Possibilities of Dixon sequences in magnetic resonance imaging for fat fraction quantification: a phantom study Possibilities of Dixon sequences in magnetic resonance imaging for fat fraction quantification: a phantom study

Full Text

BACKGROUND

There is a growing interest in fat fraction (FF) quantification using magnetic resonance imaging (MRI), computed tomography, and ultrasound imaging, primarily in the context of diagnosing hepatic steatosis. FF quantification on MRI is typically performed using Dixon pulse sequences, which are incorporated into the standard functionality of most modern MRI scanners. Their advantages include not only the ability to accurately separate water and fat signals but also the simultaneous acquisition of four images in a single short scan: fat-only (Fat), water-only (Water), in-phase (In-Phase), and out-of-phase (Out-of-Phase) images [1, 2]. This feature of the pulse sequence enables precise and visually demonstrative detection of fat within organ parenchyma or pathological lesions.

Two-point Dixon sequences are available, such as LAVA® (General Electric Healthcare, United States), mDIXON® (Philips Healthcare, Netherlands), as well as multipoint variants (e.g., VIBE®, Siemens Healthcare, Germany). All are based on Dixon techniques that use different echo-time (TE) values depending on scanner manufacturer and model. Specialized automated post-processing solutions have also been developed, including IDEAL IQ® (General Electric Healthcare, United States), Liver Lab® (Siemens Healthcare, Germany), and QUANT® (Philips Healthcare, Netherlands). These solutions enable automated calculation of proton density fat fraction (PDFF). However, these modules are not available on all scanners, as they are often optional add-ons purchased separately [3]. In such situations, radiologists can still calculate FF manually using standard Dixon pulse sequence data and formulas based on signal intensity (SI) characteristics. FF is calculated using in-phase and out-of-phase images, as well as water-weighted and fat-weighted images [1, 4].

However, quantitative values obtained in this manner may vary considerably depending on acquisition order, scanner manufacturer and model, magnetic field strength, and other technical parameters. The scientific data contains reports of substantial errors in FF estimation in certain situations [2, 5].

Thus, the accuracy and reliability of FF quantification during MRI are essential for the reliable detection of pathological changes [6]. One effective approach to achieving this is the validation of Dixon pulse sequences through phantom modeling on the specific MRI scanner being used [7, 8].

AIM

To evaluate the possibility of fat fraction quantification using standard Dixon pulse sequences through phantom modeling.

METHODS

Study Design

This was a multicenter, cross-sectional, open-label experimental study to evaluate the performance of Dixon sequences using a phantom.

Phantom Description

The experiment was conducted using a phantom developed at the Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies of the Moscow Department of Health (Fig. 1a). The phantom is an airtight acrylic cylinder containing test tubes filled with emulsions prepared with different fat-fraction concentrations: 10%, 20%, 30%, 40%, 50%, and 60% (Fig. 1b).

 

Fig. 1. Phantom: a, phantom appearance; b, set of tubes with prepared emulsions.

 

To model these fat concentrations, direct oil-in-water emulsions were prepared using vegetable oils (sunflower and soybean) with uniform phase distribution [9–11]. To ensure prolonged stability and homogeneity of the samples, the BTMS (Behentrimonium Methosulfate) emulsifier was used. Dispersion (emulsification) was performed by heating the emulsifier and subsequently mixing it with vegetable oil. Final emulsification was carried out using an IKA Ultra-Turrax T 25® ultrasonic disperser (IKA-Werke GmbH & Co. KG, Germany).

During the quality-control stage, the developed phantom was tested using magnetic resonance spectroscopy and software-based calculation on a Philips Ingenia Achieva 3.0 T® scanner (Philips Healthcare, Netherlands). For each test tube, 1H MR spectra were acquired (Stimulated Echo Acquisition Mode [STEAM]; TE = 12, 24, 36, 48, 60 ms; Repetition Time [TR] = 6000 ms). These spectra were processed using custom MATLAB® (MathWorks, USA) code. Water and fat signal intensities were corrected for T2 relaxation, enabling quantitative determination of PDFF. The accuracy of the predefined fat-fraction values was confirmed: the maximum deviation from the assigned value was 5% for the 20% fat sample. Thus, the phantom’s fat-fraction concentrations were validated as reliable.

Intervention

A series of phantom tests was conducted from February 2022 to February 2024. The phantom was scanned using magnetic resonance imaging systems from three different manufacturers operating at different magnetic field strengths (1.5 T and 3.0 T) (Fig. 2):

• Optima® MR450w 1.5 T (General Electric Healthcare, USA) using IDEAL IQ (IDEAL) and LAVA Flex (LAVA) sequences on two separate scanners;
• MAGNETOM® Skyra 3 T (Siemens Healthcare, Germany) using the T1 VIBE (VIBE) sequence on scanners from two different medical centers;
• Ingenia® 3 T and Ingenia® Achieva 3.0 T (Philips Healthcare, Netherlands) using mDIXON and DIXON QUANT (QUANT) sequences on scanners from two centers.

 

Fig. 2. Phantom setup and experimental procedure.

 

A standard scanning protocol preset on each specific scanner was used.

Fat fraction assessment was performed on the phantom MR images by placing a region of interest (ROI) on the corresponding images (Fig. 3a). Signal intensity values within each ROI were recorded on five different slices (see Fig. 3b) for the In-phase, Out-of-phase, Fat-only, and Water-only image series. The mean values were then calculated and entered into the final test report.

 

Fig. 3. Magnetic resonance images of the phantom: a, axial fat-only scan with a circular region of interest placement; b, coronal scan of the phantom with the levels at which signal intensity measurements were performed. ROI, region of interest; D, diameter; S, area.

 

Fat Fraction Assessment

To quantify the FF for each sample, the fat percentage was calculated based on the final SI values using two established formulas [12].

• Standard formula using in-phase and out-of-phase images:

$F{F}_1=\frac{SI\left(\text{In}\right)-\text{SI}\left(\text{Out}\right)}{2\times \text{SI}\left(\text{In}\right)}\times 100$, (1)

where FF₁, fat fraction calculated from in-phase and out-of-phase images (%); SI(In), signal intensity on in-phase images; SI(Out), signal intensity on out-of-phase images.

• Formula incorporating water-only and fat-only images:

$F{F}_2=\frac{\text{SI}\left(\text{Fat}\right)}{\text{SI}\left(\text{Fat}\right)+\text{SI}\left(\text{water}\right)}\times 100$, (2)

where FF₂, fat fraction calculated from water-only and fat-only images (%); SI(Fat), fat-signal intensity; SI(water), water-signal intensity.

A comparison of the calculated (measured) values obtained using both formulas with the predefined fat concentrations in the phantom was then performed through construction and analysis of the linear relationship.

Ethics Approval

This study did not involve human participants or laboratory animals. Therefore, ethical review of the study protocol was not required.

Statistical Analysis

The linear relationship between the measured FF values and the predefined concentrations was assessed by constructing linear plots. Regression analysis was used to evaluate the coefficients of linear regression and their significance for the measured FF values obtained using formulas (1) and (2).

To assess the reproducibility of FF₁ and FF₂ measurements derived from Dixon sequences, the mean, standard deviation, and coefficient of variation were calculated.

The assessment was performed for the following options:

• Overall coefficient of variation:

$CV=\frac{s{d}_i}{M_i}$, (3)

where M and sd represent the mean and standard deviation, respectively, across all phantom scans; i, scan number.

Coefficient of variation for a single scanner model and a single Dixon sequence (intra-scanner):

$C{V}_j=\frac{s{d}_{ji}}{M_{ji}}$, (4)

where M and sd represent the mean and standard deviation for phantom scans performed on the same scanner model using the same Dixon sequence; i, scan number; j, scanner model and Dixon sequence.

To compare measurement variance across different scanner models, an F-test was applied after preliminary assessment of data normality using the Shapiro–Wilk test. Statistical analyses were performed using SciPy (version 1.10.1) in the Python programming language and Microsoft Excel® version 16 (Microsoft, USA). The statistical significance threshold was set at p < 0.05 for all analyses. Five repeated FF measurements were performed according to standard measurement methodology, and MRI scanners were selected under the condition that each center had at least two scanners capable of Dixon imaging.

RESULTS

Comparison of the obtained FF₁ and FF₂ values calculated using formulas (1) and (2) for the different phantom samples revealed the following patterns (Fig. 4).

 

Fig. 4. Results of fat fraction determination on various magnetic resonance imaging scanners using Dixon sequences and formula-based calculations: a, calculation performed from in-phase and out-of-phase images (1); b, calculation performed from water-only and fat-only images (2). GE Optima, Optima® MR450w 1.5 T (General Electric Healthcare, United States); Philips Achieva, Ingenia® Achieva 3.0 T (Philips Healthcare, Netherlands); Philips Ingenia, Ingenia® 3 T (Philips Healthcare, Netherlands); Siemens Skyra, MAGNETOM® Skyra 3 T (Siemens Healthcare, Germany).

 

Analysis of the plots generated for each scanner and DIXON sequence showed that when formula (1) was applied (see Fig. 4a), all measured FF concentrations demonstrated a nonlinear relationship with the predefined values. In contrast, data calculated using formula (2) (see Fig. 4b) exhibited a pronounced linear relationship in most measurement sets, although certain values showed a shift along the y-axis.

The results of the linear regression analysis for FF₁ and FF₂ obtained using formulas (1) and (2) are presented in Table 1 for each scanner and DIXON sequence.

 

Table 1. Analysis of linear regression coefficients for calculated fat fraction values relative to predefined concentrations

MRI scanner model (Dixon sequence)	FF1		FF2
	b (95% CI)	a (95% CI)	b (95% CI)	a (95% CI)
Optima® MR450w, 1.5 T (General Electric Healthcare, USA)
Optima (IDEAL)	27.691 (−8.88…64.26)	0.071 (−0.87…1.01)	2.201 (−3.79…8.19)	1.34 (1.19–1.50)
Optima (LAVA)	48.31 (38.68–57.95)	−0.7 (−0.95…−0.45)	30.29 (17.50–43.07)	1.07 (0.74–1.40)
Optima 2,0 (IDEAL)2	10.681 (−7.09…28.44)	0.63 (0.17–1.09)	3.92 (2.00–5.84)	0.96 (0.91–1.01)
Optima 2,0 (LAVA)2	37.67 (24.03–51.31)	−0.31 (−0.66…0.04)	23.88 (6.25–41.50)	1.09 (0.64–1.54)
Ingenia® 3 T and Ingenia® Achieva 3.0 T (Philips Healthcare, Netherlands)
Achieva (QUANT)	11.611 (−3.79…27.02)	0.56 (0.16–0.95)	4.17 (1.76–6.58)	0.94 (0.87–1.00)
Ingenia (mDixon)	51.57 (27.16–75.98)	−0.591 (−1.21…0.04)	24.9 (15.73–34.08)	1.08 (0.84–1.31)
Ingenia 2,0 (mDixon)2	1.66 (−8.10…11.42)	0.831 (0.58–1.08)	−2.991 (−12.73…6.74)	1.02 (0.77–1.27)
Ingenia 2,0 (QUANT)2	2.081 (−7.54…11.69)	0.82 (0.57–1.07)	−3.061 (−12.76…6.64)	1.01 (0.77–1.26)
MAGNETOM® Skyra 3 T (Siemens Healthcare, Germany)
Skyra (VIBE)	40.59 (33.12–48.07)	−0.61 (−0.80…−0.41)	33.11 (4.22–61.99)	1.19 (0.44–1.93)
Skyra 2,0 (VIBE)2	33.58 (9.16–58.00)	0.091 (−0.54…0.72)	10.581 (−0.21…21.37)	1.13 (0.85–1.40)
Note. 1, coefficient values not demonstrating statistical significance according to regression analysis (p > 0.05); 2, second scanner of the same manufacturer located in a different medical center; a, slope coefficient; b, intercept coefficient; FF1, fat fraction calculated using the formula based on parameters derived from in-phase and out-of-phase images; FF2, fat fraction calculated using the formula based on parameters derived from water-only and fat-only images; CI, confidence interval.

 

The linear regression equation was expressed as:

$y=a\times x+b$, (5)

where b is the intercept and a is the slope coefficient.

Analysis of the data summarized in Table 1 indicates a statistically significant shift in the measured FF₁ and FF₂ values exceeding 5%. This shift was observed, for example, for:

• both Optima® MR450w 1.5 T (General Electric Healthcare, USA) scanners using the LAVA sequence;
• the Ingenia® 3 T (Philips Healthcare, Netherlands) scanner using the mDIXON sequence;
• both MAGNETOM® Skyra 3 T (Siemens Healthcare, Germany) scanners using the VIBE sequence.

The slope coefficients of the linear regression for FF₁ ranged widely from −0.70 to 0.83, whereas for FF₂ the values were close to unity, ranging from 0.94 to 1.34 (see Table 1).

To assess the reproducibility of measurements for each specific MRI scanner model and DIXON sequence, statistical parameters were calculated as presented in Table 2.

 

Table 2. Scanning results for six phantom samples (10–60٪) using two methods of fat fraction assessment

Parameter	MRI scanner model (Dixon sequence)	FF1						FF2
		10%	20%	30%	40%	50%	60%	10%	20%	30%	40%	50%	60%
Overall assessment (all scanners and Dixon sequences)
Mean value Standard deviation Overall coefficient of variation. %	All scanners (Dixon sequences)	20.84 10.58 50.77	30.10 9.95 33.07	34.11 8.78 25.75	34.16 10.08 29.49	34.16 10.08 29.49	31.45 10.08 32.04	19.89 10.52 52.91	34.71 16.12 46.45	46.84 17.26 36.85	57.62 16.55 28.73	57.62 16.55 28.73	67.99 16.55 24.35
Optima® MR450w. 1.5 T (General Electric Healthcare. USA)
Mean value	Optima (IDEAL) Optima (LAVA) Optima 2.0 (IDEAL)1 Optima 2.0 (LAVA)1	14.72 37.27 12.86 27.43	27.97 40.29 22.01 36.31	45.34 27.42 31.77 33.52	44.91 18.85 42.42 26.79	28.36 12.23 48.14 21.49	19.44 6.81 39.14 16.27	14.72 35.16 13.01 30.56	27.96 53.15 23.06 41.46	45.34 67.41 32.59 65.73	55.13 76.65 43.06 73.07	71.77 83.85 52.24 78.48	80.58 89.60 60.33 83.33
Coefficient of variation. %	Optima (IDEAL) Optima (LAVA) Optima 2.0 (IDEAL)1 Optima 2.0 (LAVA)1	2.89 2.66 21.19 4.65	2.85 1.52 11.92 1.67	1.14 3.19 4.85 2.05	0.52 5.94 2.27 2.42	4.06 7.97 1.11 4.74	3.98 18.22 1.89 6.69	2.80 4.03 3.20 4.40	2.82 1.16 2.95 11.47	1.16 1.21 1.45 0.80	0.38 1.48 1.30 1.10	1.65 1.27 1.06 1.25	0.96 1.06 0.95 0.64
Ingenia® 3 T and Ingenia® Achieva 3.0 T (Philips Healthcare. Netherlands)
Mean value	Ingenia (mDixon) Achieva (QUANT) Ingenia 2.0 (mDixon)1 Ingenia 2.0 (QUANT)1	33.66 12.40 8.27 8.67	46.39 23.04 16.94 17.21	45.84 30.50 27.17 27.34	29.18 40.45 38.80 38.78	18.73 43.19 46.27 46.35	12.58 37.44 46.27 46.30	33.34 12.77 7.40 7.38	50.41 24.07 15.86 15.83	53.56 31.43 26.69 26.53	70.59 42.39 38.23 37.97	80.71 51.14 53.63 53.56	87.23 59.89 53.53 53.50
Coefficient of variation. %	Ingenia (mDixon) Achieva (QUANT) Ingenia 2.0 (mDixon)1 Ingenia 2.0 (QUANT)1	2.00 7.95 8.96 15.58	0.40 2.68 3.84 5.94	1.45 6.09 1.21 2.09	2.02 1.53 0.11 0.16	2.52 0.22 0.50 0.69	3.22 1.47 0.07 0.18	2.27 2.02 2.55 2.62	14.06 1.14 1.48 1.53	0.27 4.74 0.88 1.95	0.73 2.68 1.03 2.25	1.02 1.62 0.67 0.77	0.62 1.22 0.24 0.31
MAGNETOM® Skyra 3 T (Siemens Healthcare. Germany)
Mean value	Skyra (VIBE) Skyra 2.0 (VIBE)1	32.09 23.27	33.10 38.34	21.86 47.91	14.46 43.05	9.33 35.98	5.38 32.02	29.87 16.70	66.24 35.40	77.85 47.46	85.98 58.79	91.69 67.57	95.96 73.91
Coefficient of variation. %	Skyra (VIBE) Skyra 2.0 (VIBE)1	2.75 0.84	4.51 0.37	8.79 0.42	14.59 0.57	24.46 0.32	42.44 7.40	9.06 0.88	2.67 0.36	1.66 0.53	1.12 0.36	0.51 0.19	0.29 0.17
Note. 1. second scanner of the same manufacturer located in another medical center; FF1. fat fraction calculated using the formula based on parameters derived from in-phase and out-of-phase images; FF2. fat fraction calculated using the formula based on parameters derived from water-only and fat-only images.

 

To evaluate the variance of measurements across scanner models and DIXON sequences, an F-test was performed. The analysis was conducted pairwise for each of the six phantom samples (10%–60%) using the measured FF₁ and FF₂ values. For FF₁ (sample with 10% fat content), statistically significant differences in variance were identified when comparing measurements obtained on the Optima® MR450w 1.5 T (General Electric Healthcare, USA) scanners using the IDEAL sequence (p = 0.002) and the MAGNETOM® Skyra 3 T (Siemens Healthcare, Germany) scanners using the VIBE sequence (p = 0.007). In contrast, no statistically significant differences were observed for the Ingenia® 3 T (Philips Healthcare, Netherlands) scanners using the mDIXON sequence. For FF₂ (sample with 10% fat content), statistically significant differences in variance were found for the Ingenia® 3 T (Philips Healthcare, Netherlands) scanners using the mDIXON sequence (p = 0.010) and the MAGNETOM® Skyra 3 T (Siemens Healthcare, Germany) scanners using the VIBE sequence (p < 0.001). Furthermore, when identical Dixon sequences (IDEAL and LAVA) were used on Optima® MR450w 1.5 T (General Electric Healthcare, USA) scanners, equivalent coefficients of variation for FF₂ were obtained. However, when these results were compared with each other, statistically significant differences in the FF₂ coefficient of variation were observed (p = 0.020).

DISCUSSION

This multicenter phantom study revealed both the capabilities and the limitations of Dixon pulse sequences for quantitative FF assessment across MRI scanners of different models and manufacturers. For accurate non-software-based quantification of FF, calculations based on formula (2), which are measurements derived from water-only and fat-only images, are preferable. Calculations using formula (1) produce inconsistent and difficult-to-interpret results. The findings emphasize the need for phantom-based calibration to assess measurement reproducibility and to calculate correction coefficients for aligning measured values with predefined concentrations. However, calculations based on formula (2) also require prior calibration.

Current scientific research is focused on identifying reliable noninvasive biomarkers, quantitative metrics derived from objective digital data rather than visual assessment. Percentage-based FF estimation in MRI provides additional information within the region of interest. This is made possible by multiple variants of Dixon pulse sequences available across scanners from all manufacturers [12]. The quantitative parameters FF and PDFF (when using automated software modules) derived from Dixon pulse sequences are widely used in clinical practice for the differential diagnosis of adrenal lesions, hepatic conditions, and detection of chylous abdominal neoplasms [13]. Moreover, the method is applied to evaluate pathological changes in the musculoskeletal system in benign and malignant neoplasms, skeletal muscle dystrophy, osteoporosis, hematologic disorders, as well as for detection and grading of hepatic steatosis [14–18].

The phantom represents a complex test object designed to model intracellular fat specifically. The contents of the test tubes serve as a tissue-mimicking material, meaning that the fat emulsion reproduces the precise intracellular fat content found in normal and pathologically altered tissues at varying concentrations (for example, in adrenal adenoma or hepatic steatosis). A phantom incorporating iron-containing emulsions has been described in the published data [19]. The phantom model used in our experiment does not account for the presence of additional components, particularly iron. On the one hand, this may be considered a limitation; on the other hand, the feasibility of routinely including iron-containing samples in standard tests remains insufficiently studied, as the formulas in use do not account for its influence on the magnetic resonance signal.

We evaluated the performance of standard Dixon pulse sequences relative to the predefined reference FF values in the phantom. The findings demonstrate the feasibility of reliable FF quantification using formula (2), which is based on water-only and fat-only signal intensities. Across all scanner models and Dixon sequences, a linear relationship was observed between FF₂ values and the predefined ones. The closest agreement with the true line was achieved on the Ingenia® 2.0 scanners using the QUANT and mDixon sequences, on the Ingenia® Achieva 3.0 T (Philips Healthcare, Netherlands) using the QUANT sequence, and on the Optima® MR450w 1.5 T (General Electric Healthcare, USA) using the IDEAL sequence. For formula (1), a linear trend with a positive slope was seen only at FF values from 0% to 30%, whereas at 30%–60%, visual assessment showed a slope that was either negative or near zero.

The dispersion analysis based on five repeated measurements for each scanner model and Dixon sequence revealed variability in results obtained on identical equipment across different medical centers. These findings highlight the need for cross-validation between institutions, even when the same scanner model and Dixon sequence are used.

Study Limitations

A limitation of this study was the absence of samples with an FF greater than 60%. This limitation was related to the phase separation observed during the preparation of highly concentrated fat emulsions using the employed methodology, resulting in separation of the fat and aqueous components. Reliable assessment of signal characteristics requires homogeneous and stable emulsions. Nonetheless, according to clinical observations, FF values exceeding 60% are uncommon in tissues; therefore, the obtained results may reasonably be extrapolated to higher concentrations.

Another limitation was the lack of post-experiment data regarding the stability of the phantom’s magnetic resonance characteristics.

CONCLUSION

The phantom described in this study enables monitoring of measurement reproducibility across different MRI scanners and validation of obtained results regardless of scanner manufacturer or model, thereby ensuring appropriate quality control in MRI examinations. When automated FF calculation software is unavailable, preliminary phantom testing allows for the derivation of correction coefficients and, when needed, adjustment of scanner-specific values. These findings contribute to improving the diagnostic quality of MRI, providing radiologists with greater accuracy in establishing the diagnosis.

This phantom study, conducted using MRI scanners from multiple vendors, demonstrated that Dixon sequences can be effectively used for quantitative FF analysis provided that preliminary phantom testing is performed. Based on the obtained results, phantom testing is recommended to improve the accuracy and reproducibility of quantitative measurements on specific scanner models. For reliable FF quantification, calculations should preferably be performed using water-only and fat-only maps according to formula (2).

ADDITIONAL INFORMATION

Author contributions: O.Yu. Panina: conceptualization, investigation, writing—original draft; A.I. Gromov: supervision, data interpretation, writing—review & editing; E.S. Akhmad: data interpretation, formal analysis, writing—original draft; D.S. Semenov: phantom development and manufacturing, investigation; A.V. Petraikin: supervision, data interpretation, writing—review & editing; S.A. Kivasev: investigation; V.A. Nechaev: writingformal analysis— review & editing, final approval of the results. All the authors approved the version of the manuscript to be published and agreed to be accountable for all aspects of the work, ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Acknowledgments: The authors express their gratitude to K.A. Sergunova, Cand. Sci. (Engineering) for the initial idea that inspired this study.

Ethics approval: This study did not involve human participants or laboratory animals; therefore, ethics approval was not required.

Funding sources: This article was prepared by the author team as part of the research project Scientific Support of the Standardization, Safety, and Quality of Magnetic Resonance Imaging (Unified State Information Accounting System No. 123031500007-6), in accordance with Order No. 1196 dated December 21, 2022, On Approval of State Assignments Funded by the Budget of the City of Moscow for State Budgetary (Autonomous) Institutions Under the Jurisdiction of the Moscow City Health Department for 2023 and the Planned Period of 2024–2025, issued by the Moscow City Health Department.

Disclosure of interests: The authors have no relationships, activities, or interests for the last three years related to for-profit or not-for-profit third parties whose interests may be affected by the content of the article.

Statement of originality: No previously published material (text, images, or data) was used in this work.

Data availability statement: The editorial policy regarding data sharing does not apply to this work.

Generative AI: No generative artificial intelligence technologies were used to prepare this article.

Provenance and peer review: This paper was submitted unsolicited and reviewed following the standard procedure. The peer review process involved two members of the editorial board and the in-house science editor.

About the authors

Olga Yu. Panina

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies; Moscow City Hospital named after S.S. Yudin

Author for correspondence.
Email: olgayurpanina@gmail.com
ORCID iD: 0000-0002-8684-775X
SPIN-code: 5504-8136

MD

Russian Federation, 24 Petrovka st, bldg 1, Moscow, 127051; Moscow

Alexander I. Gromov

Russian University of Medicine; National Medical Research Radiological Center

Email: gai8@mail.ru
ORCID iD: 0000-0002-9014-9022
SPIN-code: 6842-8684

MD, Dr. Sci. (Medicine), Professor

Russian Federation, Moscow; Moscow

Ekaterina S. Ahkmad

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies

Email: akhmades@zdrav.mos.ru
ORCID iD: 0000-0002-8235-9361
SPIN-code: 5891-4384
Russian Federation, 24 Petrovka st, bldg 1, Moscow, 127051

Dmitry S. Semenov

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies

Email: semenovds4@zdrav.mos.ru
ORCID iD: 0000-0002-4293-2514
SPIN-code: 2278-7290

Cand. Sci. (Engineering)

Russian Federation, 24 Petrovka st, bldg 1, Moscow, 127051

Stanislav A. Kivasev

Central Clinical Hospital “RZD-Medicine”

Email: Kivasev@yahoo.com
ORCID iD: 0000-0003-1160-5905
SPIN-code: 9883-3406

MD

Russian Federation, Moscow

Alexey V. Petraikin

Research and Practical Clinical Center for Diagnostics and Telemedicine Technologies

Email: PetryajkinAV@zdrav.mos.ru
ORCID iD: 0000-0003-1694-4682
SPIN-code: 6193-1656

MD, Dr. Sci. (Medicine)

Russian Federation, 24 Petrovka st, bldg 1, Moscow, 127051

Valentin A. Nechaev

Moscow City Hospital named after S.S. Yudin

Email: NechaevVA1@zdrav.mos.ru
ORCID iD: 0000-0002-6716-5593
SPIN-code: 2527-0130

MD, Cand. Sci. (Medicine)

Russian Federation, Moscow

References

Supplementary files