Assessing cancer risk in the anterior part of the prostate using micro-ultrasound: validation of a novel distinct protocol

We selected consecutive patients who had undergone microUS biopsies and radical prostatectomy between May 2019 and June 2022 in the three participating institutions. We included patients with presumed non-metastatic disease at pre-surgery workup, available prostate mpMRI with visual synopsis, available microUS prostate scans and final pathology. For every patient included in this phase, the local expert uro-genital pathologist provided a visual synopsis displaying all prostate cancer foci, the lesion aggressiveness in terms of Gleason Grade Group [12], individual lesion volume and extracapsular extension. This phase aimed to determine the particular microUS features associated with the presence or absence of clinically significant prostate cancer (any Gleason pattern > / = 4) in the anterior part of the prostate which was defined as the area above the prostatic urethra. 102 selected prostate scans were used in the development of the prostate’s anterior zone cancer risk assessment. Investigators initially reviewed both microUS images and final pathology together to describe the appearance of the microUS in areas of the prostate harboring clinically significant prostate cancer and normal tissue. Training material was then produced to describe these differing appearances to new readers.

Seven investigators evaluated the risk scale in an independent set of 50 microUS prostate scans. We selected two groups of patients who underwent both prostate mpMRI and a microUS evaluation in the three participating institutions from May 2018 to October 2020. In the “Anterior Cancer” group, 25 patients with confirmed clinically significant prostate cancer detected in the anterior part of the prostate on either radical prostatectomy or biopsy were selected. In the “Benign Anterior” group, 25 patients with no cancer identified on radical prostatectomy in the anterior prostate were selected; cancers below the level of urethra were permitted. All 50 cases provided full observation of the gland and anterior capsule, with no artifacts significantly obscuring the anterior prostate.

MicroUS prostate scans of the selected patients were shown to investigators with varying level of experience with microUS examination. Each reader was previously trained with the specific module developed above and available online. Every reader assessed the 50 prostate scans and noted for every film whether the presence of significant cancer was likely, equivocal or unlikely in the anterior part of the prostate.

We calculated the sensitivity, specificity and positive and negative predictive values for every reader. Descriptive statistics were used, including the median and interquartile range (IQR) for each metric. To evaluate the coherence between the readers, we calculated the Fleiss kappa and the Cronbach alpha values. Area under the receiver-operator curve (AUC) was assessed for each reader. Statistical analysis was performed with MATLAB (The Mathworks, Natick, MA).

Median patients’ age was 65 and 67 years, median prostate volume 38 and 40 ml and median PSA 6.6 and 8.6 ng/ml in the development and the validation population, respectively (Table 1).

Initial review of the imaging set produced a large variety of descriptive terms for both benign and cancerous anterior prostate tissue which are listed in Table 2. These terms were then reviewed as a group and simplified into three consensus statements: 1. low-risk anterior prostate tissue has a hyperechoic appearance often with ductal patches and smooth, contiguous capsule; 2. high-risk anterior prostate tissue has a hypoechoic appearance with irregular, poorly defined margins; 3. care should be taken to rule out common artifacts such as edge artifact, shadowing from trapped gas or calcifications, and BPH nodules.

High-risk tissue as described in Table 2 was further grouped by shape and type of irregularity in the margin into four categories: 1. focal lesions with poorly defined, but generally ovular or lenticular margins; 2. lesions with irregular or scalloped margins (“storm cloud” lesions); 3. lesions with irregular finger-like projections; 4. lesions of the anterior apical horn.

A novel PRI-MUS anterior score was created using high-risk and low-risk features (Fig. 1).

The median sensitivity, specificity and positive and negative predictive value for all readers were 72% (IQR 68–84), 68% (IQR 64–84), 75% (IQR 72–81) and 73% (IQR 71–80), respectively. If equivocal findings were considered as positive, the sensitivity, specificity and negative and positive predictive values were 72% (IQR 72–80), 68% (IQR 68 –76), 73% (IQR 72–77) and 75% (IQR 75–81), respectively. The accuracy was assessed using the AUC under the individual ROC curves (Fig. 2), with a mean SD of 0.75 ± 0.02 (range 0.71–0.77) The Cronbach alpha value was 0.56, indicating a reliable consistency. Absolute consistency measured by the Fleiss kappa value was 0.179, indicating a slight agreement.