Correlation of magnetic resonance imaging tumor volume with histopathology

Abstract

Purpose: The biology of prostate cancer may be influenced by the index lesion. The definition of index lesion volume is important for appropriate decision making, especially for image guided focal treatment. We determined the accuracy of magnetic resonance imaging for determining index tumor volume compared with volumes derived from histopathology.

Materials and methods: We evaluated 135 patients (mean age 59.3 years) with a mean prostate specific antigen of 6.74 ng/dl who underwent multiparametric 3T endorectal coil magnetic resonance imaging of the prostate and subsequent radical prostatectomy. Index tumor volume was determined prospectively and independently by magnetic resonance imaging and histopathology. The ellipsoid formula was applied to determine histopathology tumor volume, whereas manual tumor segmentation was used to determine magnetic resonance tumor volume. Histopathology tumor volume was correlated with age and prostate specific antigen whereas magnetic resonance tumor volume involved Pearson correlation and linear regression methods. In addition, the predictive power of magnetic resonance tumor volume, prostate specific antigen and age for estimating histopathology tumor volume (greater than 0.5 cm(3)) was assessed by ROC analysis. The same analysis was also conducted for the 1.15 shrinkage factor corrected histopathology data set.

Results: There was a positive correlation between histopathology tumor volume and magnetic resonance tumor volume (Pearson coefficient 0.633, p <0.0001), but a weak correlation between prostate specific antigen and histopathology tumor volume (Pearson coefficient 0.237, p = 0.003). On linear regression analysis histopathology tumor volume and magnetic resonance tumor volume were correlated (r(2) = 0.401, p <0.00001). On ROC analysis AUC values for magnetic resonance tumor volume, prostate specific antigen and age in estimating tumors larger than 0.5 cm(3) at histopathology were 0.949 (p <0.0000001), 0.685 (p = 0.001) and 0.627 (p = 0.02), respectively. Similar results were found in the analysis with shrinkage factor corrected tumor volumes at histopathology.

Conclusions: Magnetic resonance imaging can accurately estimate index tumor volume as determined by histology. Magnetic resonance imaging has better accuracy in predicting histopathology tumor volume in tumors larger than 0.5 cm(3) than prostate specific antigen and age. Index tumor volume as determined by magnetic resonance imaging may be helpful in planning treatment, specifically in identifying tumor margins for image guided focal therapy and possibly selecting better active surveillance candidates.

Figure 1

Imaging of 53-year-old man with serum PSA of 23.7 ng/dl. Axial T2W MRI demonstrates midline peripheral zone lesion (asterisk) with calculated MRI volume of 1.43 cc (A), and corresponding histopathology map confirms presence of Gleason 4 + 4 tumor (inked in black) in midline peripheral zone with volume of 1.48 cc (B).

Figure 2

Linear regression plots of ITV (cm³) at pathology (with and without tissue shrinkage factor correction) and at MRI in 135 patients (for both analyses r²=0.401, p <0.00001). Discrete line represents 95% CI.

Figure 3

Box plots demonstrating changes in shrinkage factor noncorrected (A) and corrected (B) ITV (cm³) at pathology, and at MRI in 3 different Gleason score groups (low—Gleason score 3 + 3, moderate—Gleason score 3 + 4 and 4 + 3, high — Gleason score 4 + 4 and higher). High Gleason group was significantly different from low and moderate groups (p <0.05) at MRI and both pathology analysis groups.

Figure 4

ROC curve estimating shrinkage noncorrected (A) and shrinkage corrected (B) pathology volumes using cutoff of 0.5 cm³ from MRTV, age and serum PSA. Using threshold pathological tumor volume (without shrinkage factor correction) of 0.5 cm³, MRTV predicted HTV with sensitivity of 88% and specificity of 90% (A), whereas sensitivity and specificity were 88% and 95% for predicting HTV after shrinkage factor correction was applied (B).