Intraindividual comparison of 3-T and 5-T gadoxetic acid-enhanced MRI for evaluating HCC: initial results

doi:10.21203/rs.3.rs-8275712/v1

Download PDF

Research Article

Intraindividual comparison of 3-T and 5-T gadoxetic acid-enhanced MRI for evaluating HCC: initial results

https://doi.org/10.21203/rs.3.rs-8275712/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background

We aimed to evaluate the utility of 5-T gadoxetic acid (Gd-ethoxybenzyl-diethylenetriaminepentaacetic acid, Gd-EOB-DTPA)-enhanced magnetic resonance imaging (MRI) by intraindividual comparison with 3-T, focusing on image quality and diagnosis of hepatocellular carcinoma (HCC).

Methods

We prospectively enrolled 28 patients with suspected HCC who underwent dynamic Gd-EOB-enhanced MRI using both 5-T and 3-T scanners. Artificial intelligence-assisted compressed sensing (ACS) and parallel imaging (PI) were both used for hepatobiliary phase (HBP) imaging at 5-T. Two radiologists performed the qualitative and quantitative assessments of image quality, and the evaluation of imaging features. Wilcoxon signed-rank, paired χ², and Cochran Q test as well as intraclass correlation coefficients and Cohen κ were used.

Results

All subjective image quality scores were rated as good to excellent. The subjective scores of contrast-enhanced phases at 5 T were higher than those at 3 T (p ≤ 0.016) except for image artifacts. Quantitative measures were also greater at 5 T (p ≤ 0.019). Subjective and quantitative assessment of HBP imaging were higher with ACS (p ≤ 0.046). The detection rate of enhancing HCC capsule was higher at 5 T (p = 0.031), as well as the peritumoral hypointensity on the HBP image at 5 T using ACS (p = 0.039).

Conclusions

Liver dynamic Gd-EOB-DTPA-enhanced 5-T MRI demonstrated superior image quality for contrast-enhanced phases and greater sensitivity in detecting the HCC enhancing capsule compared with 3-T MRI. The integration of 5-T MRI and ACS technology holds the potential to further improve image quality and the assessment of imaging features.

Relevance statement

Gd-EOB-DTPA-enhanced 5-T MRI provides promising potential for accurate HCC evaluation.

Artificial intelligence

Carcinoma (hepatocellular)

Gadolinium ethoxybenzyl DTPA

Liver neoplasms

Magnetic resonance imaging

Hepatocellular carcinoma (HCC) is the most prevalent primary malignant tumor of the liver, characterized by a high incidence and mortality rate. Early diagnosis of HCC is crucial for timely treatment and improved prognosis. Research indicates that, compared to ultrasound and computed tomography, multiparametric magnetic resonance imaging (MRI) is more effective in detecting liver focal lesions with non-typical ultrasound/computed tomography features, thereby avoiding unnecessary biopsies due to its superior spatial resolution and optimal soft tissue contrast [1]. Contrast-enhanced MRI is widely used in detecting and characterizing liver lesions. The liver cell-specific contrast agent gadoxetic acid (Gd-ethoxybenzyl-diethylenetriaminepentaacetic acid, Gd-EOB-DTPA) possesses a dynamic enhancement effect similar to that of extracellular contrast agents (ECAs) in arterial phase (AP) and portal venous phase (PVP), and is specifically absorbed by liver cells at hepatobiliary phase (HBP), making it a bifunctional contrast agent [2]. It has been proved that Gd-EOB-DTPA-enhanced MRI significantly improved the detection rate and diagnostic accuracy of small (≤ 20 mm) and subcentimeter (<10 mm) focal lesions [3, 4].

The development of ultrahigh-field MRI has improved the limited spatial resolution and signal-to-noise ratio (SNR) of MRI. There is a growing interest in employing ultrahigh magnetic fields in whole-body MRI systems to explore their potential clinical applications. Recently, a 5-T whole-body MRI scanner has been introduced, which may offer advantages for abdominal imaging. Previous studies have shown that 5-T MRI provides superior image quality than 3-T, including higher SNR, better tumor identification, and clearer bile duct display in functional imaging and conventional plain scan [5 − 8].

On the other hand, there is a trade-off between the acquisition time and spatial resolution of MRI [9]. In order to obtain high spatial resolution images, the scanning time will increase and the SNR may decrease [10]. Thus, it is both crucial and challenging to conduct high-resolution liver MRI examinations with appropriate breath-holding times without compromising image quality [11]. In comparison to current acceleration techniques such as parallel imaging (PI), the advent of artificial intelligence-assisted compressive sensing (ACS) has established a balance between scanning time and image quality [12, 13]. Previous studies have demonstrated that ACS significantly reduced scanning time without sacrificing image quality when compared to PI, thereby enhancing efficiency and minimizing motion artifacts [12 − 14].

We aimed to evaluate and compare image quality of plain and Gd-EOB-DTPA-enhanced MRI of HCC as well as the display of imaging features − including major and ancillary features of Liver Imaging Reporting and Data System (LI-RADS) version 2018 [15] − obtained at 5 and 3 T.

Study design and participants

This study is a single center, head to head controlled prospective study in accordance with the Declaration of Helsinki and has been approved by the Ethics Review Committee of the First Affiliated Hospital of the University of Science and Technology of China (2022KY-267). All patients provided written informed consent before the examination. Between September 2023 to December 2024, patients suspected of having HCC were enrolled. Inclusion criteria: (1) ≥ 18 years old; (2) Underwent 5-T gadoxetic acid-enhanced MRI. Exclusion criteria: (1) The patient didn't meet the LI-RADS v2018 criteria for high-risk populations;(2) Participants who withdrew from the trial and did not undergo the scheduled 3-T gadoxetic acid-enhanced MRI; (3) Not indicated for surgical resection [16];(4) Final pathological results negative for HCC/ICC or without pathological. Finally, 28 patients were included in this study (Fig. 1).

Image acquisition

After admission, all patients underwent MRI using a 5-T scanner (uMR Jupiter, United Imaging Healthcare, Shanghai, China) and a 3-T scanner (uMR 870, United Imaging Healthcare, Shanghai, China), both of them with a 24-channel body coil. The interval time between the 5-T and 3-T acquisitions was from two days to one week. All participants were informed of the potential risks associated with high-field MRI, including mild nausea, dizziness, headache, etc., as well as the purpose of the examination and the risk of contrast agent allergies. Respiratory coordination training was conducted before the scan to reduce the potential impact of respiratory artifacts. In order to purely compare effects of magnetic field strength on the image quality, same sequences and parameters were used at 3 and 5 T, and the spatial resolution was consistent in paired sequences. The imaging sequences included unenhanced axial three-dimensional (3D fast spoiled gradient-echo T1-weighted imaging (“quick-3D” in United Imaging Healthcare systems), axial fast spin echo T2-weighted imaging, diffusion-weighted imaging, and dynamic contrast enhanced scanning (Supplementary Tables S1 and S2).

For pre- and post-contrast scanning including HBP imaging, quick-3D sequence was applied. The liver cell-specific contrast agent, gadoxetic acid (Gd-EOB-DTPA, Primovist; Bayer Healthcare, Berlin, Germany), was injected into the elbow vein at a rate of 1 − 2 mL/s using a high-pressure injector with a dose of 0.1 mL/kg, and then rinsed with 20 mL of saline. Early and late AP phase, PVP, transitional phase (TP), and HBP images were collected at 15 s, 60 s, 120 s, and 20 min after the injection of the contrast agent, respectively. During dynamic imaging, all quick-3D sequences were accelerated by PI method at 3 and 5 T. For HBP imaging, PI and ACS were both applied at 5-T.

Image analysis

All sets of pre- and post-contrast images were transmitted to the postprocessing workstation (uWS-MR, United Imaging Healthcare, Shanghai, China) for image analysis and related parameter measurement. Two abdominal radiologists (with five and ten years of work experience) evaluated all images without knowing the patients information, pathological diagnosis, and field strength.

Subjective evaluation

The two radiologists independently and subjectively evaluated pre- and post-contrast images at 5 and 3 T with respect to image artifacts, lesion edge clarity, liver edge clarity, and overall image quality. The scoring criteria for image artifacts and overall image quality adopt the 5-point Likert scale: 1 point, severe artifacts with image distortion, not diagnostic; 2 points, obvious artifacts, the image is blurry and affects diagnosis; 3 points, moderate artifacts, possibly diagnostic; 4 points, mild artifacts, good image quality; 5 points, no artifacts, excellent image quality. A 5-point scale was used to assess the edge clarity, 1 = no visible boundary; 2 = ill-defined boundary; 3 = obscure boundary; 4 = well defined boundary; 5 = excellent boundary. The average subjective scores of the two radiologists were taken as the final scores.

Objective evaluation

The two radiologists determined regions of interest (ROIs) independently. Four ROIs of fixed size (approximately 100 mm²) were placed at the second porta hepatis level of each sequence to measure the signal intensity (SI) of the liver parenchyma in the left inner, left outer, right anterior, and right posterior lobes, avoiding blood vessels, bile ducts, and lesions. Then these four ROIs were copied to other sequences. To measure the SI of the lesion, the slice displaying the maximum cross-section area of the lesion, and its adjacent upper and lower slices were selected for ROI drawing. The lesion ROIs were determined on the three slices to cover the entire lesion as much as possible and avoid necrosis and bleeding. The SNR, contrast-to-noise ratio (CNR) and contrast ratio (CR) were calculated as follows:

SNR = SI_liver / SD_background

CNR = |SI_liver - SI_lesion| / SD_background

CR = |SI_liver - SI_lesion| / SI_liver + SI_lesion

where SI_liver is the averaged SI within four liver parenchyma ROIs, SI_lesion is the averaged SI within three lesion ROIs, and SD_background is the mean standard deviation (SD) of the four corners of the background on the same slice where the ROIs for normal liver tissues were placed (Supplementary Fig S1). The average objective scores of two radiologists were taken as the final scores.

Display of HCC features

The two abdominal radiologists documented anatomical location and size of the target nodules together. They evaluated all LI-RADS major features (non-rim AP hyperenhancement, nonperipheral washout, enhancing capsule, and threshold growth), features of LI-M (rim AP hyperenhancement, peripheral washout, delayed central enhancement, targetoid restriction, and targetoid TP/HBP appearance), as well as specific ancillary features (nonenhancing capsule, nodule in nodule, mosaic structure, blood products in mass, and fat in mass), based on LI-RADS v2018. In addition, due to the important value of corona enhancement and peritumoral hypointensity on the HBP image [17], as well as arterial vessels in HCC [18], the two readers also evaluated these three features. When there was a discrepancy in image interpretation between two radiologists, the final decision was determined by a third radiologist, with 24 years of experience in liver imaging.

Statistical analysis

Sample size estimation for the paired design was conducted using PASS 2025, version 25.0.2 [19]. The comparison was made using a one-sided, paired-difference t-test, with a type I error (α) rate of 0.05. The underlying standard deviation of the paired difference distribution was assumed to be 0.7. To detect a paired mean difference of 0.4 with 90% statistical power (1-β), the number of needed pairs was 28. Other statistical analyses were conducted using SPSS 26.0 software (IBM Corp., Armonk, New York, USA). Shapiro-Wilk test was used to perform normality test on the data. Continuous data are reported as mean ± SD and were compared using the paired Student’s t-test for a normal distribution and Wilcoxon signed rank test for a non-normal distribution. Count data are expressed as frequencies and percentages. Paired χ² test was used for comparisons between 3-T and 5-T scans and Cochran Q test was applied to compare three groups of data. Intraclass correlation coefficient (ICC) and Cohen κ were used to assess the consistency between the two readers. Agreement levels were interpreted as follows: not consistent (0.01 − 0.20); poor (0.21 − 0.40); moderate (0.41 − 0.60); good (0.61 − 0.80); and excellent (0.81 − 0.99). Two-tailed p < 0.05 indicates statistically significant difference.

Patient characteristics

We included 28 high-risk HCC patients, comprising 24 males and 4 females, aged between 42 and 76 years, aged 60.07 ± 9.07 years. All patients underwent surgery or biopsy to obtain pathological results and the results confirmed HCC. Among these patients, there were 27 cases of single lesion, and one case of two lesions. Furthermore, all enrolled patients successfully completed 3-T and 5-T MRI without complications and there were no patients whose images could not be evaluated due to severe respiratory artifacts. The patient clinical characteristics are presented in Table 1.

Subjective evaluation and inter-observer consistency

The κ values of subjective scores from the two abdominal radiologists for all image sets obtained at 3 and 5 T ranged from 0.644 to 0.867, indicating good to excellent consistencies (Supplementary Tables S3 and S4).

There was no statistically significant difference in all subjective scores of pre-contrast sequences and artifact scores of all sequences between 3 and 5 T. Compared with 3-T MRI, 5-T MRI had significantly higher scores in lesion edge clarity, liver edge clarity, and overall image quality for all post-contrast enhanced images (Table 2). Moreover, all subjective scores except image artifacts of ACS images were significantly higher than that of PI images for HBP imaging at 5 T (Table 3).

Objective evaluation and inter-observer consistency

The ICCs of SNR and CNR from the two abdominal radiologists for all image sets obtained at 3 and 5 T ranged from 0.730 to 0.930, indicating excellent consistencies (Supplementary Table S5). Summaries of SNR, CNR and CR of pre- and post-contrast enhanced quick-3D images at 3 and 5 T are shown in Fig. 2. There was no statistically significant difference between pre-contrast enhanced quick-3D images between different field strengths, while SNR, CNR and CR of post-contrast enhanced images of 5-T MRI were significantly higher. The SNR, CNR and CR of ACS-accelerated HBP sequence were also significantly superior to those of PI-accelerated sequence at 5 T (Table 4).

Detection of imaging features in HCC and non-HCC malignancy

Using LI-RADS v2018, 27 lesions were categorized as LR-5, and 2 lesions were categorized as LR-M. No lesion was categorized as LR-TIV by the two radiologists (Supplementary Table S6). Regarding the enhancing capsule, 16 cases were detected on PVP/TP images at 3 T, 22 cases were detected on PVP/TP images at 5 T (Fig. 3). The detection rate of enhancing capsule was significantly higher at 5 than at 3 T (p = 0.031) (Table 5). On HBP images, 4 and 5 cases of peritumoral hypointensity were detected by PI-accelerated quick-3D at 3 and 5 T, respectively, and 8 cases were detected based on ACS-accelerated HBP imaging at 5 T, with a significant difference (p = 0.039) (Fig. 4). The pairwise comparison showed that the detection rate of ACS sequence at 5 T was higher than that of PI sequence at 3 T (p = 0.043) (Supplementary Table S7). For corona enhancement on AP images, 3 cases were observed at 3 T and 5 cases at 5 T (Fig. 5). Arterial vessels within the lesion were detected in 11 cases at 3, and 13 at 5 T (Fig. 6). The detection rate of other major/ancillary features and features of LI-M were consistent between the two field strengths (Supplementary Table S8). All enrolled patients were found to have liver lesions on their first examination, and the major feature of threshold growth was not applicable.

In this study, we compared 3-T and 5-T MRI in the diagnosis of HCC. The results showed that Gd-EOB-DTPA-enhanced MRI was superior at 5 T than at 3 T in terms of subjective evaluations of lesion edge clarity, liver edge clarity, and overall image quality, as well as considering the objective evaluations of SNR, CNR, and CR. In terms of displaying imaging features of HCC, 5-T MRI combined with ACS technology was superior to 3-T MRI. The combination of 5-T MRI and ACS technology can improve the quality of liver images and facilitate the display of HCC imaging features.

Influence of magnetic field strength on image quality

The image quality of MRI depends on factors such as SNR, CNR, spatial resolution, and artifacts. As the SI is linearly proportional to B₀ field strength, higher field could help improve SNR, CNR, and allow the application of high resolution imaging [20]. However, at high magnetic field strengths (> 3 T), the risk of distortion increases due to the inhomogeneity of B₀ and B₁ fields, and the specific absorption rate increases, especially in abdominal imaging [21, 22]. Therefore, most ultrahigh-field studies were limited to neuroimaging and musculoskeletal imaging [23, 24].

Recently, the 5-T ultrahigh-field scanner has been put into clinical application, and studies have shown that both functional imaging such as diffusion-weighted and conventional plain scanning have obtained high-quality images of abdominal organs including liver, pancreas, kidneys [5 − 7,25]. These applications indicate that 5-T MRI has advantages on better SNR, CNR, CR and high spatial resolution imaging, giving the potential in displaying fine structures such as small tumors and complex anatomical structures, which can effectively reduce the risk of missed diagnosis [26].

This study demonstrated that the image quality of contrast-enhanced images at 5 T were significantly superior to that at 3 T. Previous in vitro experiments have shown that native T1 relaxation time increased at higher field strengths, which may be the reason of improved contrast enhancement [27]. As the field strength increases, due to the combined effect of protein binding, the increase in T1 value of tissues leads to a corresponding increase in relative contrast, ultimately resulting in a significant change in T1 relaxation value under high magnetic field strength [28]. Jiong et al. [29] set similar repetition time, echo time, and flip angle in sequences at 3 and 5 T and the results showed that the contrast between tumor and brain tissue in the half-dose contrast-enhanced images at 5 T was significantly higher than that in the full-dose contrast at 3 T. There was no statistically significant difference between the unenhanced images obtained at 3 and 5- T in this study, which may be due to the fact that the sequence parameters at 5 T were adjusted to be consistent with 3 T for comparison instead of being optimized to meet hardware limits.

To mitigate industry-related bias, we applied several strategies. First, the 3-T and 5-T scanners were from the same manufacturer. Systematic errors have been minimized by the consistency in hardware and software. Second, same imaging sequences were used and parameters were set as identical as possible across different scanners. Besides, all image sets were reviewed on the same workstation and radiologists were blinded to any image information including the field strength. These strategies reinforced that image quality differences were attribute to field strength rather than industry-related confounders.

Advantages of ACS technology in contrast-enhanced sequences

MRI has unique advantages in displaying the structure of lesions and abdominal organs. However, the drawbacks includes long scanning time and is prone to artifacts, which is especially inevitable in abdominal imaging [30]. Common abdominal MRI artifacts include partial volume effects and motion/pulsation artifacts. In clinical settings, thin-slice and high spatial resolution imaging reduces local volume effects, and improves visibility and detection rates of small lesions, while increases scanning time and affects SNR. At present, commonly used acceleration techniques including CS, PI, and half-Fourier acquisition. However, when using high acceleration factors, these techniques may generate various artifacts and amplified noise during the image reconstruction process, thereby reducing image quality [31–35].

ACS follows the principles of half-Fourier, PI, and CS, and combines artificial intelligence modules in the reconstruction process [36], which reduces noise and artifacts, and corrects shortcomings of conventional acceleration techniques, providing reliable images for clinical diagnosis in a shorter time [37, 38]. Li et al. [12] compared the ACS-accelerated single breath-hold T2-weighted images with traditional respiratory-triggered T2-weighted images, and showed that ACS provided better abdominal image quality, lesion detection rate, and greatly reduced the scanning time. In another nasopharyngeal study [39], ACS not only shortened the scanning time, but also improved image quality. Our study applied ACS to thin-slice iso-resolution HBP imaging and obtained similar results to previous studies that subjective scores, SNR, CNR and CR were significantly higher than PI-accelerated sequence scanned with thicker slices, indicating that ACS-accelerated HBP imaging could clearly display the liver and lesions without increasing scan time.

Detection of imaging features of HCC

According to LI-RADS v2018 [15], enhancing capsule is a smooth peripheral edge enhancement during PVP in ECA-enhanced MRI or PVP/TP in Gd-EOB-DTPA-enhanced MRI. Compared with ECA-enhanced MRI, the sensitivity for LR-5 is reduced when using Gd-EOB-DTPA-enhanced MRI [40], which is a major challenge as LI-RADS was originally designed for ECA-enhanced MRI [41], and the degree of tumor parenchymal enhancement in Gd-EOB-DTPA-enhanced MRI is lower [42]. As the SI of the background liver tissue gradually increases in Gd-EOB-DTPA-enhanced MRI, the enhanced capsule is usually masked by the background liver tissue, thereby reducing the detection rate of this feature [43]. In order to improve the diagnostic sensitivity of enhanced capsule, Chung et al. [44] proposed using subtraction method to increase the detection rate of enhanced capsule. Kim et al. [45] modified the scanning sequence by replacing the traditional single PVP with the double PVP, which also improved the detection rate of enhanced capsule. The results of our study showed that compared to 3-T MRI, 5-T MRI can improve the detection enhanced capsule, improving the accuracy of HCC diagnosis.

As an ancillary feature of malignant tumor, corona enhancement has been shown to be an important predictor for HCC microvascular invasion [46]. Although arterial vessels within tumors and peritumoral hypointensity on the HBP image are not major or ancillary features, they are crucial for HCC diagnosis, prognosis, and determination of treatment strategies [47, 48]. Our study showed that the above features could be more clearly displayed at 5 T compared to 3 T, or ACS compared to PI sequences. The difference in detection rate of peritumoral hypointensity on the HBP image was significantly higher at 5-T ACS sequence, and the detection rates of corona enhancement and arterial vessels within tumors had also been improved with 5-T MRI, although the difference was not statistically significant. Since we had only two intrahepatic cholangiocarcinomas, the detection rate of LR-M features were consistent between 5-T and 3-T MRI. LR-M features including rim APHE, peripheral washout, delayed central enhancement and targetoid TP/HBP appearance can all be seen in these two cases.

This study has several limitations. First, 5-T MRI is still in its early stages, and the development of technology makes it challenging to ensure the use of completely consistent software platforms and phased array circles to obtain 3-T and 5-T scans. In fact, phased array coils produced for systems with different magnetic field strengths will have different performance characteristics, which makes it challenging to compare image quality and SNR that rely solely on magnetic field strength. Second, our study population is relatively small. This may also be the reason why there is no statistical difference in partial imaging features between 3 and 5 T. Another limitation is that all subjects underwent 3-T scan at after the 5-T scan. Although there was an at least 48-hour interval between the two injections of contrast agent, residual contrast agent in tumors and organs may have influenced the image quality, SNR and CNR. However, our research results showed no significant difference in image quality between unenhanced 3-T and 5-T, suggesting that the impact of residual contrast agents on image quality may not interfere with subsequent image quality analysis. Finally, we did not include non-HCC lesions to evaluate diagnostic performance. However, the primary objective of this study was to compare HCC detection and image quality.

In summary, 5-T MRI showed significant advantages over 3-T MRI with respect image quality and diagnosis of HCC. Furthermore, Gd-EOB-DTPA-enhanced MRI combined with ACS technology at 5 T obtained better image quality with same spatial resolution to display HCC imaging features, such as peritumoral hypointensity on HBP images, without increasing the scanning time.

3D Three-dimensional

ACS Artificial intelligence-assisted compressive sensing

AP Arterial phase

CNR Contrast-to-noise ratio

CR Contrast ratio

ECA Extracellular contrast agent

Gd-EOB-DTPA Gd-ethoxybenzyl-diethylenetriaminepentaacetic acid (gadoxetic acid)

HBP Hepatobiliary phase

HCC Hepatocellular carcinoma

LI-RADS Liver Imaging Reporting and Data System

MRI Magnetic resonance imaging

PI Parallel imaging

PVP Portal venous phase

quick-3D 3D fast spoiled gradient-echo T1-weighted sequence

ROI Region of interest

SD Standard deviation

SI Signal intensity

SNR Signal-to-noise ratio

TP Transitional phase

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Review Committee of the First Affiliated Hospital of the University of Science and Technology of China (approval number 2022KY-267, date of approval 27/10/2022).

Consent for publication

Not applicable.

Availability of data and material

All results of this study are presented in the manuscript in the form of tables, including scanning parameters. The datasets generated or analyzed during the study are available from the corresponding author upon reasonable request.

Competing interests

Zhichao Feng, Runyu Tang and Xiaopeng Song were employees of United Imaging Healthcare throughout their involvement in the study. The other authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

This study has received funding by National Key RandD Program of China (2019YFA0709300).

Authors' contributions

Shaopeng LI: onceptualization, data curation, methodology, project administration, validation, writing – original draft and review, and editing. Shuhang Liang, PhD: data curation, formal analysis, project administratio

Mengqiu Liu: data curation, methodology, validation. Xudan Chen: data curation, resources, validation

Dawei YIN: data curation, project administration, validation. Zhichao Feng: investigation, methodology, software, writing—review, and editing. Runyu Tang and Xiaopeng Song: formal analysis, methodology, software, writing–review, and editing. Peng Wan: project administration, supervision, validation. Lianxin Liu: conceptualization, formal analysis, funding acquisition, supervision, visualization, writing–review, and editing. Ying Liu: conceptualization, project administration, supervision, visualization, writing-original draft, review, and editing.

Acknowledgements

The authors would like to thank all participants for their contribution to this study. We did not use Large Language Models for the conceptualization or creation of this manuscript.

Chartampilas E, Rafailidis V, Georgopoulou V et al (2022) Current imaging diagnosis of hepatocellular carcinoma. Cancers (Basel) 14:3997. https://doi.org/10.3390/cancers14163997
Murakami T, Sofue K, Hori M (2022) Diagnosis of hepatocellular carcinoma using Gd-EOB-DTPA MR imaging. Magn Reson Med Sci 21:168–181. https://doi.org/10.2463/mrms.rev.2021-0031
Raman SS, Leary C, Bluemke DA et al (2010) Improved characterization of focal liver lesions with liver-specific gadoxetic acid disodium-enhanced magnetic resonance imaging: a multicenter phase 3 clinical trial. J Comput Assist Tomogr 34:163–172. https://doi.org/10.1097/RCT.0b013e3181c89d87
Zech CJ, Herrmann KA, Reiser MF et al (2007) MR imaging in patients with suspected liver metastases: value of liver-specific contrast agent Gd-EOB-DTPA. Magn Reson Med Sci 6:43–52. https://doi.org/10.2463/mrms.6.43
Zhang Y, Yang C, Liang L et al (2022) Preliminary experience of 5.0 T higher field abdominal diffusion-weighted MRI: agreement of apparent diffusion coefficient with 3.0 T Imaging. J Magn Reson Imaging 56:1009–1017. https://doi.org/10.1002/jmri.28097
Zheng L, Yang C, Sheng R et al (2022) Renal imaging at 5 T versus 3 T: a comparison study. Insights Imaging 13:155. https://doi.org/10.1186/s13244-022-01290-9
Yin L, Li Z, Shang M et al (2024) Magnetic resonance cholangiopancreatography at 5.0 T: quantitative and qualitative comparison with 3.0 T. BMC Med Imaging 24:331. https://doi.org/10.1186/s12880-024-01512-0
Xu Q, Zhao H, Gao R et al (2025) Insulinoma detection and surgery planning: a comparative study of 5.0T MRI versus 3.0T MRI and MDCT. Abdom Radiol (NY) 50:4148–4159. https://doi.org/10.1007/s00261-024-04680-3
Clarke SE, Saranathan M, Rettmann DW et al (2015) High resolution multi-arterial phase MRI improves lesion contrast in chronic liver disease. Clin Invest Med 38:E90–E99. https://doi.org/10.25011/cim.v38i3 .22704
Ihara K, Onoda H, Tanabe M et al (2023) Breath-hold high-resolution T1-weighted gradient echo liver MR imaging with compressed sensing obtained during the gadoxetic acid-enhanced hepatobiliary phase: image quality and lesion visibility compared with a standard T1-weighted sequence. Magn Reson Med Sci 23:146–152. https://doi.org/10.2463/mrms.mp.2022-0137
Yang AC, Kretzler M, Sudarski S et al (2016) Sparse reconstruction techniques in magnetic resonance imaging: methods, applications, and challenges to clinical adoption. Invest Radiol 51:349–364. https://doi.org/10.1097/RLI.0000000000000274
Li H, Hu C, Yang Y et al (2022) Single-breath-hold T2WI MRI with artificial intelligence-assisted technique in liver imaging: As compared with conventional respiratory-triggered T2WI. Magn Reson Imaging 93:175–180. https://doi.org/10.1016/j.mri.2022.08.012
Zhao Y, Peng C, Wang S et al (2022) The feasibility investigation of AI -assisted compressed sensing in kidney MR imaging: an ultra-fast T2WI imaging technology. BMC Med Imaging 22:119. https://doi.org/10.1186/s12880-022-00842-1
Sheng RF, Zheng LY, Jin KP et al (2021) Single-breath-hold T2WI liver MRI with deep learning-based reconstruction: A clinical feasibility study in comparison to conventional multi-breath-hold T2WI liver MRI. Magn Reson Imaging 81:75–81. https://doi.org/10.1016/j.mri.2021.06.014
American College of Radiology. Liver Imaging Reporting and Data System. Available online: https://www.acr.org/Clinical-Resources/Clinical-Tools-and-Reference/ Reporting-and-Data-Systems/LI-RADS
Zhou J, Sun H, Wang Z et al (2023) Guidelines for the diagnosis and treatment of primary liver cancer (2022 Edition). Liver Cancer 12(5):405–444. https://doi.org/10.1159/000530495
Lu M, Qu Q, Xu L et al (2023) Prediction for aggressiveness and postoperative recurrence of hepatocellular carcinoma using gadoxetic acid-enhanced magnetic resonance imaging. Acad Radiol 30:841–852. https://doi.org/10.1016/j.acra.2022.12.018
Koda M, Matsunaga Y, Ueki M et al (2004) Qualitative assessment of tumor vascularity in hepatocellular carcinoma by contrast-enhanced coded ultrasound: comparison with arterial phase of dynamic CT and conventional color/power Doppler ultrasound. Eur Radiol 14:1100–1108. https://doi.org/10.1007/s00330-003-2172-5
Bujang MA, Adnan TH (2016) Requirements for minimum sample size for sensitivity and specificity analysis. J Clin Diagn Res 10:YE01-YE06. https://doi.org/10.7860/JCDR/2016/18129.8744
Caruso D, Zerunian M, De Santis D et al (2020) Magnetic resonance of rectal cancer response to therapy: an image quality comparison between 3.0 and 1.5 Tesla. Biomed Res Int 10:9842732. https://doi.org/10.1155/2020/9842732
Platt T, Ladd ME, Paech D (2021) 7 Tesla and beyond: advanced methods and clinical applications in magnetic resonance imaging. Invest Radiol 56:705–725. https://doi.org/10.1097/RLI.0000000000000820
Vachha B, Huang SY (2021) MRI with ultrahigh field strength and high-performance gradients: challenges and opportunities for clinical neuroimaging at 7 T and beyond MRI with ultrahigh field strength and high-performance gradients: challenges and opportunities for clinical neuroimaging at 7 T and beyond. Eur Radiol Exp 5:35. https://doi.org/10.1186/s41747-021-00216-2
Verma G, Balchandani P (2019) Ultrahigh field MR neuroimaging. Top Magn Reson Imaging 28:137. https://doi.org/10.1097/RMR.0000000000000210
Juras V, Mlynarik V, Szomolanyi P et al (2019) Magnetic resonance imaging of the musculoskeletal system at 7 T: morphological imaging and beyond. Top Magn Reson Imaging 28:125. https://doi.org/10.1097/RMR.0000000000000205
Zheng L, Yang C, Liang L et al (2023) T2-weighted MRI and reduced-FOV diffusion-weighted imaging of the human pancreas at 5 T: A comparison study with 3 T. Med Phys 50:344–353. https://doi.org/10.1002/mp.15970
Amrami KK, Chebrolu VV, Felmlee JP et al (2023) 7T for clinical imaging of benign peripheral nerve tumors: preliminary results. Acta Neurochir (Wien) 165:3549–3558. https://doi.org/10.1007/s00701-023-05724-1
Krautmacher C, Willinek WA, Tschampa HJ et al(2005) Brain tumors: full- and half-dose contrast-enhanced MR imaging at 3.0 T compared with 1.5 T-initial experience. Radiology 237:1014–1019. https://doi.org/10.1148/radiol.2373041672
Obmann VC, Catucci D, Berzigotti A et al (2022) T1 reduction rate with Gd-EOB-DTPA determines liver function on both 1.5 T and 3 T MRI. Sci Rep 12:4716. https://doi.org/10.1038/s41598-022-08659-2
Min JH, Kim YK, Choi SY et al (2017) Differentiation between cholangiocarcinoma and hepatocellular carcinoma with target sign on diffusion-weighted imaging and hepatobiliary phase gadoxetic acid-enhanced MR imaging: Classification tree analysis applying capsule and septum. Eur J Radiol 92:1–10. https://doi.org/10.1016/j.ejrad.2017.04.008
Graves MJ, Mitchell DG (2013) Body MRI artifacts in clinical practice: a physicist's and radiologist's perspective. J Magn Reson Imaging 38:269–287. https://doi.org/10.1002/jmri.24288
Yoon JK, Kim MJ, Lee S (2019) Compressed sensing and parallel imaging for double hepatic arterial phase acquisition in gadoxetate-enhanced dynamic liver magnetic resonance imaging. Invest Radiol 54:374–382. https://doi.org/10.1097/RLI.0000000000000548
Bydder M, Ali F, Ghodrati V et al (2021) Minimizing echo and repetition times in magnetic resonance imaging using a double half-echo k-space acquisition and low-rank reconstruction. NMR Biomed 34:e4458. https://doi.org/10.1002/nbm.4458
Schulz J, P Marques J, Ter Telgte A et al (2018) Clinical application of half fourier acquisition single shot turbo spin echo (HASTE) imaging accelerated by simultaneous multi-slice acquisition.Eur J Radiol 98:200–206. https://doi.org/10.1016/j.ejrad. 2017.11.022
Feng L, Benkert T, Block KT et al (2017) Compressed sensing for body MRI. J Magn Reson Imaging 45:966–987. https://doi.org/10.1002/jmri.25547
Wang S, Cao G, Wang Y et al (2021) Review and prospect: artificial intelligence in advanced medical imaging. Front Radiol 1:781868. https://doi.org/10.3389/fradi.2021.781868
Priyanka, Kadavigere R, Nayak SS et al (2024) Impact of artificial intelligence assisted compressed sensing technique on scan time and image quality in musculoskeletal MRI - a systematic review. Radiography (Lond) 30:1704–1712. https://doi.org/10.1016/j.radi.2024.08.012
Zhao Q, Xu J, Yang YX et al (2023) AI-assisted accelerated MRI of the ankle: clinical practice assessment. Eur Radiol Exp 7:62. https://doi.org/10.1186/s41747-023-00374-5
Wang Q, Zhao W, Xing X et al (2023) Feasibility of AI-assisted compressed sensing protocols in knee MR imaging: a prospective multi-reader study. Eur Radiol 33:8585–8596. https://doi.org/10.1007/s00330-023-09823-6
Liu H, Deng D, Zeng W et al (2023) AI-assisted compressed sensing and parallel imaging sequences for MRI of patients with nasopharyngeal carcinoma: comparison of their capabilities in terms of examination time and image quality. Eur Radiol 33:7686–7696. https://doi.org/10.1007/s00330-023-09742-6
Min JH, Kim JM, Kim YK et al (2018) Prospective intraindividual comparison of magnetic resonance imaging with gadoxetic acid and extracellular contrast for diagnosis of hepatocellular carcinomas using the liver imaging reporting and data system. Hepatology 68:2254–2266. https://doi.org/10.1002/hep.30122
Hope TA, Fowler KJ, Sirlin CB et al (2015) Hepatobiliary agents and their role in LI-RADS. Abdom Imaging 40:613–625. https://doi.org/10.1007/s00261-014-0227-5
Collettini F, Elkilany A, Seta MD et al (2022) MR imaging of hepatocellular carcinoma: prospective intraindividual head-to-head comparison of the contrast agents gadoxetic acid and gadoteric acid. Sci Rep 12:18583. https://doi.org/10.1038/s41598-022-23397-1
Cho ES, Choi JY (2015) MRI features of hepatocellular carcinoma related to biologic behavior. Korean J Radiol 16:449–464. https://doi.org/10.3348/kjr.2015.16.3.449
Chung JW, Yu JS, Choi JM et al (2020) Subtraction images from portal venous phase gadoxetic acid-enhanced MRI for observing washout and enhancing capsule features in LI-RADS version 2018. AJR Am J Roentgenol 214:72–80. https://doi.org/10.2214/AJR.18.20797
Kim H, Kim SS, Shin HC et al (2023) Gadoxetate disodium-enhanced MRI for diagnosis of hepatocellular carcinoma in patients with chronic liver disease: late portal venous phase may improve identification of enhancing capsule. Abdom Radiol (NY) 48:621–629. https://doi.org/10.1007/s00261-022-03756-2
Wang L, Jia M, Wen X et al (2022) Diagnostic value of magnetic resonance imaging features of microvascular invasion in hepatocellular carcinoma: a meta-analysis. Diagn Interv Radiol 28:428–440. https://doi.org/10.5152/dir.2022.21108
Wang H, Xiao W, Han Y et al (2022) Study on safety and efficacy of regorafenib combined with transcatheter arterial chemoembolization in the treatment of advanced hepatocellular carcinoma after first-line targeted therapy. J Gastrointest Oncol 13:1248–1254. https://doi.org/10.21037/jgo-22-395
Rhee H, Park YN, Choi JY (2024) Advances in understanding hepatocellular carcinoma vasculature: implications for diagnosis, prognostication, and treatment. Korean J Radiol 25:887–901. https://doi.org/10.3348/kjr.2024.0307

Table 1
Clinical characteristics of the patients
Characteristics	Values
Number	28
Age (years)	60.07 ± 9.07
Sex
Male	24 (85.71%)
Female	4 (14.29%)
Tumor size (cm)	5.41 ± 2.60
Etiology of liver disease
Hepatitis B	25 (89.29%)
Hepatitis C	2 (7.14%)
Alcoholic liver disease	1 (3.57%)
Cirrhosis
Presence	15 (53.57%)
Absence	13 (46.43%)
Tumor markers
Alphafetoprotein (ng/mL)
>200	8 (28.57%)
7 − 200	7 (25.00%)
< 7	13 (46.43%)
CA19-9 (U/mL)
< 43	23(82.14%)
43 − 100	3 (10.72%)
>100	2 (7.14%)
Number of lesions
1	27 (96.43%)
2	1 (3.57%)
Per-patient final diagnosis
Intrahepatic cholangiocarcinoma	2 (6.90%)
Hepatocellular carcinoma	27 (93.10%)
Edmondson-Steiner grade
Ⅰ	3 (11.11%)
Ⅱ	10 (37.04%)
Ⅲ	13 (48.15%)
Ⅳ	1 (3.70%)
CA19-9 Carbohydrateantigen19-9.

Table 2
Image quality scores for 3-T and 5-T images
	T	Precontrast	Early-AP	Late-AP	PVP	TP	HBP
Image artifacts	3	3.75 ± 0.51	3.82 ± 0.51	3.91 ± 0.58	3.63 ± 0.49	3.73 ± 0.65	3.75 ± 0.55
Image artifacts	5	3.89 ± 0.49	3.95 ± 0.55	4.07 ± 0.50	3.79 ± 0.53	3.95 ± 0.7	3.91 ± 0.61
p value		0.088	0.226	0.095	0.085	0.083	0.140
Clarity of lesion margins	3	3.70 ± 0.53	3.73 ± 0.62	3.71 ± 0.71	3.61 ± 0.64	3.66 ± 0.64	3.68 ± 0.61
Clarity of lesion margins	5	3.89 ± 0.53	4.14 ± 0.44	4.27 ± 0.49	4.18 ± 0.43	4.09 ± 0.44	4.05 ± 0.52
p value		0.067	< 0.001	< 0.001	< 0.001	0.002	0.008
Liver edge clarity	3	3.82 ± 0.54	3.70 ± 0.59	3.63 ± 0.58	3.77 ± 0.55	3.59 ± 0.65	3.70 ± 0.59
Liver edge clarity	5	4.00 ± 0.54	4.03 ± 0.44	4.05 ± 0.56	4.11 ± 0.45	4.14 ± 0.44	4.02 ± 0.56
p value		0.065	0.005	0.004	0.003	< 0.001	0.016
Overall image quality	3	3.77 ± 0.62	3.70 ± 0.66	3.73 ± 0.56	3.66 ± 0.55	3.63 ± 0.56	3.70 ± 0.63
Overall image quality	5	3.91 ± 0.48	4.14 ± 0.52	4.20 ± 0.22	4.18 ± 0.47	4.09 ± 0.55	4.11 ± 0.53
p value		0.159	0.004	< 0.001	< 0.001	0.002	0.008
Data are means ± standard deviations. AP Arterial phase,, HBP Hepatobiliary phase, PVP Portal venous phase, TP Transitional phase.

Table 3
Image quality scores for PI- and ACS-accelerated HBP images
	Image artifacts	Clarity of lesion margins	Liver edge clarity	Overall image quality
PI	3.91 ± 0.61	4.05 ± 0.52	4.02 ± 0.56	4.11 ± 0.53
ACS	4.14 ± 0.55	4.32 ± 0.47	4.38 ± 0.49	4.39 ± 0.52
p value value	0.046	0.001	0.002	0.006
Data are means ± standard deviations. ACS Artificial intelligence-assisted compressed sensing, HBP Hepatobiliary phase, PI Parallel imaging.

Table 4
Comparisons of SNR, CNR and CR between PI and ACS images
	SNR_HBP	CNR_HBP	CR_HBP
PI	52.58 ± 6.32	15.19 ± 5.46	0.215 ± 0.082
ACS	58.94 ± 11.99	20.36 ± 5.84	0.246 ± 0.077
p value	0.003	< 0.001	0.017
Data are means ± standard deviations. ACS Artificial intelligence-assisted compressed sensing, CNR Contrast-to-noise ratio, CR Contrast ratio, HBP Hepatobiliary phase, PI Parallel imaging, SNR Signal-to-noise ratio;

Table 5
Detection rate of imaging features of 3-T and 5-T, as well as PI and ACS images
	Enhancing capsule	Intralesion arterial vessels on AP	Corona enhancement	Peritumoral hypointensity on HBP
3-T-PI	16/27 (59.26%)	11/27 (40.74%)	3/27 (11.11%)	4/27 (14.81%)
5-T-PI	22/27 (81.48%)	13/27 (48.15%)	5/27 (18.52%)	5/27 (18.52%)
5-T-ACS	−	−	−	8/27 (29.63%)
p value	0.031	0.500	0.500	0.039
ACS Artificial intelligence-assisted compressed sensing, AP Arterial phase, HBP Hepatobiliary phase, PI Parallel imaging;

No competing interests reported.

supplementarymaterialsAB.docx

Download PDF

Editorial decision: Revision requested
04 Jan, 2026
Reviews received at journal
27 Dec, 2025
Reviews received at journal
20 Dec, 2025
Reviewers agreed at journal
11 Dec, 2025
Reviewers agreed at journal
11 Dec, 2025
Reviewers agreed at journal
10 Dec, 2025
Reviewers agreed at journal
09 Dec, 2025
Reviewers invited by journal
08 Dec, 2025
Editor assigned by journal
05 Dec, 2025
Submission checks completed at journal
05 Dec, 2025
First submitted to journal
04 Dec, 2025

You are reading this latest preprint version

Intraindividual comparison of 3-T and 5-T gadoxetic acid-enhanced MRI for evaluating HCC: initial results

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Relevance statement

Figures

Purpose

Methods

Results

Discussion

Abbreviations

Declarations

References

Tables

Additional Declarations

Supplementary Files

Status:

Version 1