Motion-optimized diffusion MRI for assessment of liver fibrosis

Author

Lu Mao

Published

May 9, 2026

Statistical analysis

Categorical variables were summarized as counts and percentages; continuous variables are reported as median (interquartile range). Associations between each ADC sequence (MONO, MODI, MODI LRES) and fibrosis stage were assessed using Spearman rank correlation, with differences across fibrosis stages evaluated by the Kruskal-Wallis test followed by pairwise Mann-Whitney U tests with Bonferroni correction. Pearson correlation was used to quantify the relationships between ADC values and 2D MRE liver stiffness and between ADC values and collagen proportionate area (CPA).

Diagnostic performance of each ADC sequence and 2D MRE stiffness for three binary fibrosis thresholds — significant fibrosis (F≥2 vs F0–1), advanced fibrosis (F≥3 vs F0–2), and cirrhosis (F4 vs F0–3) — was evaluated by receiver operating characteristic (ROC) analysis; area under the ROC curve (AUC) with 95% DeLong confidence intervals was computed for each predictor. Three clinically meaningful operating points were identified per curve: a Youden-based cutoff maximizing the sum of sensitivity and specificity, a rule-in cutoff targeting ≥90% specificity, and a rule-out cutoff targeting ≥90% sensitivity.

To assess the incremental diagnostic value of ADC for significant fibrosis, BIC-based stepwise logistic regression (forward and backward selection, penalty k = log n) was applied separately for each ADC sequence using candidate pools with and without ADC alongside 2D MRE stiffness, PDFF, age, sex, BMI, ALT, and AST; AUCs derived from model-fitted probabilities were compared pairwise using DeLong’s test for dependent ROC curves. All analyses were performed in R (version 4.5.3).

Results

Patient characteristics

Table 1 summarizes the demographic and clinical characteristics of the 30 patients, stratified by sex.

Table 1: Patient characteristics by sex.
Characteristic F
N = 151
M
N = 151
Overall
N = 301
Age (years) 46.0 (37.0, 61.0) 49.0 (41.0, 61.0) 47.5 (40.0, 61.0)
BMI (kg/m²) 29.9 (26.8, 35.8) 30.5 (27.4, 34.1) 30.2 (27.4, 34.1)
Etiology of CLD


    Autoimmune, viral 2 (13%) 2 (13%) 4 (13%)
    MASLD/MASH/EtOH 11 (73%) 9 (60%) 20 (67%)
    Other (Cardiac Hepatopathy) 0 (0%) 1 (6.7%) 1 (3.3%)
    Other (Granulomatous Hepatitis) 1 (6.7%) 0 (0%) 1 (3.3%)
    Other (Normal Liver) 1 (6.7%) 0 (0%) 1 (3.3%)
    Other (Unknown) 0 (0%) 1 (6.7%) 1 (3.3%)
    PBC, PSC 0 (0%) 2 (13%) 2 (6.7%)
Fibrosis Stage


    0 2 (13%) 1 (6.7%) 3 (10%)
    1 3 (20%) 3 (20%) 6 (20%)
    2 6 (40%) 7 (47%) 13 (43%)
    3 3 (20%) 2 (13%) 5 (17%)
    4 1 (6.7%) 2 (13%) 3 (10%)
NAS Score


    0 4 (27%) 5 (33%) 9 (30%)
    1 1 (6.7%) 2 (13%) 3 (10%)
    3 4 (27%) 5 (33%) 9 (30%)
    4 4 (27%) 1 (6.7%) 5 (17%)
    5 1 (6.7%) 1 (6.7%) 2 (6.7%)
    6 1 (6.7%) 1 (6.7%) 2 (6.7%)
Pathology Steatosis Grade (%) 20.0 (0.0, 35.0) 15.0 (0.0, 25.0) 17.5 (0.0, 30.0)
ALT (U/L) 70.0 (41.0, 254.0) 102.0 (49.0, 199.0) 82.5 (48.0, 199.0)
AST (U/L) 75.0 (38.0, 195.0) 67.0 (45.0, 149.0) 71.0 (45.0, 149.0)
ALP (U/L) 95.0 (69.0, 122.0) 74.0 (60.0, 143.0) 83.5 (60.0, 127.0)
Total Bilirubin (mg/dL) 0.5 (0.4, 0.8) 0.7 (0.6, 0.9) 0.6 (0.4, 0.8)
INR


    0.84 0 (0%) 1 (6.7%) 1 (3.3%)
    0.9 1 (6.7%) 2 (13%) 3 (10%)
    1 8 (53%) 8 (53%) 16 (53%)
    1.1 5 (33%) 3 (20%) 8 (27%)
    1.2 0 (0%) 1 (6.7%) 1 (3.3%)
    1.3 1 (6.7%) 0 (0%) 1 (3.3%)
2D MRE Liver Stiffness (kPa) 3.0 (2.5, 3.7) 3.2 (2.9, 4.0) 3.1 (2.5, 3.7)
    Unknown 0 1 1
PDFF (%) 9.3 (2.3, 11.7) 7.4 (2.7, 11.0) 8.4 (2.5, 11.5)
    Unknown 0 2 2
MONO ADC (×10⁻⁶ mm²/s) 971.7 (941.8, 1,011.9) 1,100.1 (1,060.5, 1,234.9) 1,036.2 (950.7, 1,123.1)
MODI ADC (×10⁻⁶ mm²/s) 1,134.1 (1,046.3, 1,265.1) 1,135.5 (1,090.8, 1,253.3) 1,134.8 (1,072.7, 1,255.0)
MODI LRES ADC (×10⁻⁶ mm²/s) 1,196.1 (1,111.0, 1,306.0) 1,191.1 (1,119.4, 1,265.5) 1,191.8 (1,119.4, 1,265.5)
1 Median (Q1, Q3); n (%)

ADC vs fibrosis stage

Correlation of ADC (MONO, MODI, and MODI LRES) with Fibrosis Stage (Spearman correlation): Boxplot of ADC values by fibrosis stage with p-values from trend test.

Figure 1: Boxplot of ADC values by fibrosis stage with p-values from trend test.

ADC vs 2D MRE stiffness

Figure 2: Scatter plot of ADC vs 2D MRE stiffness with Spearman correlation coefficient and p-value.

ADC vs CPA

Figure 3: Scatter plot of ADC vs CPA with Spearman correlation coefficient and p-value.

ADC and MRE stiffness by fibrosis groups

Table 2: Median (IQR) of ADC and MRE stiffness by fibrosis groups with p-values from Wilcoxon rank-sum test.
Significant fibrosis
Measure F0–1 (n=9) F2–4 (n=21) p
MONO ADC 1065 (947–1123) 1012 (972–1102) p=0.928
MODI ADC 1231 (1143–1273) 1110 (1073–1243) p=0.189
MODI LRES ADC 1306 (1196–1355) 1178 (1111–1218) p=0.027
2D MRE (kPa) 3 (2–3) 3 (3–5) p=0.048
Advanced fibrosis
Measure F0–2 (n=22) F3–4 (n=8) p
MONO ADC 1036 (943–1118) 1036 (998–1101) p=0.590
MODI ADC 1220 (1077–1271) 1099 (1076–1116) p=0.079
MODI LRES ADC 1214 (1186–1319) 1114 (1098–1138) p=0.001
2D MRE (kPa) 3 (2–3) 5 (4–6) p<0.001
Cirrhosis
Measure F0–3 (n=27) F4 (n=3) p
MONO ADC 1012 (949–1101) 1206 (1104–1235) p=0.147
MODI ADC 1143 (1075–1260) 1091 (1062–1099) p=0.167
MODI LRES ADC 1196 (1163–1286) 1074 (1055–1102) p=0.032
2D MRE (kPa) 3 (2–4) 5 (4–6) p=0.058

ROC analysis - univariate

Univariate ROC analysis of ADC (MONO, MODI, MODI LRES) and 2D MRE as predictors of fibrosis stage. Multiple ROCs (F0–1 vs F2–4, F0–2 vs F3–4, F0–3 vs F4). With

  1. Youden-based cutoff that maximizes the sum of sensitivity and specificity;
  2. a rule-in cutoff (a cutoff providing 90% or highest possible specificity);
  3. a rule-out cutoff (a cutoff providing 90% or highest possible sensitivity).

F0-1 vs F2-4 (significant fibrosis)

Figure 4: ROC curves for significant fibrosis (F0–1 vs F2–4).
Table 3: ROC analysis for significant fibrosis (F0–1 vs F2–4).
Predictor AUC (95% CI)
Youden
Rule-in
Rule-out
Threshold Sn (%) Sp (%) Threshold Sn (%) Sp (%) Threshold Sn (%) Sp (%)
MONO ADC 0.51 (0.27–0.76) 964.6 76 44 1244.6 10 100 913.5 90 22
MODI ADC 0.66 (0.42–0.89) 1139.0 67 78 964.8 10 100 1324.8 95 22
MODI LRES 0.76 (0.54–0.98) 1285.8 90 56 1044.8 10 100 1285.8 90 56
2D MRE 0.74 (0.56–0.93) 3.5 48 100 3.5 48 100 2.2 100 25

F0-2 vs F3-4 (advanced fibrosis)

Figure 5: ROC curves for advanced fibrosis (F0–2 vs F3–4).
Table 4: ROC analysis for advanced fibrosis (F0–2 vs F3–4).
Predictor AUC (95% CI)
Youden
Rule-in
Rule-out
Threshold Sn (%) Sp (%) Threshold Sn (%) Sp (%) Threshold Sn (%) Sp (%)
MONO ADC 0.57 (0.35–0.79) 948.6 100 32 1236.5 12 91 948.6 100 32
MODI ADC 0.72 (0.53–0.90) 1139.0 100 64 964.8 0 91 1139.0 100 64
MODI LRES 0.89 (0.77–1.00) 1169.6 100 86 1109.4 50 91 1169.6 100 86
2D MRE 0.91 (0.77–1.00) 4.2 75 100 3.7 75 90 2.8 100 43

F0-3 vs F4 (cirrhosis)

Figure 6: ROC curves for cirrhosis (F0–3 vs F4).
Table 5: ROC analysis for cirrhosis (F0–3 vs F4).
Predictor AUC (95% CI)
Youden
Rule-in
Rule-out
Threshold Sn (%) Sp (%) Threshold Sn (%) Sp (%) Threshold Sn (%) Sp (%)
MONO ADC 0.77 (0.43–1.00) 1166.1 67 85 1236.5 33 93 1002.0 100 44
MODI ADC 0.75 (0.56–0.95) 1108.5 100 67 964.8 0 93 1108.5 100 67
MODI LRES 0.89 (0.74–1.00) 1145.6 100 78 1090.1 67 93 1145.6 100 78
2D MRE 0.85 (0.68–1.00) 3.5 100 73 5.5 33 92 3.5 100 73

Multivariate ROC analysis

To assess the incremental diagnostic value of ADC beyond established imaging and clinical biomarkers for significant fibrosis (F≥2), BIC-based stepwise logistic regression was applied to three candidate pools — one per ADC sequence — each including 2D MRE stiffness, PDFF, age, sex, BMI, ALT, and AST (n = 27 with complete data).

In all cases, BIC selection retained 2D MRE stiffness as the sole predictor, with neither ADC nor PDFF entering any model; the BIC-selected models with and without ADC were therefore identical. Among single-predictor models, MODI LRES achieved the highest AUC (0.73), comparable to 2D MRE alone (0.72), while no ADC sequence differed significantly from MRE by DeLong’s test. These results suggest that ADC does not provide incremental value over 2D MRE for predicting significant fibrosis in this cohort, though the small sample size limits definitive conclusions.

Table 6: Multivariate ROC analysis for significant fibrosis (F0–1 vs F2–4). BIC-based stepwise logistic regression with and without ADC; DeLong’s test comparisons.
Model Selected predictors AUC (95% CI)
MODI LRES
MODI LRES alone modi_lres 0.73 (0.49–0.97)
BIC model with ADC mre_2d 0.72 (0.52–0.93)
BIC model without ADC mre_2d 0.72 (0.52–0.93)
MODI ADC
MODI ADC alone modi_adc 0.64 (0.39–0.90)
BIC model with ADC mre_2d 0.72 (0.52–0.93)
BIC model without ADC mre_2d 0.72 (0.52–0.93)
MONO ADC
MONO ADC alone mono_adc 0.55 (0.29–0.80)
BIC model with ADC mre_2d 0.72 (0.52–0.93)
BIC model without ADC mre_2d 0.72 (0.52–0.93)