XMXenios Milidonis

Xenios Milidonis

*This is an algorithmically generated author record based on publicly available information.

Xenios Milidonis has a total of 72 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Deep Chroma Compression of Tone-Mapped Images

    by Xenios Milidonis, Alessandro Artusi, Francesco Banterle

    Acquisition of High Dynamic Range (HDR) images is thriving due to the increasing use of smart devices and the demand for high-quality output. Extensive research has focused on developing methods for reducing the luminance range in HDR images using conventional and deep learning-based tone mapping operators to enable accurate reproduction on conventional 8- and 10-bit digital displays. However, these methods often fail to account for pixels that may lie outside the target display’s gamut, resulting in visible chromatic distortions or color clipping artifacts. Previous studies suggested that a gamut management step ensures that all pixels remain within the target gamut. However, such approaches are computationally expensive and cannot be deployed on devices with limited computational resources. We propose a generative adversarial network for fast and reliable chroma compression of HDR tone-mapped images. We design a loss function that considers the hue property of generated images to improve color accuracy and train the model on an extensive image dataset. Quantitative experiments demonstrate that the proposed model outperforms state-of-the-art image generation and enhancement networks in color accuracy, while a subjective study suggests that the generated images are on par or superior to those produced by conventional chroma compression methods in terms of visual quality. Additionally, the model achieves real-time performance, showing promising results for deployment on devices with limited computational resources.


  • Open Access

    Measurement of myocardial blood flow in atrial fibrillation using high-resolution, free-breathing in-line quantitative cardiovascular magnetic resonance

    by Richard J. Crawley, Karl-Philipp Kunze, Anmol Kaushal, Xenios Milidonis, Jack Highton, Blanca Domenech-Ximenos, Irum D. Kotadia, Can Karamanli, Nathan C.K. Wong, Robbie Murphy, et al.


  • Top Cited Papers
    Open Access

    Hybrid-Segmentor: Hybrid approach for automated fine-grained crack segmentation in civil infrastructure

    by June Moh Goo, Xenios Milidonis, Alessandro Artusi, Jan Boehm, Carlo Ciliberto


  • Open Access

    Deep chroma compression of tone-mapped images

    by Xenios Milidonis, Francesco Banterle, Alessandro Artusi

    Acquisition of high dynamic range (HDR) images is thriving due to the increasing use of smart devices and the demand for high-quality output. Extensive research has focused on developing methods for reducing the luminance range in HDR images using conventional and deep learning-based tone mapping operators to enable accurate reproduction on conventional 8 and 10-bit digital displays. However, these methods often fail to account for pixels that may lie outside the target display's gamut, resulting in visible chromatic distortions or color clipping artifacts. Previous studies suggested that a gamut management step ensures that all pixels remain within the target gamut. However, such approaches are computationally expensive and cannot be deployed on devices with limited computational resources. We propose a generative adversarial network for fast and reliable chroma compression of HDR tone-mapped images. We design a loss function that considers the hue property of generated images to improve color accuracy, and train the model on an extensive image dataset. Quantitative experiments demonstrate that the proposed model outperforms state-of-the-art image generation and enhancement networks in color accuracy, while a subjective study suggests that the generated images are on par or superior to those produced by conventional chroma compression methods in terms of visual quality. Additionally, the model achieves real-time performance, showing promising results for deployment on devices with limited computational resources.

  • Open Access

    High-resolution free-breathing automated quantitative myocardial perfusion by cardiovascular magnetic resonance for the detection of functionally significant coronary artery disease

    by R Crawley, K P Kunze, X Milidonis, J Highton, S McElroy, S M Frey, D Hoefler, C Karamanli, N C K Wong, S J Backhaus, et al.

    Current assessment of myocardial ischaemia from stress perfusion cardiovascular magnetic resonance (SP-CMR) largely relies on visual interpretation. This study investigated the use of high-resolution free-breathing SP-CMR with automated quantitative mapping in the diagnosis of coronary artery disease (CAD). Diagnostic performance was evaluated against invasive coronary angiography (ICA) with fractional flow reserve (FFR) measurement. Seven hundred and three patients were recruited for SP-CMR using the research sequence at 3 Tesla. Of those receiving ICA within 6 months, 80 patients had either FFR measurement or identification of a chronic total occlusion (CTO) with inducible perfusion defects seen on SP-CMR. Myocardial blood flow (MBF) maps were automatically generated in-line on the scanner following image acquisition at hyperaemic stress and rest, allowing myocardial perfusion reserve (MPR) calculation. Seventy-five coronary vessels assessed by FFR and 28 vessels with CTO were evaluated at both segmental and coronary territory level. Coronary territory stress MBF and MPR were reduced in FFR-positive (≤0.80) regions [median stress MBF: 1.74 (0.90–2.17) mL/min/g; MPR: 1.67 (1.10–1.89)] compared with FFR-negative regions [stress MBF: 2.50 (2.15–2.95) mL/min/g; MPR 2.35 (2.06–2.54) P < 0.001 for both]. Stress MBF ≤ 1.94 mL/min/g and MPR ≤ 1.97 accurately detected FFR-positive CAD on a per-vessel basis (area under the curve: 0.85 and 0.96, respectively; P < 0.001 for both). A novel scanner-integrated high-resolution free-breathing SP-CMR sequence with automated in-line perfusion mapping is presented which accurately detects functionally significant CAD. High-resolution free-breathing stress perfusion cardiovascular magnetic resonance (SP-CMR) with in-line automated quantitative mapping is accurate in the diagnosis of functionally significant coronary artery disease. Diagnostic performance was compared with invasive fractional flow reserve (FFR) with myocardial blood flow (MBF) at hyperaemic stress and myocardial perfusion reserve (MPR) reduced in patients with inducible ischaemia. CTO, chronic total occlusion; AUC, area under the curve.


  • Open Access

    Impact of Temporal Resolution and Methods for Correction on Cardiac Magnetic Resonance Perfusion Quantification

    by Xenios Milidonis, Muhummad Sohaib Nazir, Amedeo Chiribiri

    Background: Acquisition of magnetic resonance first‐pass perfusion images is synchronized to the patient's heart rate (HR) and governs the temporal resolution. This is inherently linked to the process of myocardial blood flow (MBF) quantification and impacts MBF accuracy but to an unclear extent.Purpose: To assess the impact of temporal resolution on quantitative perfusion and compare approaches for accounting for its variability.Study Type: Prospective phantom and retrospective clinical study.Population and Phantom: Simulations, a cardiac perfusion phantom, and 30 patients with (16, 53%) or without (14, 47%) coronary artery disease.Field Strength/Sequence: 3.0 T/2D saturation recovery spoiled gradient echo sequence.Assessment: Dynamic perfusion data were simulated for a range of reference MBF (1 mL/g/min–5 mL/g/min) and HR (30 bpm–150 bpm). Perfusion imaging was performed in patients and a phantom for different temporal resolutions. MBF and myocardial perfusion reserve (MPR) were quantified without correction for temporal resolution or following correction by either MBF scaling based on the sampling interval or data interpolation prior to quantification. Simulated data were quantified using Fermi deconvolution, truncated singular value decomposition, and one‐compartment modeling, whereas phantom and clinical data were quantified using Fermi deconvolution alone.Statistical Tests: Shapiro–Wilk tests for normality, percentage error (PE) for measuring MBF accuracy in simulations, and one‐way repeated measures analysis of variance with Bonferroni correction to compare clinical MBF and MPR. Statistical significance set at P < 0.05.Results: For Fermi deconvolution and an example simulated 1 mL/g/min, the MBF PE without correction for temporal resolution was between 55.4% and −62.7% across 30–150 bpm. PE was between −22.2% and −6.8% following MBF scaling and between −14.2% and −14.2% following data interpolation across the same HR. An interpolated HR of 240 bpm reduced PE to ≤10%. Clinical rest and stress MBF and MPR were significantly different between analyses.Data Conclusion: Accurate perfusion quantification needs to account for the variability of temporal resolution, with data interpolation prior to quantification reducing MBF variability across different resolutions.Level of Evidence: 3Technical Efficacy Stage: 1


  • Open Access

    Quantification of balanced SSFP myocardial perfusion imaging at 1.5 T: Impact of the reference image

    by Sarah McElroy, Karl P. Kunze, Xenios Milidonis, Li Huang, Muhummad Sohaib Nazir, Carl Evans, Filippo Bosio, Nabila Mughal, Pier Giorgio Masci, Radhouene Neji, et al.

    Purpose To investigate the use of a high flip-angle (HFA) balanced SSFP (bSSFP) reference image (in comparison to conventional proton density [PD]–weighted reference images) for conversion of bSSFP myocardial perfusion images into dynamic T1 maps for improved myocardial blood flow (MBF) quantification at 1.5 T. Methods The HFA-bSSFP (flip angle [FA] = 50°), PD gradient-echo (PD-GRE; FA = 5°), and PD-bSSFP (FA = 8°) reference images were acquired before a dual-sequence bSSFP perfusion acquisition. Simulations were used to study accuracy and precision of T1 and MBF quantification using the three techniques. The accuracy and precision of T1, and the precision and intersegment variability of MBF were compared among the three techniques in 8 patients under rest conditions. Results In simulations, HFA-bSSFP demonstrated improved T1/MBF precision (higher T1/MBF SD of 30%-80%/50%-100% and 30%-90%/60%-115% for PD-GRE and PD-bSSFP, respectively). Proton density–GRE and PD-bSSFP were more sensitive to effective FA than HFA-bSSFP (maximum T1/MBF errors of 13%/43%, 20%/43%, and 1%/3%, respectively). Sensitivity of all techniques (defined as T1/MBF errors) to native T1, native T2, and effective saturation efficiency were negligible (1/MBF precision (T1/MBF SD: 41 ± 19 ms/0.24 ± 0.08 mL/g/min vs PD-GRE: 48 ± 20 ms/0.29 ± 0.09 mL/g/min and PD-bSSFP: 59 ± 23 ms/0.33 ± 0.11 mL/g/min; p ≤ .02) and lower MBF intersegment variability (0.14 ± 0.09 mL/g/min vs PD-GRE: 0.21 ± 0.09 mL/g/min and PD-bSSFP: 0.20 ± 0.10 mL/g/min; p ≤ .046). Conclusion We have demonstrated the feasibility of using a HFA-bSSFP reference image for MBF quantification of bSSFP perfusion imaging at 1.5 T. Results from simulations demonstrate that the HFA-bSSFP reference image results in improved precision and reduced sensitivity to effective FA compared with conventional techniques using a PD reference image. Preliminary in vivo data acquired at rest also demonstrate improved precision and intersegment variability using the HFA-bSSFP technique compared with PD techniques; however, a clinical study in patients with coronary artery disease under stress conditions is required to determine the clinical significance of this finding.


  • Open Access

    Influence of the arterial input sampling location on the diagnostic accuracy of cardiovascular magnetic resonance stress myocardial perfusion quantification

    by Xenios Milidonis, Russell Franks, Torben Schneider, Javier Sánchez-González, Eva C. Sammut, Sven Plein, Amedeo Chiribiri


  • Open Access

    Pixel‐wise assessment of cardiovascular magnetic resonance first‐pass perfusion using a cardiac phantom mimicking transmural myocardial perfusion gradients

    by Xenios Milidonis, Muhummad Sohaib Nazir, Torben Schneider, Myles Capstick, Sita Drost, Gertjan Kok, Nikola Pelevic, Christian Poelma, Tobias Schaeffter, Amedeo Chiribiri

    Purpose: Cardiovascular magnetic resonance first‐pass perfusion for the pixel‐wise detection of coronary artery disease is rapidly becoming the clinical standard, yet no widely available method exists for its assessment and validation. This study introduces a novel phantom capable of generating spatially dependent flow values to enable assessment of new perfusion imaging methods at the pixel level.Methods: A synthetic multicapillary myocardial phantom mimicking transmural myocardial perfusion gradients was designed and manufactured with high‐precision 3D printing. The phantom was used in a stationary flow setup providing reference myocardial perfusion rates and was scanned on a 3T system. Repeated first‐pass perfusion MRI for physiological perfusion rates between 1 and 4 mL/g/min was performed using a clinical dual‐sequence technique. Fermi function‐constrained deconvolution was used to estimate pixel‐wise perfusion rate maps. Phase contrast (PC)‐MRI was used to obtain velocity measurements that were converted to perfusion rates for validation of reference values and cross‐method comparison. The accuracy of pixel‐wise maps was assessed against simulated reference maps.Results: PC‐MRI indicated excellent reproducibility in perfusion rate (coefficient of variation [CoV] 2.4‐3.5%) and correlation with reference values (R2 = 0.985) across the full physiological range. Similar results were found for first‐pass perfusion MRI (CoV 3.7‐6.2%, R2 = 0.987). Pixel‐wise maps indicated a transmural perfusion difference of 28.8‐33.7% for PC‐MRI and 23.8‐37.7% for first‐pass perfusion, matching the reference values (30.2‐31.4%).Conclusion: The unique transmural perfusion pattern in the phantom allows effective pixel‐wise assessment of first‐pass perfusion acquisition protocols and quantification algorithms before their introduction into routine clinical use.


  • Open Access

    Simultaneous 13N-Ammonia and gadolinium first-pass myocardial perfusion with quantitative hybrid PET-MR imaging: a phantom and clinical feasibility study

    by Muhummad Sohaib Nazir, Sarah-May Gould, Xenios Milidonis, Eliana Reyes, Tevfik F. Ismail, Radhouene Neji, Sébastien Roujol, Jim O’dOherty, Hui Xue, Sally F. Barrington, et al.

    Positron emission tomography (PET) is the non-invasive reference standard for myocardial blood flow (MBF) quantification. Hybrid PET-MR allows simultaneous PET and cardiac magnetic resonance (CMR) acquisition under identical experimental and physiological conditions. This study aimed to determine feasibility of simultaneous 13N-Ammonia PET and dynamic contrast-enhanced CMR MBF quantification in phantoms and healthy volunteers. Images were acquired using a 3T hybrid PET-MR scanner. Phantom study: MBF was simulated at different physiological perfusion rates and a protocol for simultaneous PET-MR perfusion imaging was developed. Volunteer study: five healthy volunteers underwent adenosine stress. 13N-Ammonia and gadolinium were administered simultaneously. PET list mode data was reconstructed using ordered subset expectation maximisation. CMR MBF was quantified using Fermi function-constrained deconvolution of arterial input function and myocardial signal. PET MBF was obtained using a one-tissue compartment model and image-derived input function. Phantom study: PET and CMR MBF measurements demonstrated high repeatability with intraclass coefficients 0.98 and 0.99, respectively. There was high correlation between PET and CMR MBF (r = 0.98, p < 0.001) and good agreement (bias − 0.85 mL/g/min; 95% limits of agreement 0.29 to − 1.98). Volunteer study: Mean global stress MBF for CMR and PET were 2.58 ± 0.11 and 2.60 ± 0.47 mL/g/min respectively. On a per territory basis, there was moderate correlation (r = 0.63, p = 0.03) and agreement (bias − 0.34 mL/g/min; 95% limits of agreement 0.49 to − 1.18). Simultaneous MBF quantification using hybrid PET-MR imaging is feasible with high test repeatability and good to moderate agreement between PET and CMR. Future studies in coronary artery disease patients may allow cross-validation of techniques.