Deep Learning–Based SmartSpeed Precise DWI for Hepatocellular Carcinoma: Comparison with Conventional DWI

Deep Learning–Based SmartSpeed Precise DWI for Hepatocellular Carcinoma: Comparison with Conventional DWI | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Deep Learning–Based SmartSpeed Precise DWI for Hepatocellular Carcinoma: Comparison with Conventional DWI Joonyong Jang, Heejin Kwon, Arim Ji, Gyubin Lee, Enuju Kang, Sanghyun Kim, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9442562/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Purpose To evaluate the image quality and diagnostic performance of SmartSpeed Precise diffusion-weighted imaging (DWI), a deep learning–based denoising and super-resolution reconstruction technique, compared with conventional DWI for the detection of hepatocellular carcinoma (HCC). Materials and Methods This retrospective analysis included 84 consecutive patients (mean age, 66.8 ± 9.1 years; 65 men) with liver cirrhosis who underwent gadoxetic acid–enhanced MRI including both conventional DWI (C-DWI) and SmartSpeed Precise DWI (P-DWI) between September 2025 and January 2026. A total of 137 HCC lesions were evaluated. Two radiologists independently assessed qualitative image quality and measured signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and apparent diffusion coefficient (ADC) values. Diagnostic performance for HCC detection was compared among three image sets: dynamic MRI alone, dynamic MRI plus C-DWI, and dynamic MRI plus P-DWI. Sensitivity was compared using McNemar testing, and receiver operating characteristic (ROC) analysis was performed to identify the optimal tumor size cut-off for detectability. A p value < 0.05 was considered statistically significant. Results Acquisition time was identical for C-DWI and P-DWI, as both were reconstructed from the same raw dataset. Qualitative image quality, SNR, and CNR did not differ significantly between techniques (all p > 0.05). However, ADC values were significantly lower with P-DWI than with C-DWI for both readers (Reader 1: 1.15 ± 0.25 vs 1.29 ± 0.25 × 10⁻ 3 mm²/s, p = 0.013; Reader 2: 1.21 ± 0.30 vs 1.32 ± 0.28 × 10⁻ 3 mm²/s, p = 0.036). Sensitivity for HCC detection was 65.7% (90/137) for dynamic MRI alone, 92.7% (127/137) for dynamic MRI plus C-DWI, and 98.5% (135/137) for dynamic MRI plus P-DWI. Both DWI-containing image sets showed significantly higher sensitivity than dynamic MRI alone (both p < 0.001). However, the difference between P-DWI and C-DWI was not statistically significant (p = 0.125). ROC analysis revealed that tumor size ≤ 1.4 cm was a significant predictor of missed lesions on C-DWI (AUC = 0.777, p < 0.001), where P-DWI demonstrated a numerical diagnostic advantage in detecting these small HCCs. Conclusion Deep learning–based SmartSpeed Precise DWI provided image quality and HCC detection performance comparable to those of conventional DWI without additional acquisition time. It showed lower ADC values with AI-based reconstruction, and there was no difference in HCC diagnostic performance between the two techniques. These findings suggest that SmartSpeed Precise DWI may serve as a feasible alternative to conventional DWI for liver diffusion imaging. Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Hepatocellular carcinoma (HCC) is the most prevalent primary liver malignancy, accounting for approximately 80% of cases [ 1 ]. In 2022, liver cancer was the third leading cause of cancer-related mortality and the sixth most commonly diagnosed cancer worldwide [ 2 ]. The global burden of liver cancer is expected to increase substantially, with projections estimating a rise of more than 55% by 2040, corresponding to approximately 1.4 million new cases and 1.3 million deaths annually [ 3 ]. Diffusion-weighted imaging (DWI) characterizes the microscopic motion of water molecules within tissues and provides indirect information regarding tissue cellularity and cell membrane integrity [ 4 ]. More specifically, liver DWI plays an important role in detection and characterization of focal hepatic lesions [ 4 – 6 ], assessment of treatment response, and prediction of therapeutic outcomes in malignant tumors [ 7 ]. Because HCC typically exhibits increased cellular density with smaller tumor cells compared with normal hepatocytes, a greater number of cell membranes per voxel results in restricted diffusion and correspondingly increased signal intensity on DWI. Despite these advantages, liver DWI remains technically challenging due to long acquisition times, respiratory and cardiac motion artifacts, susceptibility-related distortions, and intrinsic limitations of echo-planar imaging [ 4 , 6 , 8 , 9 ]. SmartSpeed Precise is a vendor-developed, artificial intelligence–based reconstruction framework that incorporates dual convolutional neural networks (CNNs) to enable accelerated image acquisition while maintaining or improving image quality [ 10 ]. During the coil-combination step, Adaptive CS-NET [ 11 ] is applied to images acquired with Compressed SENSE [ 12 ], effectively reducing noise and undersampling-related artifacts. This mechanism brings about faster acquisition of higher-quality images [ 12 ]. Precise Image Net subsequently suppresses ringing artifacts and enhances spatial resolution [ 10 ]. Prior studies have demonstrated that denoising and super-resolution deep learning (DL) techniques improve image quality and reduce acquisition time in breast [ 13 ] and prostate MRI [ 14 , 15 ] while studies in knee [ 16 ] and cine cardiovascular MRI [ 17 ] have reported preservation of image integrity with accelerated image acquisition. However, the impact of denoising and super-resolution DL-based reconstruction on liver DWI, particularly in the context of HCC diagnosis, is yet to be established. Therefore, the purpose of this study was to evaluate whether liver DWI reconstructed with denoising and super-resolution DL framework (SmartSpeed Precise, hereafter referred to as P-DWI) provided a comparable or improved diagnostic performance for HCC compared with conventional DWI (hereafter referred to as C-DWI). Materials and Methods This study is a retrospective analysis of data collected as part of a prospective study, which was approved by our institutional review board, and with written informed consent of all patients included in this study. Philips Healthcare (Best, The Netherland) provided the hardware and software support for the reconstruction of P-DWI MRI. Study population We searched consecutive patients who underwent gadoxetic acid-enhanced MRI including C-DWI and P-DWI between September 2025 and January 2026. The inclusion criteria were as follows: (1) patients with liver cirrhosis who underwent liver MRI for HCC surveillance and (2) patients confirmed to have HCC based on typical contrast-enhanced MRI imaging findings, characterized by arterial enhancement followed by washout on portal venous or delayed phase images according to the Liver Imaging Reporting and Data System (LI-RADS) version 2018 [ 18 ], or confirmed by biopsy or subsequent imaging over 6 months. The exclusion criteria were as follows: (1) patients with only observations difficult to characterize because of small lesion size (< 5 mm) or suboptimal image quality (including transient severe motion during arterial phase), (2) patients with LR-M, and LR-TIV observations. A detailed flowchart of the study population is shown in Fig. 1 . Liver MRI protocol All MRI examinations were performed using a 3.0-T system (Ingenia Elition X; Philips Healthcare, Best, The Netherlands) equipped with a multi-channel phased-array body coil. The standard liver MRI protocol included a T1-weighted dual-echo sequence (in-phase and opposed-phase), a T2-weighted turbo spin-echo sequence, and dynamic contrast-enhanced imaging. For dynamic imaging, Gd-EOB-DTPA (Primovist; Bayer Schering Pharma, Berlin, Germany) was administered intravenously at a dose of 0.1 mL/kg at a rate of 1.0 mL/s, followed by a 15-mL saline flush. The C-DWI sequence was acquired using a respiratory-triggered single-shot echo-planar imaging technique with b-values of 50, 400, and 800 s/mm². The detailed acquisition parameters were as follows: repetition time (TR): 2064 ms, echo time (TE): 65 ms, field of view (FOV): 440 × 440 mm, matrix size: 192 × 192, slice thickness: 4.5 mm, intersection gap: 0.8 mm, number of excitations: 1, and SENSE acceleration factor: 2.5. The average acquisition time for the C-DWI sequence was 3 minutes and 36 seconds. The P-DWI was subsequently generated from the exact same raw data of the C-DWI, using a DL-based denoising and super-resolution algorithm of SmartSpeed Precise with no additional scanning time. Observation Registration A board-certified abdominal radiologist with 20 years of liver imaging experience identified consecutive cases that met the inclusion and exclusion criteria using a picture archiving and communication system. For each selected patient, the size and location of the observations were reported, and the three largest lesions were chosen for further image analysis. Subsequently, two board-certified radiologists with over 5 and 4 years of experience in liver imaging, respectively, reviewed the selected observations according to the imaging analysis protocol described below. Qualitative Image Analysis Two radiologists independently reviewed the anonymized images on a PACS system (Centricity, GE Healthcare). The readers were blinded to the specific pulse sequence information and clinical details. Using a 5-point Likert scale, the readers assessed four qualitative parameters: (1) Overall image quality, (2) Image quality at the liver dome, (3) Image sharpness, and (4) Artifacts. Both overall image quality and image quality at the liver dome were evaluated according to the following scale: 5. Excellent with sharp liver margin and minimal artifacts; 4. Good; 3. Moderate; 2. Poor; and 1. Non-diagnostic. Image sharpness was examined according to the following scale: 5. Sharp liver margin; 4. Mild blurring; 3. Moderate blurring; 2. Severe blurring; and 1. Non-diagnostic. The level of artifacts was assessed according to the following scale: 5. No artifacts; 4. Mild artifacts without interpretation interference; 3. Moderate; 2. Severe artifacts that affect interpretation; and 1. Non-diagnostic. Quantitative Image Analysis Quantitative image analyses were performed by placing regions of interest (ROIs) on the solid portion of the HCC lesions (SI lesion ) and the background liver parenchyma (SI liver ). Signal intensities were measured on the b-value 800 s/mm² images, while mean HCC ADC values were measured on the corresponding ADC maps. Care was taken to avoid major blood vessels, bile ducts, and artifacts. Signal intensity in the background air was estimated by placing an ROI just outside the body wall and the standard deviation was calculated (SD air ). The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated using the following equations: $$\:\text{S}\text{N}\text{R}=\frac{\text{S}\text{I}\:lesion}{\text{S}\text{D}\:air}$$ $$\:\text{C}\text{N}\text{R}=\frac{\text{S}\text{I}\:lesion-\text{S}\text{I}\:liver}{\text{S}\text{D}\:air}$$ HCC Detectability Analysis A total of 137 HCC lesions were assessed by two radiologists in three separate reading sessions using the following image sets: (1) Gadoxetic acid-enhanced Dynamic MRI alone; (2) Dynamic MRI combined with C-DWI (including b-value 800 s/mm² images and ADC maps); and (3) Dynamic MRI combined with P-DWI (including b-value 800 s/mm² images and ADC maps). Detectability of HCC lesions were scored according to the following scale: 5. Definite, 4. Probable, 3. Equivocal, 2. Less likely, and 1. Definitely absent. Scores from the two readers were averaged to obtain a mean score for each lesion. Mean scores ≥ 3 were considered positive for lesion detection, whereas scores < 3 were considered negative. Statistical analysis We utilized commercially available statistical software, specifically SPSS version 23.0 (IBM Corp., Armonk, NY, USA) and MedCalc version 16.4 (MedCalc Software bvba, Mariakerke, Belgium), for data analysis. A two-sided P value of less than 0.05 was considered statistically significant. Inter-reader agreement for MRI images was assessed using intraclass correlation coefficients (ICC), with the following thresholds: 0.80 excellent agreement. Qualitative parameters were compared using the Mann-Whitney U test while quantitative parameters were compared using Wilcoxon signed-rank test or the Mann-Whitney U test, depending on data distribution. The diagnostic performance of detecting HCC lesions was calculated for the three image sets (Dynamic MRI alone, Dynamic MRI + C-DWI, and Dynamic MRI + P-DWI) and compared using McNemar’s test. To evaluate the impact of tumor size on detectability, the sizes of detected and missed lesions were compared using the independent-samples t-test or Mann-Whitney U test. A receiver operating characteristic (ROC) curve analysis was performed, and the optimal cut-off size for HCC detection was determined using the Youden index. Result Clinical characteristics Baseline characteristics are summarized in Table 1 . The final study population consisted of 84 patients (mean age, 66.8 ± 9.1 years; range, 44–86 years), including 65 men and 19 women. All patients had liver cirrhosis. Etiologies included alcohol-related cirrhosis (n = 32, 38.1%), hepatitis B virus infection (n = 26, 30.9%), hepatitis C virus infection (n = 9, 10.7%), combined alcohol and HBV (n = 6, 7.1%), combined alcohol and HCV (n = 4, 4.8%), combined alcohol, HBV, and HCV (n = 1, 1.2%), nonalcoholic steatohepatitis (n = 3, 3.6%), and cryptogenic cirrhosis (n = 3, 3.6%). Four patients (4.8%) underwent biopsy and eight (9.5%) underwent surgical resection; the remaining lesions were diagnosed based on imaging criteria. Table 1 Baseline clinical characteristics. Data are presented as mean ± standard deviation or number of patients (percentages). (HBV: hepatitis B virus, HCV: hepatitis C virus, NASH: nonalcoholic steatohepatitis, SD: standard deviation) Characteristics Total (n: 84) Age (years, mean ± SD) All patients 66.8 ± 9.1 Males 64.8 ± 8.3 Females 74.8 ± 6.7 Sex ratio (M:F) 65:19 Etiology of cirrhosis, n (%) Alcohol 32 (38.1%) HBV 26 (30.9%) HCV 9 (10.7%) Alcohol + HBV 6 (7.1%) Alcohol + HCV 4 (4.8%) Alcohol + HBV + HCV 1 (1.2%) NASH 3 (3.6%) Cryptogenic 3 (3.6%) Diagnostic method, n (%) Biopsy 4 (4.8%) Surgery 8 (9.5%) Typical image 72 (85.7%) Qualitative Image Quality Assessment Inter-reader agreement for qualitative image quality was fair to moderate. ICC values ranged from 0.249 to 0.541 for C-DWI and from 0.294 to 0.468 for P-DWI (Table 2 ). Table 2 Inter-reader agreement for qualitative and quantitative image parameters of C-DWI and P-DWI. Inter-reader agreement was assessed using the intraclass correlation coefficient. (ADC: apparent diffusion coefficient, C-DWI: conventional diffusion-weighted imaging, CI: confidence interval, CNR: contrast-to-noise ratio, ICC: intraclass correlation coefficient, IQ: image quality, P-DWI: precise diffusion-weighted imaging, SNR: signal-to-noise ratio) C-DWI P-DWI ICC 95% CI ICC 95% CI Lower Limit Upper Limit Lower Limit Upper Limit Qualitative Data Overall image quality .541 .115 .753 .403 − .129 .684 IQ at dome .338 − .252 .650 .463 − .015 .716 Sharpness .474 .007 .722 .468 − .005 .719 Artifact .249 − .420 .603 .294 − .336 .626 Quantitative Data SNR .678 .397 .831 .067 − .767 .506 CNR .500 .053 .735 .154 − .605 .552 ADC value 10 − 3 mm 2 /s .874 .685 .913 .430 − .078 .698 No statistically significant differences were observed between C-DWI and P-DWI for qualitative parameter for either reader (all p > 0.05) (Table 3 ). Overall image quality scores were comparable for Reader 1 (3.08 ± 0.92 vs 3.25 ± 0.87; p = 0.389) and Reader 2 (2.82 ± 0.72 vs 3.00 ± 0.77; p = 0.328). Image quality at the liver dome was similar for Reader 1 (3.33 ± 1.21 vs 3.33 ± 1.18; p > 0.99) and Reader 2 (3.00 ± 0.78 vs 3.18 ± 0.78; p = 0.320). Sharpness did not differ significantly between sequences for Reader 1 (2.78 ± 1.17 vs 2.98 ± 1.14; p = 0.441) or Reader 2 (3.10 ± 0.78 vs 3.30 ± 0.65; p = 0.215). Artifact scores were also comparable for Reader 1 (3.28 ± 1.11 vs 3.33 ± 1.05; p = 0.836) and Reader 2 (3.08 ± 0.80 vs 3.18 ± 0.77; p = 0.563). Table 3 Comparison of qualitative and quantitative parameters between Conventional and Precise DWI. Data are presented as means ± standard deviations. A P value of < 0.05 was considered to indicate a statistically significant difference. (ADC: apparent diffusion coefficient, C-DWI: conventional diffusion-weighted imaging, CNR: contrast-to-noise ratio, IQ: image quality, P-DWI: precise diffusion-weighted imaging, SNR: signal-to-noise ratio) Reader 1 Reader 2 C-DWI P-DWI p-value C-DWI P-DWI p-value Qualitative Data Overall Image quality 3.08 ± 0.92 3.25 ± 0.87 0.389 2.82 ± 0.72 3.00 ± 0.77 0.328 IQ at dome 3.33 ± 1.21 3.33 ± 1.18 1.002 3.00 ± 0.78 3.18 ± 0.78 0.320 Sharpness 2.78 ± 1.17 2.98 ± 1.14 0.441 3.10 ± 0.78 3.30 ± 0.65 0.215 Artifact 3.28 ± 1.11 3.33 ± 1.05 0.836 3.08 ± 0.80 3.18 ± 0.77 0.563 Quantitative Data SNR 48.65 ± 34.47 47.25 ± 31.28 0.842 42.16 ± 23.10 39.84 ± 19.60 0.628 CNR 97.09 ± 99.52 93.75 ± 69.32 0.852 63.98 ± 49.75 60.45 ± 40.27 0.724 ADC value 10 − 3 mm 2 /s 1.29 ± 0.25 1.15 ± 0.25 0.013 1.32 ± 0.28 1.21 ± 0.30 0.036 Quantitative Image Quality Assessment There were no significant differences in SNR or CNR between C-DWI and P-DWI for either reader (all p > 0.05) (Table 3 ). For Reader 1, SNR was 48.65 ± 34.47 for C-DWI and 47.25 ± 31.28 for P-DWI (p = 0.842). For Reader 2, SNR was 42.16 ± 23.10 and 39.84 ± 19.60, respectively (p = 0.628). CNR values were similarly comparable between sequences for Reader 1 (97.09 ± 99.52 vs 93.75 ± 69.32; p = 0.852) and Reader 2 (63.98 ± 49.75 vs 60.45 ± 40.27; p = 0.724). ADC values were significantly lower on P-DWI than on C-DWI for both Reader 1 (1.15 ± 0.25 vs 1.29 ± 0.25 × 10⁻ 3 mm ² /s; p = 0.013) and Reader 2 (1.21 ± 0.30 vs 1.32 ± 0.28 × 10⁻ 3 mm²/s; p = 0.036). HCC Diagnostic Performance Analysis Sensitivity was highest for Dynamic MRI + P-DWI (98.5%, 135/137; 95% CI, 92.0–99.7%), followed by Dynamic MRI + C-DWI (92.7%, 127/137; 95% CI, 83.7–97.8%), and lowest for Dynamic MRI alone (65.7%, 90/137; 95% CI, 52.7–75.9%) (Table 4 ). Both combined image sets demonstrated significantly higher sensitivity than Dynamic MRI alone (both p < 0.001, McNemar test). Although sensitivity was numerically higher for P-DWI than for C-DWI, the difference was not statistically significant (p = 0.125). Figures 3 and 4 are representative examples of HCC lesions detected exclusively on P-DWI, as well as those visualized on both P-DWI and C-DWI. Table 4 Diagnostic performance for the detection of hepatocellular carcinoma. Sensitivity was calculated for the detection of 137 HCC lesions. P values were determined using McNemar’s test for pairwise comparisons of diagnostic sensitivity among the image sets. (C-DWI: conventional diffusion-weighted imaging, CI: confidence interval, P-DWI: precise diffusion-weighted imaging) Image Set Sensitivity(%) 95% CI p-value (vs. Dynamic MRI alone) p-value (vs. Dynamic MRI + C-DWI) Dynamic MRI alone 65.7 (90/137) 52.7–75.9 - - Dynamic MRI + C-DWI 92.7 (127/137) 83.7–97.8 < 0.001 - Dynamic MRI + P-DWI 98.5 (135/137) 92.0–99.7 < 0.001 0.125 HCC lesions missed on Dynamic MRI alone were significantly smaller than detected lesions (13.42 ± 22.19 mm vs 30.48 ± 26.52 mm; p < 0.001). Similarly, lesions missed on C-DWI were smaller than detected lesions (9.00 ± 2.79 mm vs 24.18 ± 26.34 mm; p = 0.039). ROC analysis (Fig. 2 ) demonstrated that tumor size predicted detectability on C-DWI with an optimal cutoff value of 1.4 cm (AUC = 0.777; p < 0.001). For lesions ≥ 1.4 cm, detection rates were comparable among all image sets. Discussion We evaluated whether liver DWI reconstructed with SmartSpeed Precise improves image quality and HCC detection performance compared with conventional DWI. Our results showed that P-DWI provided numerically higher qualitative image quality while maintaining comparable SNR and CNR relative to C-DWI, without additional acquisition time. ADC values were significantly lower on P-DWI than on C-DWI for both readers. In addition, overall sensitivity for HCC detection was numerically higher with P-DWI. These findings suggest that SmartSpeed Precise reconstruction may offer incremental benefits, particularly for small HCCs over conventional liver DWI, particularly for small HCCs. Although P-DWI demonstrated numerically higher sensitivity, the difference did not reach statistical significance, likely reflecting the already high baseline performance of conventional DWI. This finding is consistent with prior studies comparing deep learning (DL)-accelerated liver MRI with conventional techniques. Park et al. [ 19 ] reported that a DL-accelerated abbreviated MRI protocol achieved comparable diagnostic sensitivity to a standard protocol, with no significant difference in HCC detection (p > 0.05). Similarly, Kim et al. [ 20 ] found no significant difference in per-lesion sensitivity between DL-DWI and conventional DWI (p ≥ 0.062). Collectively, these studies suggest that DL-based reconstruction techniques can maintain, and in some cases modestly improve, the already high diagnostic performance of conventional DWI without additional scan time. Notably, P-DWI demonstrated a diagnostic advantage for lesions smaller than 1.4 cm. ROC analysis indicated that tumor size significantly influenced detectability on C-DWI, with an optimal cutoff of 1.4 cm. While detection rates were comparable for lesions larger than 1.4 cm, P-DWI combined with dynamic MRI identified numerically more sub-1.4 cm lesions than C-DWI. This improvement is likely related to the combined effects of noise reduction, artifact suppression, and enhanced spatial resolution provided by the dual convolutional neural network architecture. The Adaptive-CS-Net reduces background noise, whereas the Precise Image Net suppresses ringing artifacts and sharpens lesion margins. Although not statistically significant, qualitative assessments showed a trend toward improved sharpness and fewer artifacts with P-DWI. These findings are in line with Zhao et al. [ 21 ], who demonstrated improved lesion conspicuity and diagnostic sensitivity using a similar DL-based framework. Comparable benefits have also been reported in breast, prostate, and musculoskeletal MRI. Thus, while conventional DWI remains adequate for larger tumors, P-DWI may provide added value in detecting small HCCs, which are often more challenging to identify. We observed significantly lower ADC values on P-DWI compared with C-DWI. This phenomenon has been reported in prior studies using DL-based reconstruction techniques [ 21 , 22 ], although it’s underlying mechanism remains incompletely understood. One plausible explanation is related to noise suppression at high b-values. In conventional DWI, low SNR at high b-values results in a noise floor effect, which can artificially elevate signal intensity. By suppressing background noise, the DL algorithm may reduce this artificial signal elevation, leading to a greater apparent signal decay between b-values and consequently lower ADC measurements. This explanation remains hypothetical and warrants further investigation. This study has several limitations. First, its retrospective single-center design introduces potential selection bias and limits generalizability. Second, the DL reconstruction algorithm is vendor-specific; therefore, the results may not be directly applicable to other platforms with different network architectures or training datasets. Third, quantitative analysis of SNR and CNR was limited to a b-value of 800 s/mm². The performance of the DL algorithm at higher b-values (e.g., ≥ 1000 s/mm²), where SNR is further degraded, was not evaluated and may differ. In conclusion, SmartSpeed Precise DWI provides comparable overall diagnostic performance to conventional DWI while offering potential advantages in image quality and sensitivity for small HCCs, without increasing scan time. The observed reduction in ADC values likely reflects the effects of DL-based noise suppression, although further studies are needed to clarify the underlying mechanism. Declarations Author Contribution JY.J and HJ.K wrote the main manuscript text and AR. J and GB.L analyzed the data. All authors reviewed the manuscript. References Wild CP, Weiderpass E, Stewart BWE. IARC World Cancer Reports. In. World Cancer Report: Cancer research for cancer prevention. Lyon (FR): I, 2020 Bray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024; 74:229–263 Rumgay H, Arnold M, Ferlay J, Lesi O, Cabasag CJ, Vignat J, et al. Global burden of primary liver cancer in 2020 and predictions to 2040. Journal of Hepatology 2022; 77:1598–1606 Koh DM, Collins DJ. Diffusion-weighted MRI in the body: applications and challenges in oncology. AJR Am J Roentgenol 2007; 188:1622–1635 Galea N, Cantisani V, Taouli B. Liver lesion detection and characterization: role of diffusion-weighted imaging. J Magn Reson Imaging 2013; 37:1260–1276 Taouli B, Koh DM. Diffusion-weighted MR imaging of the liver. Radiology 2010; 254:47–66 Shenoy-Bhangle A, Baliyan V, Kordbacheh H, Guimaraes AR, Kambadakone A. Diffusion weighted magnetic resonance imaging of liver: Principles, clinical applications and recent updates. World J Hepatol 2017; 9:1081–1091 Gholipour A, Kehtarnavaz N, Scherrer B, Warfield SK. On the accuracy of unwarping techniques for the correction of susceptibility-induced geometric distortion in magnetic resonance Echo-planar images. Annu Int Conf IEEE Eng Med Biol Soc 2011; 2011:6997–7000 Kondo F, Wada K, Kondo Y. Morphometric analysis of hepatocellular carcinoma. Virchows Arch A Pathol Anat Histopathol 1988; 413:425–430 PhilipsHealthcare. SmartSpeed Precise: Ultimate precision at maximum speed with Dual Al - MR Clinical Application [Brochure] (accessed on 28 December 2025). In: Philips SmartSpeed Precise, 2025 Pezzotti N, de Weerdt E, Yousefim S, Elmahdy MS, van Gemert J, Schülke C, et al. Adaptive-CS-Net: FastMRI with Adaptive Intelligence. arXiv 2019 Peeters H, Chung H, Valvano G, Yakisikli D, Van Gemert J, De Weerdt E, et al. Philips SmartSpeed. No compromise Image quality and speed at your fingertips [White paper] (accessed on 28 December 2025). In: Philips SmartSpeed, 2021 Mesropyan N, Katemann C, Leutner C, Sommer A, Isaak A, Weber OM, et al. Accelerated High-resolution T1- and T2-weighted Breast MRI with Deep Learning Super-resolution Reconstruction. Acad Radiol 2025; 32:3147–3156 Bischoff LM, Peeters JM, Weinhold L, Krausewitz P, Ellinger J, Katemann C, et al. Deep Learning Super-Resolution Reconstruction for Fast and Motion-Robust T2-weighted Prostate MRI. Radiology 2023;308: e230427 Jurka M, Macova I, Wagnerova M, Capoun O, Jakubicek R, Ourednicek P, et al. Deep-learning-based reconstruction of T2-weighted magnetic resonance imaging of the prostate accelerated by compressed sensing provides improved image quality at half the acquisition time. Quant Imaging Med Surg 2024; 14:3534–3543 Terzis R, Dratsch T, Hahnfeldt R, Basten L, Rauen P, Sonnabend K, et al. Five-minute knee MRI: An AI-based super resolution reconstruction approach for compressed sensing. A validation study on healthy volunteers. Eur J Radiol 2024; 175:111418 Kravchenko D, Isaak A, Mesropyan N, Peeters JM, Kuetting D, Pieper CC, et al. Deep learning super-resolution reconstruction for fast and high-quality cine cardiovascular magnetic resonance. Eur Radiol 2025; 35:2877–2887 Radiology ACo. Li-RADS CT/MRI version 2018 Core. https: edge.sitecorecloud.io/americancoldf5f-acrorgf92a-productioncb02-3650/media/ACR/Files/RADS/LI-RADS/LI-RADS-CT-MRI-2018-Core.pdf (accessed 26 Feb 2026) 2018 Park SH, Choi MH, Kim B, Lee HS, Yoon S, Lee YJ, et al. Deep Learning-Accelerated Non-Contrast Abbreviated Liver MRI for Detecting Malignant Focal Hepatic Lesions: Dual-Center Validation. Korean J Radiol 2025;26:333–345 Kim DH, Kim B, Lee HS, Benkert T, Kim H, Choi JI, et al. Deep Learning-Accelerated Liver Diffusion-Weighted Imaging: Intraindividual Comparison and Additional Phantom Study of Free-Breathing and Respiratory-Triggering Acquisitions. Invest Radiol 2023;58:782–790 Zhao D, Kong X, Yang K, Wan J, Liu Z, Pan F, et al. Deep learning-enhanced super-resolution diffusion-weighted liver MRI: improved image quality, diagnostic performance, and acceleration. Insights Imaging 2025;16:273 Bae SH, Hwang J, Hong SS, Lee EJ, Jeong J, Benkert T, et al. Clinical feasibility of accelerated diffusion weighted imaging of the abdomen with deep learning reconstruction: Comparison with conventional diffusion weighted imaging. Eur J Radiol 2022;154:110428 Dietrich O, Heiland S, Sartor K. Noise correction for the exact determination of apparent diffusion coefficients at low SNR. Magn Reson Med 2001;45:448–453 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9442562","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":630983461,"identity":"a6ce5142-9e9c-4f8f-80fd-222740223668","order_by":0,"name":"Joonyong Jang","email":"","orcid":"","institution":"Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Joonyong","middleName":"","lastName":"Jang","suffix":""},{"id":630983463,"identity":"2edad768-109a-4470-856b-4c4e5751fef2","order_by":1,"name":"Heejin Kwon","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAu0lEQVRIiWNgGAWjYDACZhBRYMPYAOEmEKvFII0ULWBgcJgELebt7M8kfhicl+2fkcD44QdDWj5BLTKHecwkewxuG8+4kcAs2cOQY9lASIsEMw+bBI/B7cSGGwkM0gwMFQYEbZFgZn8m+cfgXOJ8oC2/idTCYCbNY3AgccONBDagLTnEaOExtpYxSDbeeOZhm2WPQRoRWviPP7z5psJOdt7x5MM3flQkE9aCBEBRQ5KGUTAKRsEoGAU4AQAfUjOfvfpo+wAAAABJRU5ErkJggg==","orcid":"","institution":"Dong-A University","correspondingAuthor":true,"prefix":"","firstName":"Heejin","middleName":"","lastName":"Kwon","suffix":""},{"id":630983464,"identity":"25b04900-9d71-4b83-968a-3ae4cd12b7e3","order_by":2,"name":"Arim Ji","email":"","orcid":"","institution":"Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Arim","middleName":"","lastName":"Ji","suffix":""},{"id":630983467,"identity":"f1a579d0-8eff-48c9-bf8c-597ecfba5975","order_by":3,"name":"Gyubin Lee","email":"","orcid":"","institution":"Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Gyubin","middleName":"","lastName":"Lee","suffix":""},{"id":630983468,"identity":"92bb7c87-aa4d-4068-9b38-e8c598f34966","order_by":4,"name":"Enuju Kang","email":"","orcid":"","institution":"Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Enuju","middleName":"","lastName":"Kang","suffix":""},{"id":630983469,"identity":"b02da9a4-1e33-4c57-bb05-e9f69e660ad2","order_by":5,"name":"Sanghyun Kim","email":"","orcid":"","institution":"Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Sanghyun","middleName":"","lastName":"Kim","suffix":""},{"id":630983470,"identity":"f1e8c8e3-db4f-473c-ad1c-9a2049264aa5","order_by":6,"name":"Jongwook Lim","email":"","orcid":"","institution":"Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Jongwook","middleName":"","lastName":"Lim","suffix":""},{"id":630983471,"identity":"fcd5ba20-85cb-4064-a498-f6cb7e683c3b","order_by":7,"name":"Myungjin Kang","email":"","orcid":"","institution":"Dong-A University","correspondingAuthor":false,"prefix":"","firstName":"Myungjin","middleName":"","lastName":"Kang","suffix":""}],"badges":[],"createdAt":"2026-04-17 00:23:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9442562/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9442562/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108383210,"identity":"c9d69c68-5088-4242-9c8f-f73b5523538f","added_by":"auto","created_at":"2026-05-04 05:44:47","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":113316,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of the study population. Flow diagram illustrates the patient and observation selection process (HCC: hepatocellular carcinoma, LI-RADS: Liver Imaging Reporting and Data System, LR-M: probably or definitely malignant but not specific for HCC, LR-TIV: tumor in vein, MRI: magnetic resonance imaging)\u003c/p\u003e","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9442562/v1/8822af642c9ad6834347bc91.jpg"},{"id":108493083,"identity":"b0231774-5e8d-4b80-a98d-5bfe83ebe2fc","added_by":"auto","created_at":"2026-05-05 09:59:21","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":41259,"visible":true,"origin":"","legend":"\u003cp\u003eReceiver operating characteristic curve evaluating the impact of tumor size on hepatocellular carcinoma detectability using the C-DWI set. Based on the Youden index, the optimal cut-off size was 1.4 cm (area under the curve (AUC) = 0.777, \u003cem\u003ep\u003c/em\u003e \u0026lt; 0.001)\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9442562/v1/4a82d31b66cdea6be02705de.jpg"},{"id":108383211,"identity":"fc243856-72ff-465a-b883-cb5a35a79788","added_by":"auto","created_at":"2026-05-04 05:44:48","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":117751,"visible":true,"origin":"","legend":"\u003cp\u003eRepresentative dynamic contrast-enhanced and high b-value images (b = 800 s/mm²) of C-DWI and P-DWI. MRI scans of an 80-year-old male patient with a history of HCC and underlying liver cirrhosis. A focal HCC in the dome portion of segment 8 is not visualized on the dynamic contrast-enhanced late arterial phase (A) or C-DWI, but is well-delineated on P-DWI (D) and hepatobiliary phase (B). (C-DWI, conventional diffusion-weighted imaging; HCC, hepatocellular carcinoma; P-DWI, precise diffusion-weighted imaging)\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9442562/v1/26bef19ff5db43b7e5be0189.jpg"},{"id":108492325,"identity":"f7adb43f-0e9d-4984-859e-8961204a4a2e","added_by":"auto","created_at":"2026-05-05 09:57:28","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":103686,"visible":true,"origin":"","legend":"\u003cp\u003eRepresentative dynamic contrast-enhanced and high b-value images (b = 800 s/mm²) of C-DWI and P-DWI. MRI scans of a 74-year-old male patient with a history of HCC and underlying liver cirrhosis. A focal HCC in segment 7 is not visualized on the dynamic contrast-enhanced late arterial phase (A) but is well-delineated on C-DWI (C), P-DWI (D) and hepatobiliary phase (B). (C-DWI, conventional diffusion-weighted imaging; HCC, hepatocellular carcinoma; P-DWI, precise diffusion-weighted imaging)\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9442562/v1/f02b4e7aadc71f90bacfefef.jpg"},{"id":108610522,"identity":"74a64ea4-32c8-4624-93e4-bafa66802519","added_by":"auto","created_at":"2026-05-06 12:57:35","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":691743,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9442562/v1/93f3488a-3bd5-442c-855f-55bfa628bee6.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Deep Learning–Based SmartSpeed Precise DWI for Hepatocellular Carcinoma: Comparison with Conventional DWI","fulltext":[{"header":"Introduction","content":"\u003cp\u003eHepatocellular carcinoma (HCC) is the most prevalent primary liver malignancy, accounting for approximately 80% of cases [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. In 2022, liver cancer was the third leading cause of cancer-related mortality and the sixth most commonly diagnosed cancer worldwide [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. The global burden of liver cancer is expected to increase substantially, with projections estimating a rise of more than 55% by 2040, corresponding to approximately 1.4\u0026nbsp;million new cases and 1.3\u0026nbsp;million deaths annually [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eDiffusion-weighted imaging (DWI) characterizes the microscopic motion of water molecules within tissues and provides indirect information regarding tissue cellularity and cell membrane integrity [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. More specifically, liver DWI plays an important role in detection and characterization of focal hepatic lesions [\u003cspan additionalcitationids=\"CR5\" citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], assessment of treatment response, and prediction of therapeutic outcomes in malignant tumors [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Because HCC typically exhibits increased cellular density with smaller tumor cells compared with normal hepatocytes, a greater number of cell membranes per voxel results in restricted diffusion and correspondingly increased signal intensity on DWI. Despite these advantages, liver DWI remains technically challenging due to long acquisition times, respiratory and cardiac motion artifacts, susceptibility-related distortions, and intrinsic limitations of echo-planar imaging [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eSmartSpeed Precise is a vendor-developed, artificial intelligence\u0026ndash;based reconstruction framework that incorporates dual convolutional neural networks (CNNs) to enable accelerated image acquisition while maintaining or improving image quality [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. During the coil-combination step, Adaptive CS-NET [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] is applied to images acquired with Compressed SENSE [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e], effectively reducing noise and undersampling-related artifacts. This mechanism brings about faster acquisition of higher-quality images [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Precise Image Net subsequently suppresses ringing artifacts and enhances spatial resolution [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Prior studies have demonstrated that denoising and super-resolution deep learning (DL) techniques improve image quality and reduce acquisition time in breast [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e] and prostate MRI [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] while studies in knee [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] and cine cardiovascular MRI [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e] have reported preservation of image integrity with accelerated image acquisition. However, the impact of denoising and super-resolution DL-based reconstruction on liver DWI, particularly in the context of HCC diagnosis, is yet to be established.\u003c/p\u003e \u003cp\u003eTherefore, the purpose of this study was to evaluate whether liver DWI reconstructed with denoising and super-resolution DL framework (SmartSpeed Precise, hereafter referred to as P-DWI) provided a comparable or improved diagnostic performance for HCC compared with conventional DWI (hereafter referred to as C-DWI).\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e This study is a retrospective analysis of data collected as part of a prospective study, which was approved by our institutional review board, and with written informed consent of all patients included in this study. Philips Healthcare (Best, The Netherland) provided the hardware and software support for the reconstruction of P-DWI MRI.\u003c/p\u003e \u003cp\u003eStudy population\u003c/p\u003e \u003cp\u003eWe searched consecutive patients who underwent gadoxetic acid-enhanced MRI including C-DWI and P-DWI between September 2025 and January 2026. The inclusion criteria were as follows: (1) patients with liver cirrhosis who underwent liver MRI for HCC surveillance and (2) patients confirmed to have HCC based on typical contrast-enhanced MRI imaging findings, characterized by arterial enhancement followed by washout on portal venous or delayed phase images according to the Liver Imaging Reporting and Data System (LI-RADS) version 2018 [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e], or confirmed by biopsy or subsequent imaging over 6 months. The exclusion criteria were as follows: (1) patients with only observations difficult to characterize because of small lesion size (\u0026lt;\u0026thinsp;5 mm) or suboptimal image quality (including transient severe motion during arterial phase), (2) patients with LR-M, and LR-TIV observations. A detailed flowchart of the study population is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eLiver MRI protocol\u003c/p\u003e \u003cp\u003eAll MRI examinations were performed using a 3.0-T system (Ingenia Elition X; Philips Healthcare, Best, The Netherlands) equipped with a multi-channel phased-array body coil. The standard liver MRI protocol included a T1-weighted dual-echo sequence (in-phase and opposed-phase), a T2-weighted turbo spin-echo sequence, and dynamic contrast-enhanced imaging. For dynamic imaging, Gd-EOB-DTPA (Primovist; Bayer Schering Pharma, Berlin, Germany) was administered intravenously at a dose of 0.1 mL/kg at a rate of 1.0 mL/s, followed by a 15-mL saline flush.\u003c/p\u003e \u003cp\u003eThe C-DWI sequence was acquired using a respiratory-triggered single-shot echo-planar imaging technique with b-values of 50, 400, and 800 s/mm\u0026sup2;. The detailed acquisition parameters were as follows: repetition time (TR): 2064 ms, echo time (TE): 65 ms, field of view (FOV): 440 \u0026times; 440 mm, matrix size: 192 \u0026times; 192, slice thickness: 4.5 mm, intersection gap: 0.8 mm, number of excitations: 1, and SENSE acceleration factor: 2.5. The average acquisition time for the C-DWI sequence was 3 minutes and 36 seconds.\u003c/p\u003e \u003cp\u003eThe P-DWI was subsequently generated from the exact same raw data of the C-DWI, using a DL-based denoising and super-resolution algorithm of SmartSpeed Precise with no additional scanning time.\u003c/p\u003e \u003cp\u003eObservation Registration\u003c/p\u003e \u003cp\u003eA board-certified abdominal radiologist with 20 years of liver imaging experience identified consecutive cases that met the inclusion and exclusion criteria using a picture archiving and communication system. For each selected patient, the size and location of the observations were reported, and the three largest lesions were chosen for further image analysis. Subsequently, two board-certified radiologists with over 5 and 4 years of experience in liver imaging, respectively, reviewed the selected observations according to the imaging analysis protocol described below.\u003c/p\u003e \u003cp\u003eQualitative Image Analysis\u003c/p\u003e \u003cp\u003eTwo radiologists independently reviewed the anonymized images on a PACS system (Centricity, GE Healthcare). The readers were blinded to the specific pulse sequence information and clinical details. Using a 5-point Likert scale, the readers assessed four qualitative parameters: (1) Overall image quality, (2) Image quality at the liver dome, (3) Image sharpness, and (4) Artifacts. Both overall image quality and image quality at the liver dome were evaluated according to the following scale: 5. Excellent with sharp liver margin and minimal artifacts; 4. Good; 3. Moderate; 2. Poor; and 1. Non-diagnostic. Image sharpness was examined according to the following scale: 5. Sharp liver margin; 4. Mild blurring; 3. Moderate blurring; 2. Severe blurring; and 1. Non-diagnostic. The level of artifacts was assessed according to the following scale: 5. No artifacts; 4. Mild artifacts without interpretation interference; 3. Moderate; 2. Severe artifacts that affect interpretation; and 1. Non-diagnostic.\u003c/p\u003e \u003cp\u003eQuantitative Image Analysis\u003c/p\u003e \u003cp\u003eQuantitative image analyses were performed by placing regions of interest (ROIs) on the solid portion of the HCC lesions (SI \u003cem\u003elesion\u003c/em\u003e) and the background liver parenchyma (SI \u003cem\u003eliver\u003c/em\u003e). Signal intensities were measured on the b-value 800 s/mm\u0026sup2; images, while mean HCC ADC values were measured on the corresponding ADC maps. Care was taken to avoid major blood vessels, bile ducts, and artifacts. Signal intensity in the background air was estimated by placing an ROI just outside the body wall and the standard deviation was calculated (SD \u003cem\u003eair\u003c/em\u003e). The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were calculated using the following equations:\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:\\text{S}\\text{N}\\text{R}=\\frac{\\text{S}\\text{I}\\:lesion}{\\text{S}\\text{D}\\:air}$$\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:\\text{C}\\text{N}\\text{R}=\\frac{\\text{S}\\text{I}\\:lesion-\\text{S}\\text{I}\\:liver}{\\text{S}\\text{D}\\:air}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eHCC Detectability Analysis\u003c/p\u003e \u003cp\u003eA total of 137 HCC lesions were assessed by two radiologists in three separate reading sessions using the following image sets: (1) Gadoxetic acid-enhanced Dynamic MRI alone; (2) Dynamic MRI combined with C-DWI (including b-value 800 s/mm\u0026sup2; images and ADC maps); and (3) Dynamic MRI combined with P-DWI (including b-value 800 s/mm\u0026sup2; images and ADC maps). Detectability of HCC lesions were scored according to the following scale: 5. Definite, 4. Probable, 3. Equivocal, 2. Less likely, and 1. Definitely absent. Scores from the two readers were averaged to obtain a mean score for each lesion. Mean scores\u0026thinsp;\u0026ge;\u0026thinsp;3 were considered positive for lesion detection, whereas scores\u0026thinsp;\u0026lt;\u0026thinsp;3 were considered negative.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eWe utilized commercially available statistical software, specifically SPSS version 23.0 (IBM Corp., Armonk, NY, USA) and MedCalc version 16.4 (MedCalc Software bvba, Mariakerke, Belgium), for data analysis. A two-sided P value of less than 0.05 was considered statistically significant.\u003c/p\u003e \u003cp\u003eInter-reader agreement for MRI images was assessed using intraclass correlation coefficients (ICC), with the following thresholds: \u0026lt; 0.20 indicating poor agreement, 0.20\u0026ndash;0.39 fair agreement, 0.40\u0026ndash;0.59 moderate agreement, 0.60\u0026ndash;0.79 substantial agreement, and \u0026gt;\u0026thinsp;0.80 excellent agreement. Qualitative parameters were compared using the Mann-Whitney U test while quantitative parameters were compared using Wilcoxon signed-rank test or the Mann-Whitney U test, depending on data distribution.\u003c/p\u003e \u003cp\u003eThe diagnostic performance of detecting HCC lesions was calculated for the three image sets (Dynamic MRI alone, Dynamic MRI\u0026thinsp;+\u0026thinsp;C-DWI, and Dynamic MRI\u0026thinsp;+\u0026thinsp;P-DWI) and compared using McNemar\u0026rsquo;s test. To evaluate the impact of tumor size on detectability, the sizes of detected and missed lesions were compared using the independent-samples t-test or Mann-Whitney U test. A receiver operating characteristic (ROC) curve analysis was performed, and the optimal cut-off size for HCC detection was determined using the Youden index.\u003c/p\u003e \u003c/div\u003e"},{"header":"Result","content":"\u003cp\u003eClinical characteristics\u003c/p\u003e \u003cp\u003eBaseline characteristics are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The final study population consisted of 84 patients (mean age, 66.8\u0026thinsp;\u0026plusmn;\u0026thinsp;9.1 years; range, 44\u0026ndash;86 years), including 65 men and 19 women. All patients had liver cirrhosis. Etiologies included alcohol-related cirrhosis (n\u0026thinsp;=\u0026thinsp;32, 38.1%), hepatitis B virus infection (n\u0026thinsp;=\u0026thinsp;26, 30.9%), hepatitis C virus infection (n\u0026thinsp;=\u0026thinsp;9, 10.7%), combined alcohol and HBV (n\u0026thinsp;=\u0026thinsp;6, 7.1%), combined alcohol and HCV (n\u0026thinsp;=\u0026thinsp;4, 4.8%), combined alcohol, HBV, and HCV (n\u0026thinsp;=\u0026thinsp;1, 1.2%), nonalcoholic steatohepatitis (n\u0026thinsp;=\u0026thinsp;3, 3.6%), and cryptogenic cirrhosis (n\u0026thinsp;=\u0026thinsp;3, 3.6%). Four patients (4.8%) underwent biopsy and eight (9.5%) underwent surgical resection; the remaining lesions were diagnosed based on imaging criteria.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eBaseline clinical characteristics. Data are presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation or number of patients (percentages). (HBV: hepatitis B virus, HCV: hepatitis C virus, NASH: nonalcoholic steatohepatitis, SD: standard deviation)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristics\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTotal (n: 84)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eAge (years, mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAll patients\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e66.8\u0026thinsp;\u0026plusmn;\u0026thinsp;9.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMales\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e64.8\u0026thinsp;\u0026plusmn;\u0026thinsp;8.3\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemales\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e74.8\u0026thinsp;\u0026plusmn;\u0026thinsp;6.7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eSex ratio (M:F)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e65:19\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eEtiology of cirrhosis, n (%)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAlcohol\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e32 (38.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHBV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e26 (30.9%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHCV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9 (10.7%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAlcohol\u0026thinsp;+\u0026thinsp;HBV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6 (7.1%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAlcohol\u0026thinsp;+\u0026thinsp;HCV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4 (4.8%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAlcohol\u0026thinsp;+\u0026thinsp;HBV\u0026thinsp;+\u0026thinsp;HCV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1 (1.2%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNASH\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3 (3.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCryptogenic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3 (3.6%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDiagnostic method, n (%)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBiopsy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4 (4.8%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSurgery\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8 (9.5%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTypical image\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e72 (85.7%)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eQualitative Image Quality Assessment\u003c/p\u003e \u003cp\u003eInter-reader agreement for qualitative image quality was fair to moderate. ICC values ranged from 0.249 to 0.541 for C-DWI and from 0.294 to 0.468 for P-DWI (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eInter-reader agreement for qualitative and quantitative image parameters of C-DWI and P-DWI. Inter-reader agreement was assessed using the intraclass correlation coefficient. (ADC: apparent diffusion coefficient, C-DWI: conventional diffusion-weighted imaging, CI: confidence interval, CNR: contrast-to-noise ratio, ICC: intraclass correlation coefficient, IQ: image quality, P-DWI: precise diffusion-weighted imaging, SNR: signal-to-noise ratio)\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e\n \u003cp\u003eC-DWI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colspan=\"3\" nameend=\"c7\" namest=\"c5\"\u003e\n \u003cp\u003eP-DWI\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\n \u003cp\u003eICC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colspan=\"2\" nameend=\"c4\" namest=\"c3\"\u003e\n \u003cp\u003e95% CI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e\n \u003cp\u003eICC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colspan=\"2\" nameend=\"c7\" namest=\"c6\"\u003e\n \u003cp\u003e95% CI\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003eLower Limit\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003eUpper Limit\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003eLower Limit\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003eUpper Limit\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003e\u003cstrong\u003eQualitative Data\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colspan=\"6\" nameend=\"c7\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eOverall image quality\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e.541\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e.115\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e.753\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e.403\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026minus;\u0026thinsp;.129\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e.684\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eIQ at dome\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e.338\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e\u0026minus;\u0026thinsp;.252\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e.650\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e.463\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026minus;\u0026thinsp;.015\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e.716\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSharpness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e.474\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e.007\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e.722\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e.468\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026minus;\u0026thinsp;.005\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e.719\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eArtifact\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e.249\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e\u0026minus;\u0026thinsp;.420\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e.603\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e.294\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026minus;\u0026thinsp;.336\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e.626\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003e\u003cstrong\u003eQuantitative Data\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colspan=\"6\" nameend=\"c7\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eSNR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e.678\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e.397\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e.831\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e.067\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026minus;\u0026thinsp;.767\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e.506\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eCNR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e.500\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e.053\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e.735\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e.154\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026minus;\u0026thinsp;.605\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e.552\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colname=\"c1\"\u003e\n \u003cp\u003eADC value 10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003emm\u003csup\u003e2\u003c/sup\u003e/s\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c2\"\u003e\n \u003cp\u003e.874\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c3\"\u003e\n \u003cp\u003e.685\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c4\"\u003e\n \u003cp\u003e.913\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c5\"\u003e\n \u003cp\u003e.430\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c6\"\u003e\n \u003cp\u003e\u0026minus;\u0026thinsp;.078\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" colname=\"c7\"\u003e\n \u003cp\u003e.698\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eNo statistically significant differences were observed between C-DWI and P-DWI for qualitative parameter for either reader (all p\u0026thinsp;\u0026gt;\u0026thinsp;0.05) (Table \u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Overall image quality scores were comparable for Reader 1 (3.08\u0026thinsp;\u0026plusmn;\u0026thinsp;0.92 vs 3.25\u0026thinsp;\u0026plusmn;\u0026thinsp;0.87; p\u0026thinsp;=\u0026thinsp;0.389) and Reader 2 (2.82\u0026thinsp;\u0026plusmn;\u0026thinsp;0.72 vs 3.00\u0026thinsp;\u0026plusmn;\u0026thinsp;0.77; p\u0026thinsp;=\u0026thinsp;0.328). Image quality at the liver dome was similar for Reader 1 (3.33\u0026thinsp;\u0026plusmn;\u0026thinsp;1.21 vs 3.33\u0026thinsp;\u0026plusmn;\u0026thinsp;1.18; p\u0026thinsp;\u0026gt;\u0026thinsp;0.99) and Reader 2 (3.00\u0026thinsp;\u0026plusmn;\u0026thinsp;0.78 vs 3.18\u0026thinsp;\u0026plusmn;\u0026thinsp;0.78; p\u0026thinsp;=\u0026thinsp;0.320). Sharpness did not differ significantly between sequences for Reader 1 (2.78\u0026thinsp;\u0026plusmn;\u0026thinsp;1.17 vs 2.98\u0026thinsp;\u0026plusmn;\u0026thinsp;1.14; p\u0026thinsp;=\u0026thinsp;0.441) or Reader 2 (3.10\u0026thinsp;\u0026plusmn;\u0026thinsp;0.78 vs 3.30\u0026thinsp;\u0026plusmn;\u0026thinsp;0.65; p\u0026thinsp;=\u0026thinsp;0.215). Artifact scores were also comparable for Reader 1 (3.28\u0026thinsp;\u0026plusmn;\u0026thinsp;1.11 vs 3.33\u0026thinsp;\u0026plusmn;\u0026thinsp;1.05; p\u0026thinsp;=\u0026thinsp;0.836) and Reader 2 (3.08\u0026thinsp;\u0026plusmn;\u0026thinsp;0.80 vs 3.18\u0026thinsp;\u0026plusmn;\u0026thinsp;0.77; p\u0026thinsp;=\u0026thinsp;0.563).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eComparison of qualitative and quantitative parameters between Conventional and Precise DWI. Data are presented as means\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviations. A P value of \u0026lt;\u0026thinsp;0.05 was considered to indicate a statistically significant difference. (ADC: apparent diffusion coefficient, C-DWI: conventional diffusion-weighted imaging, CNR: contrast-to-noise ratio, IQ: image quality, P-DWI: precise diffusion-weighted imaging, SNR: signal-to-noise ratio)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eReader 1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c7\" namest=\"c5\"\u003e \u003cp\u003eReader 2\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eC-DWI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eP-DWI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ep-value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eC-DWI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eP-DWI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003ep-value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eQualitative Data\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"6\" nameend=\"c7\" namest=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOverall Image quality\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.08\u0026thinsp;\u0026plusmn;\u0026thinsp;0.92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.25\u0026thinsp;\u0026plusmn;\u0026thinsp;0.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.389\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.82\u0026thinsp;\u0026plusmn;\u0026thinsp;0.72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3.00\u0026thinsp;\u0026plusmn;\u0026thinsp;0.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.328\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIQ at dome\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.33\u0026thinsp;\u0026plusmn;\u0026thinsp;1.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.33\u0026thinsp;\u0026plusmn;\u0026thinsp;1.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.002\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3.00\u0026thinsp;\u0026plusmn;\u0026thinsp;0.78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3.18\u0026thinsp;\u0026plusmn;\u0026thinsp;0.78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.320\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSharpness\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.78\u0026thinsp;\u0026plusmn;\u0026thinsp;1.17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2.98\u0026thinsp;\u0026plusmn;\u0026thinsp;1.14\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.441\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3.10\u0026thinsp;\u0026plusmn;\u0026thinsp;0.78\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3.30\u0026thinsp;\u0026plusmn;\u0026thinsp;0.65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.215\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eArtifact\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.28\u0026thinsp;\u0026plusmn;\u0026thinsp;1.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.33\u0026thinsp;\u0026plusmn;\u0026thinsp;1.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.836\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3.08\u0026thinsp;\u0026plusmn;\u0026thinsp;0.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3.18\u0026thinsp;\u0026plusmn;\u0026thinsp;0.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.563\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eQuantitative Data\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSNR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e48.65\u0026thinsp;\u0026plusmn;\u0026thinsp;34.47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e47.25\u0026thinsp;\u0026plusmn;\u0026thinsp;31.28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.842\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e42.16\u0026thinsp;\u0026plusmn;\u0026thinsp;23.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e39.84\u0026thinsp;\u0026plusmn;\u0026thinsp;19.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.628\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCNR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e97.09\u0026thinsp;\u0026plusmn;\u0026thinsp;99.52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e93.75\u0026thinsp;\u0026plusmn;\u0026thinsp;69.32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.852\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e63.98\u0026thinsp;\u0026plusmn;\u0026thinsp;49.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e60.45\u0026thinsp;\u0026plusmn;\u0026thinsp;40.27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.724\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eADC value 10\u003csup\u003e\u0026minus;\u0026thinsp;3\u003c/sup\u003emm\u003csup\u003e2\u003c/sup\u003e/s\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.29\u0026thinsp;\u0026plusmn;\u0026thinsp;0.25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.15\u0026thinsp;\u0026plusmn;\u0026thinsp;0.25\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.013\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.32\u0026thinsp;\u0026plusmn;\u0026thinsp;0.28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.21\u0026thinsp;\u0026plusmn;\u0026thinsp;0.30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e0.036\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eQuantitative Image Quality Assessment\u003c/p\u003e \u003cp\u003eThere were no significant differences in SNR or CNR between C-DWI and P-DWI for either reader (all p\u0026thinsp;\u0026gt;\u0026thinsp;0.05) (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). For Reader 1, SNR was 48.65\u0026thinsp;\u0026plusmn;\u0026thinsp;34.47 for C-DWI and 47.25\u0026thinsp;\u0026plusmn;\u0026thinsp;31.28 for P-DWI (p\u0026thinsp;=\u0026thinsp;0.842). For Reader 2, SNR was 42.16\u0026thinsp;\u0026plusmn;\u0026thinsp;23.10 and 39.84\u0026thinsp;\u0026plusmn;\u0026thinsp;19.60, respectively (p\u0026thinsp;=\u0026thinsp;0.628). CNR values were similarly comparable between sequences for Reader 1 (97.09\u0026thinsp;\u0026plusmn;\u0026thinsp;99.52 vs 93.75\u0026thinsp;\u0026plusmn;\u0026thinsp;69.32; p\u0026thinsp;=\u0026thinsp;0.852) and Reader 2 (63.98\u0026thinsp;\u0026plusmn;\u0026thinsp;49.75 vs 60.45\u0026thinsp;\u0026plusmn;\u0026thinsp;40.27; p\u0026thinsp;=\u0026thinsp;0.724).\u003c/p\u003e \u003cp\u003eADC values were significantly lower on P-DWI than on C-DWI for both Reader 1 (1.15\u0026thinsp;\u0026plusmn;\u0026thinsp;0.25 vs 1.29\u0026thinsp;\u0026plusmn;\u0026thinsp;0.25 \u0026times; 10⁻\u003csup\u003e3\u003c/sup\u003e mm\u003csup\u003e\u0026sup2;\u003c/sup\u003e/s; p\u0026thinsp;=\u0026thinsp;0.013) and Reader 2 (1.21\u0026thinsp;\u0026plusmn;\u0026thinsp;0.30 vs 1.32\u0026thinsp;\u0026plusmn;\u0026thinsp;0.28 \u0026times; 10⁻\u003csup\u003e3\u003c/sup\u003e mm\u0026sup2;/s; p\u0026thinsp;=\u0026thinsp;0.036).\u003c/p\u003e \u003cp\u003eHCC Diagnostic Performance Analysis\u003c/p\u003e \u003cp\u003eSensitivity was highest for Dynamic MRI\u0026thinsp;+\u0026thinsp;P-DWI (98.5%, 135/137; 95% CI, 92.0\u0026ndash;99.7%), followed by Dynamic MRI\u0026thinsp;+\u0026thinsp;C-DWI (92.7%, 127/137; 95% CI, 83.7\u0026ndash;97.8%), and lowest for Dynamic MRI alone (65.7%, 90/137; 95% CI, 52.7\u0026ndash;75.9%) (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Both combined image sets demonstrated significantly higher sensitivity than Dynamic MRI alone (both p\u0026thinsp;\u0026lt;\u0026thinsp;0.001, McNemar test). Although sensitivity was numerically higher for P-DWI than for C-DWI, the difference was not statistically significant (p\u0026thinsp;=\u0026thinsp;0.125). Figures\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e3\u003c/span\u003e and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e4\u003c/span\u003e are representative examples of HCC lesions detected exclusively on P-DWI, as well as those visualized on both P-DWI and C-DWI.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDiagnostic performance for the detection of hepatocellular carcinoma. Sensitivity was calculated for the detection of 137 HCC lesions. P values were determined using McNemar\u0026rsquo;s test for pairwise comparisons of diagnostic sensitivity among the image sets. (C-DWI: conventional diffusion-weighted imaging, CI: confidence interval, P-DWI: precise diffusion-weighted imaging)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eImage Set\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSensitivity(%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95% CI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ep-value \u003c/p\u003e \u003cp\u003e(vs. Dynamic MRI alone)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ep-value \u003c/p\u003e \u003cp\u003e(vs. Dynamic MRI\u0026thinsp;+\u0026thinsp;C-DWI)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDynamic MRI alone\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e65.7 (90/137)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e52.7\u0026ndash;75.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDynamic MRI\u0026thinsp;+\u0026thinsp;C-DWI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e92.7 (127/137)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e83.7\u0026ndash;97.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDynamic MRI\u0026thinsp;+\u0026thinsp;P-DWI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e98.5 (135/137)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e92.0\u0026ndash;99.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.125\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eHCC lesions missed on Dynamic MRI alone were significantly smaller than detected lesions (13.42\u0026thinsp;\u0026plusmn;\u0026thinsp;22.19 mm vs 30.48\u0026thinsp;\u0026plusmn;\u0026thinsp;26.52 mm; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Similarly, lesions missed on C-DWI were smaller than detected lesions (9.00\u0026thinsp;\u0026plusmn;\u0026thinsp;2.79 mm vs 24.18\u0026thinsp;\u0026plusmn;\u0026thinsp;26.34 mm; p\u0026thinsp;=\u0026thinsp;0.039). ROC analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e2\u003c/span\u003e) demonstrated that tumor size predicted detectability on C-DWI with an optimal cutoff value of 1.4 cm (AUC\u0026thinsp;=\u0026thinsp;0.777; p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). For lesions\u0026thinsp;\u0026ge;\u0026thinsp;1.4 cm, detection rates were comparable among all image sets.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eWe evaluated whether liver DWI reconstructed with SmartSpeed Precise improves image quality and HCC detection performance compared with conventional DWI. Our results showed that P-DWI provided numerically higher qualitative image quality while maintaining comparable SNR and CNR relative to C-DWI, without additional acquisition time. ADC values were significantly lower on P-DWI than on C-DWI for both readers. In addition, overall sensitivity for HCC detection was numerically higher with P-DWI. These findings suggest that SmartSpeed Precise reconstruction may offer incremental benefits, particularly for small HCCs over conventional liver DWI, particularly for small HCCs.\u003c/p\u003e \u003cp\u003eAlthough P-DWI demonstrated numerically higher sensitivity, the difference did not reach statistical significance, likely reflecting the already high baseline performance of conventional DWI. This finding is consistent with prior studies comparing deep learning (DL)-accelerated liver MRI with conventional techniques. Park et al. [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] reported that a DL-accelerated abbreviated MRI protocol achieved comparable diagnostic sensitivity to a standard protocol, with no significant difference in HCC detection (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05). Similarly, Kim et al. [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] found no significant difference in per-lesion sensitivity between DL-DWI and conventional DWI (p\u0026thinsp;\u0026ge;\u0026thinsp;0.062). Collectively, these studies suggest that DL-based reconstruction techniques can maintain, and in some cases modestly improve, the already high diagnostic performance of conventional DWI without additional scan time.\u003c/p\u003e \u003cp\u003eNotably, P-DWI demonstrated a diagnostic advantage for lesions smaller than 1.4 cm. ROC analysis indicated that tumor size significantly influenced detectability on C-DWI, with an optimal cutoff of 1.4 cm. While detection rates were comparable for lesions larger than 1.4 cm, P-DWI combined with dynamic MRI identified numerically more sub-1.4 cm lesions than C-DWI. This improvement is likely related to the combined effects of noise reduction, artifact suppression, and enhanced spatial resolution provided by the dual convolutional neural network architecture. The Adaptive-CS-Net reduces background noise, whereas the Precise Image Net suppresses ringing artifacts and sharpens lesion margins. Although not statistically significant, qualitative assessments showed a trend toward improved sharpness and fewer artifacts with P-DWI. These findings are in line with Zhao et al. [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e], who demonstrated improved lesion conspicuity and diagnostic sensitivity using a similar DL-based framework. Comparable benefits have also been reported in breast, prostate, and musculoskeletal MRI. Thus, while conventional DWI remains adequate for larger tumors, P-DWI may provide added value in detecting small HCCs, which are often more challenging to identify.\u003c/p\u003e \u003cp\u003eWe observed significantly lower ADC values on P-DWI compared with C-DWI. This phenomenon has been reported in prior studies using DL-based reconstruction techniques [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e], although it\u0026rsquo;s underlying mechanism remains incompletely understood. One plausible explanation is related to noise suppression at high b-values. In conventional DWI, low SNR at high b-values results in a noise floor effect, which can artificially elevate signal intensity. By suppressing background noise, the DL algorithm may reduce this artificial signal elevation, leading to a greater apparent signal decay between b-values and consequently lower ADC measurements. This explanation remains hypothetical and warrants further investigation.\u003c/p\u003e \u003cp\u003eThis study has several limitations. First, its retrospective single-center design introduces potential selection bias and limits generalizability. Second, the DL reconstruction algorithm is vendor-specific; therefore, the results may not be directly applicable to other platforms with different network architectures or training datasets. Third, quantitative analysis of SNR and CNR was limited to a b-value of 800 s/mm\u0026sup2;. The performance of the DL algorithm at higher b-values (e.g., \u0026ge;\u0026thinsp;1000 s/mm\u0026sup2;), where SNR is further degraded, was not evaluated and may differ.\u003c/p\u003e \u003cp\u003eIn conclusion, SmartSpeed Precise DWI provides comparable overall diagnostic performance to conventional DWI while offering potential advantages in image quality and sensitivity for small HCCs, without increasing scan time. The observed reduction in ADC values likely reflects the effects of DL-based noise suppression, although further studies are needed to clarify the underlying mechanism.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eJY.J and HJ.K wrote the main manuscript text and AR. J and GB.L analyzed the data. All authors reviewed the manuscript.\u003c/p\u003e "},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eWild CP, Weiderpass E, Stewart BWE. IARC World Cancer Reports. In. World Cancer Report: Cancer research for cancer prevention. Lyon (FR): I, 2020\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024; 74:229\u0026ndash;263\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRumgay H, Arnold M, Ferlay J, Lesi O, Cabasag CJ, Vignat J, et al. Global burden of primary liver cancer in 2020 and predictions to 2040. Journal of Hepatology 2022; 77:1598\u0026ndash;1606\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKoh DM, Collins DJ. Diffusion-weighted MRI in the body: applications and challenges in oncology. AJR Am J Roentgenol 2007; 188:1622\u0026ndash;1635\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGalea N, Cantisani V, Taouli B. Liver lesion detection and characterization: role of diffusion-weighted imaging. J Magn Reson Imaging 2013; 37:1260\u0026ndash;1276\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTaouli B, Koh DM. Diffusion-weighted MR imaging of the liver. Radiology 2010; 254:47\u0026ndash;66\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShenoy-Bhangle A, Baliyan V, Kordbacheh H, Guimaraes AR, Kambadakone A. Diffusion weighted magnetic resonance imaging of liver: Principles, clinical applications and recent updates. World J Hepatol 2017; 9:1081\u0026ndash;1091\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGholipour A, Kehtarnavaz N, Scherrer B, Warfield SK. On the accuracy of unwarping techniques for the correction of susceptibility-induced geometric distortion in magnetic resonance Echo-planar images. Annu Int Conf IEEE Eng Med Biol Soc 2011; 2011:6997\u0026ndash;7000\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKondo F, Wada K, Kondo Y. Morphometric analysis of hepatocellular carcinoma. Virchows Arch A Pathol Anat Histopathol 1988; 413:425\u0026ndash;430\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePhilipsHealthcare. SmartSpeed Precise: Ultimate precision at maximum speed with Dual Al - MR Clinical Application [Brochure] (accessed on 28 December 2025). In: Philips SmartSpeed Precise, 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePezzotti N, de Weerdt E, Yousefim S, Elmahdy MS, van Gemert J, Sch\u0026uuml;lke C, et al. Adaptive-CS-Net: FastMRI with Adaptive Intelligence. arXiv 2019\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePeeters H, Chung H, Valvano G, Yakisikli D, Van Gemert J, De Weerdt E, et al. Philips SmartSpeed. No compromise Image quality and speed at your fingertips [White paper] (accessed on 28 December 2025). In: Philips SmartSpeed, 2021\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMesropyan N, Katemann C, Leutner C, Sommer A, Isaak A, Weber OM, et al. Accelerated High-resolution T1- and T2-weighted Breast MRI with Deep Learning Super-resolution Reconstruction. Acad Radiol 2025; 32:3147\u0026ndash;3156\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBischoff LM, Peeters JM, Weinhold L, Krausewitz P, Ellinger J, Katemann C, et al. Deep Learning Super-Resolution Reconstruction for Fast and Motion-Robust T2-weighted Prostate MRI. Radiology 2023;308: e230427\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJurka M, Macova I, Wagnerova M, Capoun O, Jakubicek R, Ourednicek P, et al. Deep-learning-based reconstruction of T2-weighted magnetic resonance imaging of the prostate accelerated by compressed sensing provides improved image quality at half the acquisition time. Quant Imaging Med Surg 2024; 14:3534\u0026ndash;3543\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTerzis R, Dratsch T, Hahnfeldt R, Basten L, Rauen P, Sonnabend K, et al. Five-minute knee MRI: An AI-based super resolution reconstruction approach for compressed sensing. A validation study on healthy volunteers. Eur J Radiol 2024; 175:111418\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKravchenko D, Isaak A, Mesropyan N, Peeters JM, Kuetting D, Pieper CC, et al. Deep learning super-resolution reconstruction for fast and high-quality cine cardiovascular magnetic resonance. Eur Radiol 2025; 35:2877\u0026ndash;2887\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRadiology ACo. Li-RADS CT/MRI version 2018 Core. https:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003eedge.sitecorecloud.io/americancoldf5f-acrorgf92a-productioncb02-3650/media/ACR/Files/RADS/LI-RADS/LI-RADS-CT-MRI-2018-Core.pdf\u003c/span\u003e\u003cspan address=\"http://edge.sitecorecloud.io/americancoldf5f-acrorgf92a-productioncb02-3650/media/ACR/Files/RADS/LI-RADS/LI-RADS-CT-MRI-2018-Core.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (accessed 26 Feb 2026) 2018\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark SH, Choi MH, Kim B, Lee HS, Yoon S, Lee YJ, et al. Deep Learning-Accelerated Non-Contrast Abbreviated Liver MRI for Detecting Malignant Focal Hepatic Lesions: Dual-Center Validation. Korean J Radiol 2025;26:333\u0026ndash;345\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim DH, Kim B, Lee HS, Benkert T, Kim H, Choi JI, et al. Deep Learning-Accelerated Liver Diffusion-Weighted Imaging: Intraindividual Comparison and Additional Phantom Study of Free-Breathing and Respiratory-Triggering Acquisitions. Invest Radiol 2023;58:782\u0026ndash;790\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao D, Kong X, Yang K, Wan J, Liu Z, Pan F, et al. Deep learning-enhanced super-resolution diffusion-weighted liver MRI: improved image quality, diagnostic performance, and acceleration. Insights Imaging 2025;16:273\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBae SH, Hwang J, Hong SS, Lee EJ, Jeong J, Benkert T, et al. Clinical feasibility of accelerated diffusion weighted imaging of the abdomen with deep learning reconstruction: Comparison with conventional diffusion weighted imaging. Eur J Radiol 2022;154:110428\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDietrich O, Heiland S, Sartor K. Noise correction for the exact determination of apparent diffusion coefficients at low SNR. Magn Reson Med 2001;45:448\u0026ndash;453\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-9442562/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9442562/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003ePurpose\u003c/h2\u003e \u003cp\u003eTo evaluate the image quality and diagnostic performance of SmartSpeed Precise diffusion-weighted imaging (DWI), a deep learning\u0026ndash;based denoising and super-resolution reconstruction technique, compared with conventional DWI for the detection of hepatocellular carcinoma (HCC).\u003c/p\u003e\u003ch2\u003eMaterials and Methods\u003c/h2\u003e \u003cp\u003eThis retrospective analysis included 84 consecutive patients (mean age, 66.8\u0026thinsp;\u0026plusmn;\u0026thinsp;9.1 years; 65 men) with liver cirrhosis who underwent gadoxetic acid\u0026ndash;enhanced MRI including both conventional DWI (C-DWI) and SmartSpeed Precise DWI (P-DWI) between September 2025 and January 2026. A total of 137 HCC lesions were evaluated. Two radiologists independently assessed qualitative image quality and measured signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and apparent diffusion coefficient (ADC) values. Diagnostic performance for HCC detection was compared among three image sets: dynamic MRI alone, dynamic MRI plus C-DWI, and dynamic MRI plus P-DWI. Sensitivity was compared using McNemar testing, and receiver operating characteristic (ROC) analysis was performed to identify the optimal tumor size cut-off for detectability. A p value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eAcquisition time was identical for C-DWI and P-DWI, as both were reconstructed from the same raw dataset. Qualitative image quality, SNR, and CNR did not differ significantly between techniques (all p\u0026thinsp;\u0026gt;\u0026thinsp;0.05). However, ADC values were significantly lower with P-DWI than with C-DWI for both readers (Reader 1: 1.15\u0026thinsp;\u0026plusmn;\u0026thinsp;0.25 vs 1.29\u0026thinsp;\u0026plusmn;\u0026thinsp;0.25 \u0026times; 10⁻\u003csup\u003e3\u003c/sup\u003e mm\u0026sup2;/s, p\u0026thinsp;=\u0026thinsp;0.013; Reader 2: 1.21\u0026thinsp;\u0026plusmn;\u0026thinsp;0.30 vs 1.32\u0026thinsp;\u0026plusmn;\u0026thinsp;0.28 \u0026times; 10⁻\u003csup\u003e3\u003c/sup\u003e mm\u0026sup2;/s, p\u0026thinsp;=\u0026thinsp;0.036). Sensitivity for HCC detection was 65.7% (90/137) for dynamic MRI alone, 92.7% (127/137) for dynamic MRI plus C-DWI, and 98.5% (135/137) for dynamic MRI plus P-DWI. Both DWI-containing image sets showed significantly higher sensitivity than dynamic MRI alone (both p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). However, the difference between P-DWI and C-DWI was not statistically significant (p\u0026thinsp;=\u0026thinsp;0.125). ROC analysis revealed that tumor size\u0026thinsp;\u0026le;\u0026thinsp;1.4 cm was a significant predictor of missed lesions on C-DWI (AUC\u0026thinsp;=\u0026thinsp;0.777, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), where P-DWI demonstrated a numerical diagnostic advantage in detecting these small HCCs.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eDeep learning\u0026ndash;based SmartSpeed Precise DWI provided image quality and HCC detection performance comparable to those of conventional DWI without additional acquisition time. It showed lower ADC values with AI-based reconstruction, and there was no difference in HCC diagnostic performance between the two techniques. These findings suggest that SmartSpeed Precise DWI may serve as a feasible alternative to conventional DWI for liver diffusion imaging.\u003c/p\u003e","manuscriptTitle":"Deep Learning–Based SmartSpeed Precise DWI for Hepatocellular Carcinoma: Comparison with Conventional DWI","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-04 05:44:43","doi":"10.21203/rs.3.rs-9442562/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"17be93e7-1475-4f1f-9f53-1ea80fd6cf49","owner":[],"postedDate":"May 4th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Rejected","date":"2026-05-06T12:46:01+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-04T10:53:24+00:00","index":19,"fulltext":""},{"type":"reviewerAgreed","content":"205433194831476375067268033683729606242","date":"2026-05-03T06:50:34+00:00","index":18,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-05-06T12:56:22+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-04 05:44:43","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9442562","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9442562","identity":"rs-9442562","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}