of corrosion with a corrosion rate of 90 μm/year after 24 h of immersion in a modified α-MEM. A hybrid PEO/PCL coating prevents the onset of corrosion for at least 5 h and reduces the H2 evolution during the following 90 h by two times by the precipitation of 5–40 μm thick Ca-P surface deposits. Gentamicin, naproxen, streptomycin, ciprofloxacin and paracetamol were found to be corrosion accelerators with respect to bare h.p. Mg, whereas aspirin was found to be an inhibitor. Streptomycin-functionalized PEO/PCL system exhibited an active protection mechanism, triggered upon the release of the coating and substrate cations, associated with the coating defect-blocking action of the insoluble Me(II)-streptomycin chelates.
- Open Access
- 19 June 2025
- Published by MDPI AG in Diagnostics
- Vol. 15 (12) , 1566
An Exploration of Discrepant Recalls Between AI and Human Readers of Malignant Lesions in Digital Mammography Screening
Background: The integration of artificial intelligence (AI) in digital mammography (DM) screening holds promise for early breast cancer detection, potentially enhancing accuracy and efficiency. However, AI performance is not identical to that of human observers. We aimed to identify common morphological image characteristics of true cancers that are missed by either AI or human screening when their interpretations are discrepant. Methods: Twenty-six breast cancer-positive cases, identified from a large retrospective multi-institutional digital mammography dataset based on discrepant AI and human interpretations, were included in a reader study. Ground truth was confirmed by histopathology or ≥1-year follow-up. Fourteen radiologists assessed lesion visibility, morphological features, and likelihood of malignancy. AI performance was evaluated using receiver operating characteristic (ROC) analysis and area under the curve (AUC). The reader study results were analyzed using interobserver agreement measures and descriptive statistics. Results: AI demonstrated high discriminative capability in the full dataset, with AUCs ranging from 0.903 (95% CI: 0.862–0.944) to 0.946 (95% CI: 0.896–0.996). Cancers missed by AI had a significantly smaller median size (9.0 mm, IQR 6.5–12.0) compared to those missed by human readers (21.0 mm, IQR 10.5–41.0) (p = 0.0014). Cancers in discrepant cases were often described as having ‘low visibility’, ‘indistinct margins’, or ‘irregular shape’. Calcifications were observed in 27% of human-missed cancers (42/154) versus 18% of AI-missed cancers (38/210). A very high likelihood of malignancy was assigned in 32.5% (50/154) of human-missed cancers compared to 19.5% (41/210) of AI-missed cancers. Overall inter-rater agreement was poor to fair (p = 0.396). Furthermore, 50/154 (32.5%) human-missed cancers were deemed to have a very high likelihood of malignancy, compared to 41/210 (19.5%) AI-missed cancers (p = 0.8). Overall inter-rater agreement on the items assessed during the reader study was poor to fair (<0.40), suggesting that interpretation of the selected images was challenging. Conclusions: Lesions missed by AI were smaller and less often calcified than cancers missed by human readers. Cancers missed by AI tended to show lower levels of suspicion than those missed by human readers. While definitive conclusions are premature, the findings highlight the complementary roles of AI and human readers in mammographic interpretation.
- 19 June 2025
- Published by MDPI AG in Diagnostics
- Vol. 15 (12) , 1566
Open AccessAn Exploration of Discrepant Recalls Between AI and Human Readers of Malignant Lesions in Digital Mammography Screening
Background: The integration of artificial intelligence (AI) in digital mammography (DM) screening holds promise for early breast cancer detection, potentially enhancing accuracy and efficiency. However, AI performance is not identical to that of human observers. We aimed to identify common morphological image characteristics of true cancers that are missed by either AI or human screening when their interpretations are discrepant. Methods: Twenty-six breast cancer-positive cases, identified from a large retrospective multi-institutional digital mammography dataset based on discrepant AI and human interpretations, were included in a reader study. Ground truth was confirmed by histopathology or ≥1-year follow-up. Fourteen radiologists assessed lesion visibility, morphological features, and likelihood of malignancy. AI performance was evaluated using receiver operating characteristic (ROC) analysis and area under the curve (AUC). The reader study results were analyzed using interobserver agreement measures and descriptive statistics. Results: AI demonstrated high discriminative capability in the full dataset, with AUCs ranging from 0.903 (95% CI: 0.862–0.944) to 0.946 (95% CI: 0.896–0.996). Cancers missed by AI had a significantly smaller median size (9.0 mm, IQR 6.5–12.0) compared to those missed by human readers (21.0 mm, IQR 10.5–41.0) (p = 0.0014). Cancers in discrepant cases were often described as having ‘low visibility’, ‘indistinct margins’, or ‘irregular shape’. Calcifications were observed in 27% of human-missed cancers (42/154) versus 18% of AI-missed cancers (38/210). A very high likelihood of malignancy was assigned in 32.5% (50/154) of human-missed cancers compared to 19.5% (41/210) of AI-missed cancers. Overall inter-rater agreement was poor to fair (p = 0.396). Furthermore, 50/154 (32.5%) human-missed cancers were deemed to have a very high likelihood of malignancy, compared to 41/210 (19.5%) AI-missed cancers (p = 0.8). Overall inter-rater agreement on the items assessed during the reader study was poor to fair (<0.40), suggesting that interpretation of the selected images was challenging. Conclusions: Lesions missed by AI were smaller and less often calcified than cancers missed by human readers. Cancers missed by AI tended to show lower levels of suspicion than those missed by human readers. While definitive conclusions are premature, the findings highlight the complementary roles of AI and human readers in mammographic interpretation.
- Open Access
- 24 April 2025
- Published by Elsevier in Physica Medica
- Vol. 133, 104982
Revisiting artifact spread function evaluation in digital breast tomosynthesis systems
- 24 April 2025
- Published by Elsevier in Physica Medica
- Vol. 133, 104982
Open AccessRevisiting artifact spread function evaluation in digital breast tomosynthesis systems
- Open Access
Speckle-Based Transmission and Dark-Field Imaging for Material Analysis with a Laboratory X-Ray Source
Multimodal imaging is valuable because it can provide additional information beyond that obtained from a conventional bright-field (BF) image and can be implemented with a widely available device. In this paper, we investigate the implementation of speckle-based transmission (T) and dark-field (DF) imaging in a laboratory X-ray setup to confirm its usefulness for material analysis. Three methods for recovering T and DF images were applied to a sample composed of six materials: plastic, nylon, cardboard, cork, expanded polystyrene and foam with different absorption and scattering properties. Contrast-to-noise ratio (CNR) and linear attenuation, absorption and diffusion coefficients obtained from BF, T and DF images are studied for two object-to-detector distances (ODDs). Two analysis windows are evaluated to determine the impact of noise on the image contrast of T and DF images and the ability to retrieve material characteristics. The unified modulated pattern analysis method proves to be the most reliable among the three studied speckle-based methods. The results showed that the CNR of T and DF images increases with larger analysis windows, while linear absorption and diffusion coefficients remain constant. The CNR of T images decreases with increasing ODD due to noise, whereas the CNR of DF images exhibits more complex behaviour, due to the material-dependent reduction in DF signal with increasing ODD. The experimental results on the ODD dependence of T and DF signals are consistent with recently reported numerical simulation results of these signals. The absorption coefficients derived from T images are largely independent of the ODD and the speckle-based method used, making them a universal parameter for material discrimination. In contrast, the linear diffusion coefficients vary with the ODD, limiting their applicability to specific experimental configurations despite their notable advantages in distinguishing materials. These findings highlight that T and DF images obtained from a laboratory X-ray setup offer complementary insights, enhancing their value for material analysis.
Open AccessSpeckle-Based Transmission and Dark-Field Imaging for Material Analysis with a Laboratory X-Ray Source
Multimodal imaging is valuable because it can provide additional information beyond that obtained from a conventional bright-field (BF) image and can be implemented with a widely available device. In this paper, we investigate the implementation of speckle-based transmission (T) and dark-field (DF) imaging in a laboratory X-ray setup to confirm its usefulness for material analysis. Three methods for recovering T and DF images were applied to a sample composed of six materials: plastic, nylon, cardboard, cork, expanded polystyrene and foam with different absorption and scattering properties. Contrast-to-noise ratio (CNR) and linear attenuation, absorption and diffusion coefficients obtained from BF, T and DF images are studied for two object-to-detector distances (ODDs). Two analysis windows are evaluated to determine the impact of noise on the image contrast of T and DF images and the ability to retrieve material characteristics. The unified modulated pattern analysis method proves to be the most reliable among the three studied speckle-based methods. The results showed that the CNR of T and DF images increases with larger analysis windows, while linear absorption and diffusion coefficients remain constant. The CNR of T images decreases with increasing ODD due to noise, whereas the CNR of DF images exhibits more complex behaviour, due to the material-dependent reduction in DF signal with increasing ODD. The experimental results on the ODD dependence of T and DF signals are consistent with recently reported numerical simulation results of these signals. The absorption coefficients derived from T images are largely independent of the ODD and the speckle-based method used, making them a universal parameter for material discrimination. In contrast, the linear diffusion coefficients vary with the ODD, limiting their applicability to specific experimental configurations despite their notable advantages in distinguishing materials. These findings highlight that T and DF images obtained from a laboratory X-ray setup offer complementary insights, enhancing their value for material analysis.
- Open Access
- 1 July 2024
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 11 (04) , 043501-14
Exploring single-shot propagation and speckle based phase recovery techniques for object thickness estimation by using a polychromatic X-ray laboratory source
X-ray attenuation-based imaging has limitations to provide acceptable contrast for structures with similar absorption properties as the surrounding medium. This is a situation easily encountered in the study of biological samples as well as in the field of diagnostic radiology (i.e., mammography).1 – 3
- 1 July 2024
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 11 (04) , 043501-14
Open AccessExploring single-shot propagation and speckle based phase recovery techniques for object thickness estimation by using a polychromatic X-ray laboratory source
X-ray attenuation-based imaging has limitations to provide acceptable contrast for structures with similar absorption properties as the surrounding medium. This is a situation easily encountered in the study of biological samples as well as in the field of diagnostic radiology (i.e., mammography).1 – 3
- Open Access
Resistance to Degradation of Silk Fibroin Hydrogels Exposed to Neuroinflammatory Environments
Central nervous system (CNS) diseases represent an extreme burden with significant social and economic costs. A common link in most brain pathologies is the appearance of inflammatory components that can jeopardize the stability of the implanted biomaterials and the effectiveness of therapies. Different silk fibroin scaffolds have been used in applications related to CNS disorders. Although some studies have analyzed the degradability of silk fibroin in non-cerebral tissues (almost exclusively upon non-inflammatory conditions), the stability of silk hydrogel scaffolds in the inflammatory nervous system has not been studied in depth. In this study, the stability of silk fibroin hydrogels exposed to different neuroinflammatory contexts has been explored using an in vitro microglial cell culture and two in vivo pathological models of cerebral stroke and Alzheimer’s disease. This biomaterial was relatively stable and did not show signs of extensive degradation across time after implantation and during two weeks of in vivo analysis. This finding contrasted with the rapid degradation observed under the same in vivo conditions for other natural materials such as collagen. Our results support the suitability of silk fibroin hydrogels for intracerebral applications and highlight the potentiality of this vehicle for the release of molecules and cells for acute and chronic treatments in cerebral pathologies.
Open AccessResistance to Degradation of Silk Fibroin Hydrogels Exposed to Neuroinflammatory Environments
Central nervous system (CNS) diseases represent an extreme burden with significant social and economic costs. A common link in most brain pathologies is the appearance of inflammatory components that can jeopardize the stability of the implanted biomaterials and the effectiveness of therapies. Different silk fibroin scaffolds have been used in applications related to CNS disorders. Although some studies have analyzed the degradability of silk fibroin in non-cerebral tissues (almost exclusively upon non-inflammatory conditions), the stability of silk hydrogel scaffolds in the inflammatory nervous system has not been studied in depth. In this study, the stability of silk fibroin hydrogels exposed to different neuroinflammatory contexts has been explored using an in vitro microglial cell culture and two in vivo pathological models of cerebral stroke and Alzheimer’s disease. This biomaterial was relatively stable and did not show signs of extensive degradation across time after implantation and during two weeks of in vivo analysis. This finding contrasted with the rapid degradation observed under the same in vivo conditions for other natural materials such as collagen. Our results support the suitability of silk fibroin hydrogels for intracerebral applications and highlight the potentiality of this vehicle for the release of molecules and cells for acute and chronic treatments in cerebral pathologies.
- Open Access
- 18 April 2023
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 10 (05) , 051807-14
Multi-vendor robustness analysis of a commercial artificial intelligence system for breast cancer detection
PurposePopulation-based screening programs for the early detection of breast cancer have significantly reduced mortality in women, but they are resource intensive in terms of time, cost, and workload and still have limitations mainly due to the use of 2D imaging techniques, which may cause overlapping of tissues, and interobserver variability. Artificial intelligence (AI) systems may be a valuable tool to assist radiologist when reading and classifying mammograms based on the malignancy of the detected lesions. However, there are several factors that can influence the outcome of a mammogram and thus also the detection capability of an AI system. The aim of our work is to analyze the robustness of the diagnostic ability of an AI system designed for breast cancer detection.ApproachMammograms from a population-based screening program were scored with the AI system. The sensitivity and specificity by means of the area under the receiver operating characteristic (ROC) curve were obtained as a function of the mammography unit manufacturer, demographic characteristics, and several factors that may affect the image quality (age, breast thickness and density, compression applied, beam quality, and delivered dose).ResultsThe area under the curve (AUC) from the scoring ROC curve was 0.92 (95% confidence interval = 0.89 − 0.95). It showed no dependence with any of the parameters considered, as the differences in the AUC for different interval values were not statistically significant.ConclusionThe results suggest that the AI system analyzed in our work has a robust diagnostic capability, and that its accuracy is independent of the studied parameters.
- 18 April 2023
- Published by SPIE-Intl Soc Optical Eng in Journal of Medical Imaging
- Vol. 10 (05) , 051807-14
Open AccessMulti-vendor robustness analysis of a commercial artificial intelligence system for breast cancer detection
PurposePopulation-based screening programs for the early detection of breast cancer have significantly reduced mortality in women, but they are resource intensive in terms of time, cost, and workload and still have limitations mainly due to the use of 2D imaging techniques, which may cause overlapping of tissues, and interobserver variability. Artificial intelligence (AI) systems may be a valuable tool to assist radiologist when reading and classifying mammograms based on the malignancy of the detected lesions. However, there are several factors that can influence the outcome of a mammogram and thus also the detection capability of an AI system. The aim of our work is to analyze the robustness of the diagnostic ability of an AI system designed for breast cancer detection.ApproachMammograms from a population-based screening program were scored with the AI system. The sensitivity and specificity by means of the area under the receiver operating characteristic (ROC) curve were obtained as a function of the mammography unit manufacturer, demographic characteristics, and several factors that may affect the image quality (age, breast thickness and density, compression applied, beam quality, and delivered dose).ResultsThe area under the curve (AUC) from the scoring ROC curve was 0.92 (95% confidence interval = 0.89 − 0.95). It showed no dependence with any of the parameters considered, as the differences in the AUC for different interval values were not statistically significant.ConclusionThe results suggest that the AI system analyzed in our work has a robust diagnostic capability, and that its accuracy is independent of the studied parameters.