Due to the complexity of the prostatic tissue, there is a substantial overlap in the diagnostic performance for tumors with Gleason score 3+4 and Gleason score 4+3. This may lead to under-treatment or over-treatment of prostatic carcinomas, since the latter is considered to be more aggressive. Texture analysis probes the spatial variation in pixel intensities to quantify intuitive qualities that characterize the inherent texture of objects in an image ¹.

In this study, we evaluated T2-weighted image (T2WI) textural entropy – the measure of disorderness or complexity in the T2W image based on the spatial distribution of pixel intensities, as a potential marker for assessing prostate cancer aggressiveness, especially in separating Gleason score 3+4 from Gleason score 4+3 tumors.

Twenty-three patients with biopsy proven prostate cancer underwent multi-parametric prostate MRI (Table 1) on a 3T scanner (Magnetom Trio; Siemens Medical Solutions, Erlangen, Germany) prior to radical prostatectomy.

For each patient, an experienced pathologist outlined and graded ² cancer foci in whole-mount histology slides. The histopathology slide with the largest cancer area was spatially matched to the closest axial T2W image (Figure 1A) using anatomical landmarks. Tumor regions of interest (ROIs) were delineated in the T2W images, and then transferred to the respective DW and DCE images (Figure 1B) via registration ³.

Two-dimensional gray level co-occurrence matrix-based textural entropy ⁴, apparent diffusion coefficient (ADC), and volume transfer constant (K^trans) and volume of extravascular extracellular space per unit volume of tissue (V_e) were calculated from the T2W, DW and DCE image ROIs, respectively (Figure 1C).

The T2WI-derived textural entropy was correlated with the Gleason scores from the histology using point-biserial correlation and with the quantitative MR physiological parameters (median ADC, Ktrans and Ve) using Spearman correlation. Furthermore, its capability in differentiating between Gleason scores 3+4 and 4+3 tumors was evaluated and compared to that of the quantitative MRI parameters using Mann-Whitney U test (α=0.05) and receiver-operating characteristic curve (ROC) analysis.

A total of 23 tumor regions (mean diameter = 22.2 mm; range = 10-40 mm) were analyzed, 14 of which were Gleason score 3+4 tumors and nine 4+3 tumors. Seventy-eight percent of the tumors were located in the peripheral zone and 22% in the central gland. T2WI textural entropy correlated significantly with Gleason score (rpb = 0.555, p = 0.006) and ADC (rho = 0.595, p = 0.005), but not with K^trans or V_e (Figure 2).

T2WI textural entropy was found to increase with increasing Gleason grade and decreasing ADC. Only T2WI textural entropy was found to differ significantly between Gleason score 3+4 and Gleason score 4+3 tumors (Figure 3). The area under the ROC curve was highest for T2WI textural entropy when discriminating these two prostate tumor patterns (Figure 4).

This study evaluated the utility of T2WI textural entropy in prostate cancer aggressiveness assessment. Texture analysis on T2W images could reveal details sensitive to subtle changes in the tissues, since these images depict prostate structures with both high signal-to-noise ratio and high spatial resolution. In addition, T2W images are easy to acquire, suitable for every patient and less prone to acquisition and technical artefacts compared to DW and DCE images.

Increased prostate cancer aggressiveness is characterized by increased cellularity with decreased extracellular space (i.e. low ADC) ⁵, as well as deterioration of the architectural patterns depicting cellular integrity of the prostate gland due to poor differentiation and glandular structure deformation (i.e. high Gleason grade) ⁶. All of these could result in increased tumor heterogeneity, which is reflected by increased textural entropy in the lesion, possibly explaining the observed moderately positive and negative correlations with Gleason score and ADC, respectively.

Despite the relatively low number of patients in our cohort, textural entropy was capable of differentiating Gleason scores 3+4 and 4+3 tumors with higher accuracy than ADC, K^trans and V_e, indicating the potential of MRI texture analysis in prostate cancer assessment.

The current study indicates that T2WI-derived textural entropy could serve as a diagnostic marker in prostate cancer, sensitive to pathological differences in prostate cancer lesions as indicated by the Gleason grading system. However, there is the need for further studies in a larger cohort to validate these findings