Apparent diffusion coefficient values in prostate cancer: Association with pathological grade groups and correlation with Ki-67 proliferation index

Patients

A single-center retrospective analysis was conducted on 119 patients with pathologically confirmed PCa who underwent prostate MRI at our hospital between January 2022 and June 2024. Inclusion criteria comprised: (I) Completion of preoperative prostate MRI with standardized imaging protocols within one month preceding surgery; (II) Histopathological confirmation of PCa with available immunohistochemical assessment of Ki-67 proliferation index. Exclusion criteria included: (I) Inadequate image quality for reliable ADC value quantification; (II) History of prior prostate-directed interventions (surgical procedures, radiation therapy, or systemic anticancer therapies) before MRI acquisition. The Institutional Review Board of The Central Hospital of Yongzhou granted an exemption from formal ethical review for this retrospective study.

Imaging techniques

MRI examinations of the seminal vesicles and prostate were performed using a 3.0T Siemens Skyra scanner with patients in the supine position. The imaging protocol included: (1) Axial T2WI: repetition time/echo time (TR/TE) 4800/101 ms; (2) Coronal and sagittal T2WI: TR/TE 5500/101 ms; (3) Axial T1-weighted imaging: TR/TE 500/9.7 ms; All structural sequences shared a 20 × 20 cm field of view (FOV) and 320 × 320 acquisition matrix. DWI parameters included TR/TE 5300/68 ms, FOV 20 × 20 cm, matrix 114 × 114, and b-values of 50, 1000, and 2000 s/mm². All sequences employed a 3.5 mm slice thickness with a 1 mm interslice gap.

Image analysis

Two board-certified radiologists specializing in genitourinary imaging (≥6 years’ experience) conducted independent analyses. Images were transferred to a dedicated postprocessing workstation for quantitative evaluation. Regions of interest (ROIs) measuring 15–25 mm² were placed within the lesions while carefully excluding adjacent normal tissue and cystic or necrotic areas to minimize partial-volume effects and ensure reproducibility. ROIs were strategically placed: (I) For ADC_min determination: Three measurements within the lesion’s lowest signal intensity area on ADC maps, selecting the minimal recorded value. (II) For ADC_mean calculation: Three ROI placements in tumoral solid components, averaged for final values. Consensus readings were derived from averaged measurements between both observers.

Imaging-pathology correlation method

Patients were stratified into two cohorts based on the origin of the pathological reference standard: a biopsy cohort (n = 76) and a radical prostatectomy cohort (n = 43). For patients in the biopsy cohort, all underwent MRI-targeted biopsy using cognitive fusion. The ROI for ADC measurement was consistently placed on the preoperative MRI at the precise location corresponding to the biopsy-proven target lesion. Two radiologists, referring to the biopsy procedure records, independently positioned the ROI on the ADC map at the slice exhibiting the largest cross-section of the targeted lesion. The definitive pathological grade and Ki-67 proliferation index were obtained from the histopathology report of the corresponding biopsy core. For the radical prostatectomy cohort, a retrospective back-mapping approach was adopted. The postoperative histopathology report – which detailed tumor location, extent, and description of the dominant lesion – was used to localize the corresponding lesion on the preoperative MRI. Two radiologists then placed the ROI on the MRI slice showing the largest cross-section of the histologically confirmed dominant lesion, meticulously aligning its boundaries with the documented pathological extent. The final pathological grade and Ki-67 index were derived from the pathology report of the most aggressive dominant lesion.

The dominant lesion was defined as: (I) The lesion demonstrating the most aggressive imaging features on MRI (i.e. the highest prostate imaging reporting and data system [PIRADS] score); or (II) in cases with multiple malignant lesions, the lesion with the highest pathological grade.

Histopathological evaluation

All specimens underwent standardized histopathological examination conducted by a board-certified pathologist specializing in uropathology (≥10 years’ experience). GS and subsequent GG classification were performed in accordance with the ISUP guidelines. The cohort was stratified as follows: GG 1: GS 6 (low-risk PCa); GG 2: GS 3 + 4 = 7 (intermediate-risk PCa); GG 3: GS 4 + 3 = 7 (intermediate-risk PCa); GG 4: GS 8 (high-risk PCa); GG 5: GS 9–10 (high-risk PCa).^[11] Clinically significant stratification categorized GG1-3 as non-high-risk PCa (nhPCa) and GG4-5 as high-risk PCa (hPCa).^[12] The Ki-67 score was evaluated under a 400x optical microscope. For each case, three representative hotspot regions (areas with the most intense staining) were selected, and the staining status of 1,000 tumor cells within these regions was manually counted. The number of positive cells/total cell count was defined as the Ki-67 proliferation index.

Statistical analysis

Statistical analyses were performed using Statistical Package for the Social Sciences 27.0 (IBM Corp.). Inter-reader agreement for ADC measurements between the two radiologists was assessed using the intraclass correlation coefficient (ICC), with an ICC >0.75 indicating excellent agreement. Data normality was assessed through Shapiro– Wilk testing. Non-normally distributed parameters were expressed as median (interquartile range) and compared using Mann–Whitney U tests. Normally distributed variables were presented as mean±standard deviation and analyzed with independent Student’s t-tests. This study explored the use of receiver operating characteristic (ROC) curve analysis to evaluate the diagnostic performance of ADC_mean and ADC_min in discriminating between hPCa and nhPCa. The optimal cutoff value on the ROC curve was determined by maximizing the Youden index. Spearman’s rank correlation coefficients quantified associations between: (I) ADC values and GG categories; (II) ADC values and Ki-67 proliferation indices. A two-tailed P < 0.05 defined statistical significance.

General information

A total of 119 patients were ultimately enrolled in this study. The cohort comprised 51 patients diagnosed with nhPCa group with an age range of 54–87 years, and 68 patients with hPCa group aged 48–91 years. Comparative analysis demonstrated no statistically significant difference in age distribution between the two cohorts (P = 0.330) [Table 1].

Comparison of ADC parameters and ROC curve analysis between nhPCa and hPCa groups

The inter-reader agreement for ADC_mean and ADC_min measurements between the two radiologists was excellent, with ICC values of 0.91 (95% confidence interval [CI]: 0.87~0.94) and 0.92 (95% CI: 0.88~0.94), respectively (P < 0.001). Both ADC_mean and ADC_min values were significantly lower in the hPCa group compared to the nhPCa group (P < 0.001), indicating more restricted diffusion on DWI in hPCa [Table 1 and Figure 1].

Close

Figure 1:

A 69-year-old male with prostate cancer (PCa). (a) T2-weighted imaging showed nodular hypointensity in the left transitional zone (white arrow); (b) Diffusion-weighted imaging shows high signal intensity for the lesion (white arrow); (c) Apparent diffusion coefficient shows low signal intensity for the lesion (white arrow); (d) Pathological findings shows non high-risk PCa (Hematoxylin and Eosin ×400); Gleason score 3 + 4 = 7, grading group 2, Ki-67 proliferation index 3% (black arrows).

Export to PPT

The ROC curve analysis demonstrated that both ADC_mean and ADC_min can effectively differentiate between nhPCa and hPCa. The area under the curves (AUCs) are 0.89 and 0.86, respectively. ADC_mean has a better discriminative performance compared to ADC_min [Table 2 and Figure 2].

Close

Figure 2:

The receiver operating characteristic curves of mean apparent diffusion coefficient and minimum apparent diffusion coefficient were used to distinguish non-high-risk prostate cancer from high-risk prostate cancer. ROC: Receiver operating characteristic; ADC: Apparent diffusion coefficient.

Export to PPT

Correlation analysis of ADC with ISUP GG and Ki-67

Both ADC_mean and ADC_min showed significant inverse correlations with ISUP GG [Spearman r = −0.73 (95% CI: −0.81 ~ −0.64), −0.66 (95% CI: −0.76 ~ −0.54), P < 0.001], showing a progressive decline in ADC values with increasing tumor grade [Figure 3].

Close

Figure 3:

Correlation analysis between mean apparent diffusion coefficient and the International Society of Urological Pathology grading group.

Export to PPT

The Ki-67 proliferation index demonstrated an inverse correlation with both ADC_mean and ADC_min values [Spearman r = −0.50 (95% CI: −0.63 ~ −0.35), −0.39 (95% CI: −0.57 ~ −0.27), P < 0.001], indicating a gradual decrease in ADC with increasing Ki-67 expression [Figure 4].

Close

Figure 4:

Correlation analysis between mean apparent diffusion coefficient and Ki-67 proliferation index.

Export to PPT

Comparative analysis of ADC values between nhPCa and hPCa groups

DWI assesses the mobility of water molecules within the tissue microenvironment, reflecting alterations in tumor cellularity and extracellular space.^[14] ADC is a quantitative parameter derived from DWI that inversely correlates with tumor cell density: as cellularity increases, ADC values decrease, and vice versa. Consequently, quantitative ADC measurements obtained from DWI are increasingly utilized in the early diagnosis, staging, treatment response evaluation, and prognosis assessment of PCa.^[15-17] Our findings revealed significantly lower ADC values in hPCa compared to nhPCa cohorts (P < 0.001), consistent with the histopathological correlation between reduced ADC metrics and elevated tumor cellularity. ROC analysis demonstrated superior diagnostic performance of ADC_mean and ADC_min in discriminating hPCa, with AUC values of 0.89 and 0.86, respectively. These observations align with established literature emphasizing the enhanced differential diagnostic capacity of ADC parameters for distinguishing nhPCa from hPCa.^[18] It should also be noted that the reported ADC threshold values are exploratory and specific to our scanner and acquisition protocol; thus, they require external validation before being generalized to other clinical settings.

ADC value applications in PCa risk stratification and Ki-67 correlation

The pathological grade and ISUP GG of PCa are closely associated with patient prognosis, with lower grades correlating with more favorable outcomes. Patients with hPCa generally have a poorer prognosis, for whom radical treatment forms the cornerstone of management, often combined with multimodal therapy. In contrast, patients with nhPCa exhibit better prognosis, and in some cases, the disease may remain manageable with active surveillance. Therefore, the treatment for nhPCa may focus on pharmacological therapy or active monitoring, thereby avoiding unnecessary surgical morbidity. Previous studies have demonstrated an inverse correlation between ADC values and ISUP GGs,^[19,20] supporting the role of ADC as a meaningful imaging marker in PCa risk stratification. Consistent with these reports, our findings also revealed a significant inverse correlation between both ADC_mean and ADC_min and ISUP GG. This suggests that as the pathological grade of prostate lesions increases, indicating higher malignancy and greater cellular density, the corresponding ADC values decrease.

The Ki-67 proliferative index serves as a well-established biomarker of cellular proliferation kinetics, reflecting tumor mitotic activity and serving as a prognostic indicator for tumor aggressiveness.^[8,21] Our analysis revealed significant inverse correlations between both ADC_mean and ADC_min with Ki-67 expression levels (r = −0.50, P < 0.001; r = −0.39, P < 0.001, respectively). This association may be explained by the more restricted diffusion microenvironment created by densely packed, rapidly proliferating tumor cells.

Therefore, the inverse correlations observed in this study between ADC values and both ISUP GG and the Ki-67 proliferation index are consistent with the underlying biology of increased cellular proliferation in PCa. These findings add to the evidence supporting a role for ADC in the noninvasive preoperative assessment of tumor aggressiveness. Furthermore, this correlation may provide a basis for future studies exploring the use of serial ADC measurements in monitoring disease course.

Clinical relevance and implications

While the PI-RADS framework has significantly improved the diagnostic performance of MRI, accurately distinguishing lesions requiring immediate intervention among equivocal PI-RADS category 3 or 4 lesions remains a challenge in clinical practice. Our study suggests that quantitative ADC measurements may provide an objective, numerical complement to the PI-RADS scoring system. Specifically, for indeterminate PI-RADS 3 or 4 lesions, a low ADC value may indicate a higher likelihood of upgrading and more aggressive pathological features, thereby aiding clinicians in making more precise biopsy decisions. In the realm of active surveillance, quantitative ADC holds potential as an imaging biomarker for monitoring disease progression. Furthermore, by demonstrating inverse correlation between ADC values and the Ki-67 proliferation index, our study provides a potential histopathological basis for the key PI-RADS feature of “diffusion restriction.” That is, lesions with higher PIRADS scores typically exhibit more pronounced reductions in ADC, which may correspond to increased cellular density and proliferative activity.

Compared to the PI-RADS assessment, which partly relies on subjective interpretation, ADC measurement offers reproducible quantitative data. Incorporating ADC values into the evaluation framework could help standardize reporting consistency across different institutions and readers. However, it must be unequivocally stated that no single imaging parameter, including ADC, can replace the histopathological gold standard for diagnosis. The ADC thresholds reported in this study are derived from a single-center, retrospective dataset, and their generalizability may be influenced by variations in scanner equipment, imaging protocols, and patient populations. Therefore, we posit that quantitative ADC analysis should serve as a complementary tool to the existing PI-RADS framework. By providing objective, reproducible numerical information alongside morphological assessment, it may collectively enhance the accuracy of PCa diagnosis and risk stratification. Future multicenter, prospective studies are warranted to validate these cutoff values.

This investigation has several methodological constraints that warrant consideration: (I) This study has a retrospective design and a relatively small sample size. The reported ADC thresholds were derived from a single-center dataset, and their generalizability requires validation in larger, multi-institutional cohorts; (II) The inclusion of both biopsy and radical prostatectomy specimens introduces potential heterogeneity in tumor grading, which may have an impact on the correlation analysis of pathological outcomes; (III) Partial volume effects could influence ADC quantification accuracy, particularly in lesions measuring <1 cm³, potentially confounding true diffusion restriction assessment.