Performance of simultaneous multislice diffusion-weighted imaging using monoexponential, intravoxel incoherent motion, and diffusion kurtosis models: assessment of microvascular invasion and histologic grade in hepatocellular carcinoma

Abstract

Objectives

This study aimed to evaluate the diagnostic performance of simultaneous multislice (SMS) acquisition combined with monoexponential, intravoxel incoherent motion (IVIM), and diffusion kurtosis imaging (DKI) models for predicting microvascular invasion (MVI) and histologic grade in hepatocellular carcinoma (HCC).

Materials and methods

A prospective study was conducted with 77 HCC patients. Diffusion-weighted imaging (DWI), IVIM, and DKI were performed on a 3T MRI using both SMS and conventional sequences. The values of diffusion parameters (ADC, D, D*, f, MD, and MK) were compared among SMS and conventional sequences, between MVI-positive and MVI-negative groups, and between high-grade and low-grade HCC groups. Receiver operating characteristic (ROC) curves were used to assess the diagnostic performance of diffusion parameters in predicting MVI and histologic grade. Inter-reader consistency was evaluated using intraclass correlation coefficients (ICC).

Results

Among the 77 patients, 29.9% were MVI-positive and 35.1% had high-grade HCC. SMS reduced scanning time by up to 44.44%. Most diffusion parameters were similar between SMS and conventional sequences, except for slightly lower ADC and f in SMS. MVI-positive and high-grade HCC cases showed lower ADC, D, D*, and MD values and higher MK values. The ICC ranged from 0.702 to 0.879. SMS-MK demonstrated the highest diagnostic performance with an AUC of 0.92 for MVI and 0.86 for histologic grade.

Conclusions

SMS acquisition, integrated with IVIM and DKI, is a feasible imaging method for preoperative evaluation of MVI and histologic grade in HCC, offering a faster alternative to conventional methods without compromising diagnostic performance.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40644-025-00930-9.

Introduction

Hepatocellular carcinoma (HCC) presents a significant clinical challenge due to poor prognosis, with the five-year recurrence rate after resection reaching up to 70% [1]. Microvascular invasion (MVI) and histologic grade are key determinants of HCC recurrence and influence treatment strategies [2, 3, 4, 5, 6, 7, 8]. Diffusion imaging techniques offer potential preoperative tools, including standard diffusion-weighted imaging (DWI), intravoxel incoherent motion (IVIM), and diffusion kurtosis imaging (DKI) [9, 10, 11, 12, 13]. Key parameters derived from these models, such as the apparent diffusion coefficient (ADC) [14] reflecting cellularity, the true diffusion coefficient (D) [11, 15, 16] related to tissue diffusion separate from microcapillary perfusion, and mean kurtosis (MK) [9, 10] quantifying tissue heterogeneity, have shown value in assessing MVI and histologic grade.

However, despite this potential, a major limitation to the broad clinical adoption of advanced diffusion models like IVIM and DKI is the prolonged scan time associated with acquiring multiple b-values (often 4–5 min). This impacts workflow and increases motion sensitivity [10]. Simultaneous multislice (SMS) imaging is an effective acceleration technique capable of significantly reducing scan times [17, 18, 19]. Its application in liver imaging has been explored, including studies focusing on measurement repeatability and image quality [19, 20, 21, 22].

Previous studies indicate SMS accelerates (specifically with an acceleration factor of 2) liver scans without compromising image quality or diffusion parameters (MK, mean diffusion (MD), ADC) in normal tissue [20], and yields comparable ADC values to conventional DWI in HCC, even post-contrast [23]. Critically, whether SMS acceleration impacts the reliability of diffusion models (monoexponential, IVIM, DKI) for predicting MVI and histologic grade in HCC remains unclear. This study aimed to assess the feasibility of SMS-accelerated diffusion MRI for predicting MVI and histologic grade by directly comparing its diagnostic performance, quantitative reliability, and scan time savings relative to conventional sequences.

Materials and methods

Study design and population

This prospective single-center study was approved by the local ethics committee (No. 2021[107]), and written informed consent was obtained from all participants. Between June 2021 and January 2023, participants meeting the following criteria were eligible for this study. Inclusion criteria: [1] Age ≥ 18 years; [2] Suspected HCC based on previous CT, ultrasonography, and other clinical evaluations; [3] Underwent preoperative MRI within 2 months prior to surgery. Exclusion criteria: [1] Contraindications to MRI; [2] Liver lesions confirmed to be non-HCC by postoperative pathology; [3] Preoperative antitumor therapy; [4] Incomplete pathological data; [5] Significant motion artifacts on MRI images. Preoperative data, including patient demographics, clinical history, and laboratory tests, were obtained from the hospital system. This study was performed and reported in adherence to the Preferred Reporting Items for STARD guidelines [24].

Imaging technique

Preoperative MRI was conducted on a 3T MAGNETOM Skyra scanner (Siemens Healthineers, Erlangen, Germany) with an 18-channel body coil. The protocol included conventional (CON) and SMS-accelerated sequences for DWI, IVIM, and DKI models. All sequences utilized single-shot echo-planar imaging (ss-EPI) under free-breathing conditions. Detailed parameters are shown in Table 1.

Table 1.

The detailed parameters for conventional and SMS-accelerated diffusion-weighted sequences

Parameters	CON-DWI	SMS-DWI	CON-IVIM	SMS-IVIM	CON-DKI	SMS-DKI
TR (ms)	3200	2000	4500	3000	5600	3000
TE (ms)	63	63	63	60	68	64
FOV (mm)	400 × 288	400 × 288	400 × 288	400 × 288	400 × 288	400 × 288
Matrix	128 × 94	128 × 94	128 × 94	128 × 94	128 × 94	128 × 94
Slice thickness (mm)	5	5	5	5	5	5
Slice gap (mm)	1.5	1.5	1.5	1.5	1.5	1.5
Bandwidth (Hz/pixel)	2442	2442	2442	2442	2442	2442
b values (s/mm²)	50,400,800	50,400,800	0,30,40,50,80,150,200,300,400,600,800,1000	0,30,40,50,80,150,200,300,400,600,800,1000	0,200,700,1400, 2100	0,200,700,1400, 2100
NEX	1,2,4	1,2,4	1,1,1,1,1,2,2,2,2,4,4,6	1,1,1,1,1,2,2,2,2,4,4,6	1,1,2,4,6	1,1,2,4,6
Accelerating factor	GRAPPA 2	GRAPPA 2 + SMS 2	GRAPPA 2	GRAPPA 2 + SMS 2	GRAPPA 2	GRAPPA 2 + SMS 2
Diffusion mode	3-scan trace	3-scan trace	3-scan trace	3-scan trace	3-scan trace	3-scan trace
Acquisition time	1:23	0:52	6:18	4:22	4:12	2:20

TR, repetition time; TE, echo time; FOV, field of view; NEX, number of excitations; CON, conventional; GRAPPA, generalized autocalibrating partially parallel acquisition; SMS, simultaneous multislice; DWI, diffusion-weighted imaging; IVIM, intravoxel incoherent motion; DKI, diffusion kurtosis imaging

Parameter calculation and analysis

All images were transferred to a syngo.via Frontiers Workstation, and subsequent analysis was performed using a prototype MR Body Diffusion toolbox (v1.4.0). The software generated the following diffusion parameters:

The ADC was calculated from DWI sequences using a monoexponential model with b-values of 50 and 800 s/mm², as follows:

1

IVIM parameters D, pseudo-diffusion coefficient (D*), and perfusion fraction (f) were derived using a bi-exponential model with 12 b-values, as follows:

2

DKI parameters MK and MD were derived using a non-Gaussian model with 5 b-values, as follows:

3

Finally, diffusion parameter maps for ADC, D, D*, f, MD, and MK were generated based on the respective models. Three radiologists (Y.Y.W., T.Yang., Z.Y.) without access to clinical or pathological information, independently reviewed the images, performing regions of interest (ROI) analysis using free-hand ROIs on the slices with the largest tumor diameter, while avoiding areas of necrosis and hemorrhage. Quantitative parameters were extracted from each ROI, and the average values across the three readers were subsequently used for statistical analysis.

Histopathological analysis

Histopathological examination was performed on all resected specimens to determine MVI and histologic grades of HCC for radiologic-pathologic correlation. MVI was defined as tumor cells in endothelial-lined vascular spaces seen under microscopy [25]. HCC grade was classified by the Edmondson-Steiner system [26], with grades I-II as low-grade and III-IV as high-grade.

Statistical analysis

Normality was assessed using the Shapiro-Wilk test. The independent-sample t-test or Mann-Whitney U test was used to compare diffusion parameters across different MVI status groups and histologic grade groups, while the paired t-test or Wilcoxon signed-rank test was applied to compare diffusion parameters between conventional (CON) and SMS-accelerated sequences. Diagnostic accuracy was evaluated using receiver operating characteristic (ROC) curve analysis, with DeLong’s test used to compare the area under the curve (AUC). Inter-reader consistency for diffusion parameter values was assessed using the intraclass correlation coefficient (ICC). ICC values were interpreted as follows: 0.81 to 1.00 indicated excellent agreement, 0.61 to 0.80 good agreement, 0.41 to 0.60 moderate agreement, 0.21 to 0.40 fair agreement, and 0.00 to 0.20 poor agreement [27]. All statistical analyses were performed using MedCalc (version 20.1), Python (version 3.7.3), and R (version 4.4.1) software. A two-tailed P value < 0.05 was considered statistically significant.

Results

Patient characteristics and baseline data

A total of 77 patients were included in the study (Fig. 1), with a mean age of 54.3 ± 10.1 years; 65 patients were male (Table 2). 23 patients (29.9%) were MVI-positive HCC, and 27 patients (35.1%) were high-grade HCC. Most patients had hepatitis B virus infection as the primary etiology (66 cases, 85.7%). Among the patients, 58 (75.8%) showed elevated alpha-fetoprotein (AFP) levels (≥ 7 ng/mL), and 40 (51.9%) showed elevated PIVKA-II levels (≥ 32.5 mAU/mL). Cirrhosis was confirmed histopathologically in 32 patients (41.6%). The mean tumor size was 3.4 ± 1.7 cm, ranging from 1.0 to 12.7 cm (Table 2). Representative diffusion parameter maps are shown in Fig. 2.

Fig. 1.

Patient inclusion flowchart

Table 2.

Baseline characteristics of the study population

Characteristics	Values
Clinical information
Age (years)†	54.3 ± 10.1 (30–79)
Male/female	65/12
BMI (kg/m²)†	23.5 ± 0.32
Etiology of liver disease
Hepatitis B virus	66 (85.7)
Tumor size (cm)†	3.4 ± 1.7 (1.0-12.7)
AFP (≥ 7 ng/mL)	58 (75.3)
AST (≥ 40 IU/L)	25 (32.5)
ALT (≥ 50 IU/L)	26 (33.8)
PIVKA_II (≥ 32.5 mAU/mL)	40 (51.9)
Cirrhosis	32 (41.6)
BCLC stage
0	6 (7.8)
A	62 (80.5)
B	6 (7.8)
C	3 (3.9)
Child-Pugh class
A	70 (90.1)
B	7 (9.9)
Pathologic findings
MVI status
MVI negative	54 (70.1)
MVI positive	23 (29.9)
Edmondson-Steiner grade
Grade I	3 (3.9)
Grade II	47 (61.4)
Grade III	27 (35.1)

BMI, body mass index; AFP, alpha-fetoprotein; ALT, alanine aminotransferase; AST, aspartate aminotransaminase; PIVKA_II, protein-induced vitamin K absence or antagonist-II; BCLC stage, Barcelona Clinic Liver Cancer stage; MVI, microvascular invasion

Unless indicated otherwise, data represent the number of patients, with percentages in parentheses. ^†, data are expressed as mean ± standard deviation with ranges in parentheses

Fig. 2.

Conventional and SMS-accelerated diffusion-weighted images and quantitative diffusion parameter maps from a 52-year-old male patient with HCC, classified as Edmondson-Steiner grade II and negative for MVI. A 3.1 cm HCC lesion in the right liver lobe (yellow arrows) demonstrated hyperintensity on both the conventional DWI (CON-DWI) (a) and simultaneous multislice DWI (SMS-DWI) (b) images acquired with a b value of 50 s/mm². Quantitative maps derived from diffusion models are shown, including CON-ADC (c) and SMS-ADC (d); CON-D (e) and SMS-D (f); CON-MD (g) and SMS-MD (h); and CON-MK (i) and SMS-MK (j). SMS, simultaneous multislice; CON, conventional; DWI, diffusion-weighted imaging; DKI, diffusion kurtosis imaging; IVIM, intravoxel incoherent motion; MVI, microvascular invasion; ADC, apparent diffusion coefficient; D, true diffusion coefficient; D*, pseudo-diffusion coefficient; f, perfusion fraction; MK, mean kurtosis; MD, mean diffusivity

Comparison of diffusion parameters between SMS and conventional sequences

The comparison between SMS and conventional diffusion sequences revealed statistically significant differences in certain parameters (Table 3). The mean ADC values obtained using SMS were slightly lower than those from conventional sequences (1.13 ± 0.19 × 10⁻³ mm²/s vs. 1.17 ± 0.23 × 10⁻³ mm²/s, p < 0.001). Similarly, the mean f values from SMS were lower than those from conventional sequences (27.06 ± 3.99% vs. 30.23 ± 4.46%, p = 0.003). Other parameters, such as D (1.01 ± 0.15 × 10⁻³ mm²/s for SMS vs. 1.02 ± 0.15 × 10⁻³ mm²/s for conventional, p = 0.555), D* (24.93 ± 5.93 × 10⁻³ mm²/s for SMS vs. 24.69 ± 6.02 × 10⁻³ mm²/s for conventional, p = 0.231), MK (0.82 ± 0.13 for SMS vs. 0.81 ± 0.14 for conventional, p = 0.123), and MD (1.62 ± 0.25 × 10⁻³ mm²/s for SMS vs. 1.62 ± 0.24 × 10⁻³ mm²/s for conventional, p = 0.767) showed no statistically significant differences. ICC analysis showed strong agreement for most parameters between the two sequences, with ICC values ranging from 0.702 to 0.879 (Table 3).

Table 3.

Quantitative measurements and inter-reader ICC comparison between conventional and SMS sequences

Parameters	Quantitative measurement		P		ICC (95%CI) for 3 readers
	CON	SMS	P		CON	SMS
ADC ‡	1.17 ± 0.23	1.13 ± 0.19	< 0.001 **		0.879 (0.873–0.882)	0.877 (0.871–0.891)
D †	1.02 ± 0.15	1.01 ± 0.15	0.555		0.817 (0.792–0.832)	0.836 (0.803–0.854)
D* ‡	24.69 ± 6.02	24.93 ± 5.93	0.231		0.804 (0.779–0.837)	0.811 (0.782–0.843)
f ‡	30.23 ± 4.46	27.06 ± 3.99	0.003 **		0.719 (0.615–0.794)	0.702 (0.641–0.778)
MK †	0.81 ± 0.14	0.82 ± 0.13	0.123		0.822 (0.802–0.851)	0.821 (0.793–0.847)
MD ‡	1.62 ± 0.24	1.62 ± 0.25	0.767		0.815 (0.789–0.843)	0.823 (0.805–0.855)

ADC, D, MD, and D* are measured in × 10⁻³ mm²/s

f represents the perfusion fraction expressed as a percentage

^†, Paired t test; ^‡, Wilcoxon signed-rank test; ^*, P-value < 0.05; ^**, P-value < 0.01

CON, conventional; SMS, simultaneous multislice; ICC, intraclass correlation coefficient; CI, confidence interval; ADC, apparent diffusion coefficient; D, true diffusion coefficient; D*, perfusion-related diffusion; f, perfusion fraction; MK, mean kurtosis; MD, mean diffusivity

Comparison of diffusion parameters between MVI-Positive and MVI-negative groups

MVI-positive group showed lower ADC values (SMS: 1.04 vs. 1.17 × 10⁻³ mm²/s; conventional: 1.07 vs. 1.22 × 10⁻³ mm²/s), lower D values (SMS: 0.88 vs. 1.07 × 10⁻³ mm²/s; conventional: 0.89 vs. 1.07 × 10⁻³ mm²/s, all p < 0.001), lower D* values (SMS: 22.54 vs. 25.96 × 10⁻³ mm²/s, conventional: 22.47 vs. 25.64 × 10⁻³ mm²/s), lower MD values (SMS: 1.40 vs. 1.72 × 10⁻³ mm²/s, conventional: 1.40 vs. 1.71 × 10⁻³ mm²/s, all p < 0.001) and higher MK values (SMS: 0.95 vs. 0.76; conventional: 0.94 vs. 0.75, all p < 0.001) (Table 4; Fig. 3). No significant differences were observed in f values (p = 0.323 for SMS, p = 0.241 for conventional).

Table 4.

Correlations and comparisons of parameters by MVI and histologic groups

Parameters	MVI status†				Edmondson-Steiner grades†
	MVI negative (N = 54)	MVI positive (N = 23)	P		Low-grade (N = 50)	High-grade (N = 27)	P
ADC
CON-DWI‡ SMS-DWI‡	1.22 ± 0.24 1.17 ± 0.20	1.07 ± 0.18 1.04 ± 0.12	0.013 * < 0.001 **		1.23 ± 0.25 1.18 ± 0.21	1.08 ± 0.18 1.04 ± 0.11	0.006 ** < 0.001 **
D
CON-DWI† SMS-DWI†	1.07 ± 0.13 1.07 ± 0.13	0.89 ± 0.11 0.88 ± 0.11	< 0.001 ** < 0.001 **		1.07 ± 0.14 1.07 ± 0.13	0.92 ± 0.12 0.91 ± 0.12	< 0.001 ** < 0.001 **
D*
CON-DWI‡ SMS-DWI‡	25.64 ± 6.30 25.96 ± 6.23	22.47 ± 4.72 22.54 ± 4.39	0.033 * 0.019 *		26.10 ± 5.55 26.50 ± 5.32	22.24 ± 6.13 22.20 ± 6.03	0.006 ** 0.002 **
f
CON-DWI‡ SMS-DWI‡	29.83 ± 4.61 26.77 ± 4.21	31.14 ± 4.04 27.75 ± 3.40	0.241 0.323		30.18 ± 4.49 26.93 ± 3.94	30.30 ± 4.45 27.30 ± 4.12	0.912 0.696
MK
CON-DWI† SMS-DWI†	0.75 ± 0.11 0.76 ± 0.10	0.94 ± 0.09 0.95 ± 0.08	< 0.001 ** < 0.001 **		0.75 ± 0.12 0.76 ± 0.11	0.91 ± 0.11 0.91 ± 0.12	< 0.001 ** < 0.001 **
MD
CON-DWI‡ SMS-DWI‡	1.71 ± 0.21 1.72 ± 0.23	1.40 ± 0.13 1.40 ± 0.13	< 0.001 ** < 0.001 **		1.72 ± 0.20 1.72 ± 0.20	1.43 ± 0.17 1.44 ± 0.23	< 0.001 ** < 0.001 **

ADC, D, MD, and D* are measured in × 10⁻³ mm²/s. f represents the perfusion fraction expressed as a percentage. Tumors with Edmondson-Steiner grade I and II were classified into the low-grade group, and grade III was classified into the high-grade group. ^†, independent-sample t test; ^‡, Mann-Whitney U test; ^*, P-value < 0.05; ^**, P-value < 0.01. MVI, microvascular invasion; ADC, apparent diffusion coefficient; D, true diffusion coefficient; D*, perfusion-related diffusion; f, perfusion fraction; MK, mean kurtosis; MD, mean diffusivity

Fig. 3.

Boxplots display diffusion parameter differences. This figure shows the distribution of diffusion parameters (ADC, D, D*, f, MK, MD) for MVI and Grade groups. Blue and orange boxes represent SMS and CON measurements, respectively. Statistical significance (p < 0.05) is marked with * for SMS and # for CON, except for the parameter f, where no significant differences were observed. MVI, microvascular invasion; ADC, apparent diffusion coefficient; D, true diffusion coefficient; D*, perfusion-related diffusion; f, perfusion fraction; MK, mean kurtosis; MD, mean diffusivity

Comparison of diffusion parameters between high- and low-grade HCC

High-grade HCC showed lower ADC values (SMS: 1.04 vs. 1.18 × 10⁻³ mm²/s; conventional: 1.08 vs. 1.23 × 10⁻³ mm²/s), lower D values (SMS: 0.91 vs. 1.07 × 10⁻³ mm²/s; conventional: 0.92 vs. 1.07 × 10⁻³ mm²/s, all p < 0.001), lower D* values (SMS: 22.20 vs. 26.50 × 10⁻³ mm²/s; conventional: 22.24 vs. 26.10 × 10⁻³ mm²/s), lower MD values (SMS: 1.44 vs. 1.72 × 10⁻³ mm²/s, conventional: 1.43 vs. 1.72 × 10⁻³ mm²/s, all p < 0.001) and higher MK values (SMS: 0.91 vs. 0.76; conventional: 0.91 vs. 0.75, all p < 0.001) (Table 4; Fig. 3). No significant differences were observed in f values (p = 0.696 for SMS; p = 0.912 for conventional).

ROC analysis for predicting MVI status and histologic grade

The ROC analysis showed no statistically significant differences between corresponding parameters of SMS and conventional sequences for predicting MVI status and histologic grade (DeLong’s test, all p > 0.05) (Table 5; Fig. 4). For MVI prediction, SMS-MK had an AUC of 0.92 (95% CI: 0.84–0.97), with a sensitivity of 85.7% and a specificity of 83.3%. Other parameters with AUCs above 0.8 included MD (SMS: 0.88; CON: 0.88) and D (SMS: 0.87; CON: 0.86). ADC and D* showed lower AUCs, ranging from 0.64 to 0.74. For tumor histologic grade differentiation, SMS-MK had an AUC of 0.86 (95% CI: 0.76–0.93), with a sensitivity of 96.4% and a specificity of 67.4%. Other parameters with AUCs above 0.8 included MD (SMS: 0.86; CON: 0.86) and D (SMS: 0.82; CON: 0.82). ADC and D* showed lower AUCs, ranging from 0.71 to 0.75. Significant differences in AUC values were observed between SMS-MK and SMS-f, as well as between SMS-MD and SMS-f (both p < 0.0001). Similarly, CON-MK and CON-MD differed significantly from CON-f (p < 0.001), while no significant differences were found between parameters with higher AUC values, such as SMS-MK and SMS-MD (p > 0.05).

Table 5.

Diagnostic performance of parameters for MVI status and Edmondson-Steiner grade differentiation

Parameters	AUC (95% CI)	Optimal cutoff value	Sensitivity (95% CI)	Specificity (95% CI)
MVI-negative HCC group vs. MVI-positive HCC group
CON-ADC	0.74 (0.63–0.84)	≤ 1.17	96.7 (78.1–99.9)	55.6 (41.4–69.1)
SMS-ADC	0.71 (0.60–0.81)	≤ 1.09	87.0 (66.4–97.2)	56.7 (41.7–70.2)
CON-D	0.86 (0.76–0.93)	≤ 1.029	86.9 (65.5–95.3)	72.2 (58.4–83.5)
SMS-D	0.87 (0.77–0.94)	≤ 1.03	91.3 (72.0–98.9)	70.4 (56.4–82.0)
CON-D*	0.64 (0.53–0.75)	≤ 22.77	52.2 (30.6–73.2)	77.8 (64.4–88.0)
SMS-D*	0.68 (0.56–0.78)	≤ 21.76	47.8 (26.8–69.4)	83.3 (70.7–92.1)
CON-MK	0.91 (0.82–0.96)	> 0.87	82.6 (61.2–95.0)	85.2 (72.9–93.4)
SMS-MK	0.92 (0.84–0.97)	> 0.85	85.7 (78.1–99.9)	83.3 (70.7–92.1)
CON-MD	0.88 (0.78–0.94)	≤ 1.65	100.0 (85.2–100.0)	64.8 (50.6–77.3)
SMS-MD	0.88 (0.79–0.95)	≤ 1.51	87.0 (66.4–97.2)	79.6 (66.5–89.4)
Low-grade HCC group vs. High-grade HCC group
CON-ADC	0.74 (0.63–0.84)	≤ 1.09	78.6 (43.0–64.0)	82.1 (73.3–93.0)
SMS-ADC	0.74 (0.63–0.83)	≤ 1.03	67.8 (47.6–84.1)	77.6 (63.4–88.2)
CON-D	0.82 (0.71–0.90)	≤ 1.03	78.6 (59.0–91.7)	73.6 (61.4–86.1)
SMS-D	0.82 (0.72–0.90)	≤ 1.05	89.3 (71.8–97.7)	67.4 (52.5–80.1)
CON-D*	0.71 (0.59–0.81)	≤25.06	82.1 (63.1–93.9)	59.2 (44.2–73.0)
SMS-D*	0.75 (0.63–0.84)	≤24.93	82.6 (62.5–94.1)	63.3 (48.3–76.6)
CON-MK	0.85 (0.75–0.92)	> 0.85	82.5 (64.0–93.7)	83.7 (70.3–92.7)
SMS-MK	0.86 (0.76–0.93)	> 0.78	96.4 (81.7–99.9)	67.4 (52.5–80.1)
CON-MD	0.86 (0.76–0.93)	≤ 1.65	96.5 (82.3–99.8)	69.4 (54.6–81.7)
SMS-MD	0.86 (0.76–0.93)	≤ 1.63	96.3 (82.0–99.9)	69.5 (54.3–81.8)

MVI, microvascular invasion; CI, confidence interval; ADC, apparent diffusion coefficient; D, true diffusion coefficient; D*, perfusion-related diffusion; f, perfusion fraction; MK, mean kurtosis; MD, mean diffusivity

Fig. 4.

ROC curve analysis for predicting MVI status and histologic grade using diffusion parameters. ROC, receiver operating characteristic; MVI, microvascular invasion; SMS, simultaneous multislice; CON, conventional; ADC, apparent diffusion coefficient; D, true diffusion coefficient; D*, perfusion-related diffusion; f, perfusion fraction; MK, mean kurtosis; MD, mean diffusivity

Discussion

This study demonstrates that SMS technology, integrated with monoexponential, IVIM, and DKI models, is effective in assessing HCC aggressiveness. Compared to conventional sequences, SMS achieved similar diagnostic accuracy for histologic grading and MVI while reducing scan times by 37.35%, 44.44%, and 30.69% for DWI, IVIM, and DKI, respectively. These results support SMS as a feasible technique for preoperative evaluation of HCC.

SMS has been widely studied in liver imaging, primarily focusing on DWI and often involving volunteer studies [17, 18, 19, 20, 28, 29, 30]. By simultaneously exciting multiple slices, SMS significantly reduces scan time while maintaining image quality [31]. In our study, SMS enabled efficient acquisition, consistent with previous findings [19, 20]. This rapid acquisition suggests SMS could be effectively integrated into routine clinical workflows. However, our results revealed significant differences in measured ADC (p < 0.001) and f-values (p = 0.003) when using the SMS sequence compared to the conventional sequence. These differences likely arise from their dependence on signal decay at low b-values and the inherent challenges in accurately fitting the IVIM model, making them more susceptible to noise and residual motion in the free-breathing SMS environment. Additional factors like respiratory artifacts, relaxation time variations, and T2 contribution to the signal may also influence these results.

Beyond the technical comparison, our findings align with previous research on the biological relevance of diffusion models in HCC. Specifically, lower ADC, D, and MD values were significantly associated with higher histologic grades and MVI positivity, reflecting increased cellular density and reduced extracellular space [11]. Studies have shown significantly lower ADC values in poorly differentiated HCC, linked to early recurrence and poor prognosis [14]. In comparison, the IVIM-derived D values exhibited superior predictive efficacy for both histologic grade and MVI [11, 13, 15], as it eliminates microvascular perfusion effects and more accurately reflects true diffusion within tumor tissues. A meta-analysis by Wang et al. reported an AUC of 0.87–0.89 for D, surpassing that of ADC [13]. D has been identified as an independent predictor of HCC grade [16]. MK from the DKI model showed distinct advantages over both ADC and D by better revealing tumor heterogeneity and structural complexity [32]. Higher MK values were significantly correlated with high-grade HCC and MVI and served as an independent predictor of early recurrence [10, 33, 34]. Our team’s previous study also supports this, demonstrating SMS-DKI’s predictive value for early recurrence in HCC [35]. However, IVIM-derived parameters f and D*, reflecting microvascular perfusion and density, did not perform well in differentiating histologic grades and MVI status [13]. Additionally, f and D* measurements have shown poor repeatability [36, 37, 38, 39], and f-values show statistically significant differences between SMS and conventional sequences in the liver and kidney [40]. These findings suggest that f and D* are less reliable indicators of HCC pathological characteristics compared to D and MK. Furthermore, the observed differences in f-values between SMS and conventional sequences add complexity and may reduce the reliability of f as a quantitative biomarker, especially with the current SMS implementation.

This study has several limitations. First, the single-center design and relatively small sample size limit the generalizability of our findings to diverse patient populations and clinical settings beyond our institution. Second, the inclusion of only patients undergoing liver resection introduces selection bias. While strict control of imaging acquisition and histopathological confirmation ensured internal consistency, these measures cannot fully address the limitations of external generalizability. Third, the current lack of universally standardized protocols for advanced diffusion MRI techniques, especially SMS-accelerated sequences, presents a challenge. This limits direct quantitative comparison of our studies and hinders the broader clinical implementation of derived quantitative thresholds without site-specific validation. Furthermore, quantitative diffusion measurements in liver MRI are influenced by technical factors like image quality (e.g., noise, motion artifacts), tissue properties (e.g., fat and iron content), and the heterogeneity of HCC lesions. Lastly, the absence of an independent validation cohort limits the robustness of our diagnostic performance and requires external confirmation.

Future research should explore the integration of advanced techniques such as artificial intelligence (AI) and radiomics with SMS technology to improve HCC characterization and prognosis prediction [41, 42]. Larger, multicenter studies are needed to validate SMS-diffusion models across different clinical settings and MRI systems, supporting its broader implementation as a reliable imaging modality for HCC.

Conclusion

SMS acquisition combined with monoexponential, IVIM, and DKI offers a feasible and faster approach for preoperative assessment of HCC MVI and histological grade. It provides diagnostic accuracy comparable to conventional methods with significantly reduced scan times.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ADC

Apparent diffusion coefficient

AFP

Alpha-fetoprotein

BCLC

Barcelona clinic liver cancer

CAIPIRINHA

Controlled aliasing in parallel imaging results in higher acceleration

CI

Confidence interval

DKI

Diffusion kurtosis imaging

DWI

Diffusion-weighted imaging

D

True diffusion coefficient

D*

Pseudo-diffusion coefficient

GRAPPA

Generalized auto-calibrating partially parallel acquisition

HBV

Hepatitis B virus

HCC

Hepatocellular carcinoma

ICC

Intraclass correlation coefficient

IQR

Interquartile range

IVIM

Intravoxel incoherent motion

MD

Mean diffusivity

MK

Mean kurtosis

MVI

Microvascular invasion

PF

(f), perfusion fraction

ROI

Region of interest

SMS

Simultaneous multislice

ss-EPI

Single-shot echo-planar imaging

Author contributions

All authors contributed to the study conception and design. Bin Song and Zheng Ye are the supervisors of this study and the recipients of the funding that supported this research. Material preparation, data collection, and analysis were performed by Yingyi Wu, Zheng Qu, Zheng Ye, Ting Yang, Shan Yao, Jie Chen, Xinye Bao, and Ting Yin. The first draft of the manuscript was written by Yingyi Wu, and revision of the document was performed by Zheng Qu and Zheng Ye. All authors commented on previous versions of the manuscript and approved the final manuscript.

Funding

This study was supported by Natural Science Foundation of Sichuan Province (Grant number 2024NSFSC1798), National Natural Science Foundation of China (Grant number 82302161, U22A20343), China Postdoctoral Science Foundation (Grant number 2023M732464), Postdoctor Research Fund of West China Hospital, Sichuan University (Grant number 2024HXBH161), Development Project of Hainan Provincial Clinical Medical Center and Post-doctoral Station Development Project of Sanya (Grant number 23CZ009).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This study was performed in line with the principles of the Declaration of Helsinki. This prospective study was approved by the ethics committee of West China Hospital, Sichuan University (No. 2021[107]), and informed consent was obtained from all individual participants included in the study.

Consent for publication

All authors named in this manuscript consented to its publication and take full responsibility for its content. All authors read and approved the final manuscript.

Competing interests

The authors declare no competing interests.