Accuracy of HistoScanning™ for the prediction of a negative surgical margin in patients undergoing radical prostatectomy


  • G.S. and J.S. contributed equally to the manuscript.

Correspondence: Georg Salomon, Martini-Clinic, Prostate Cancer Centre, University Medical Centre Eppendorf, Martinistr. 52, 20253 Hamburg, Germany.



What's known on the subject? and What does the study add?

  • HistoScanning™ is a novel ultrasonography technique for visualization of prostate cancer. The technology it uses and its encouraging results in prostate cancer foci detection in 29 men with prostate cancer have been described previously. A recent study confirmed these results in 31 patients. The sensitivity and specificity of prostate cancer foci detection were 90% and 72%, respectively. These promising results raised the question of whether HistoScanning™ technology might also be helpful in therapy planning.
  • Preoperative knowledge of the distribution and size of tumours might be useful for treatment planning of a nerve-sparing radical prostatectomy. We tested the ability of HistoScanning™ to predict a negative surgical margin in 80 patients undergoing radical prostatectomy.


  • To assess the accuracy of HistoScanning™ (HS) as a visualization tool for preoperative treatment planning for nerve-sparing (NS) radical prostatectomy (RP).

Patients and Methods

  • A retrospective study was carried out on 80 patients with prostate cancer undergoing RP from October 2009 to December 2009.
  • All patients underwent a HS procedure 1 day before surgery.
  • Frozen sections (FSs) were performed on each latero-posterior side of the prostate to assess for the presence of cancer.
  • On the HS analysis, the region corresponding to that removed at FS was assessed for suspicious lesions. The size of suspicious lesions within this volume was compared with the FS histopathological analysis.


  • HS results corresponded to a 93% probability of having a negative surgical margin in the FSs. The presence of a HS volume ≥0.2 mL in a specific side was associated with a 3.7 times increased risk of a positive surgical margin at FS.


  • HS has the potential to assist in the planning of NSRP.
  • Larger, multicentre studies need to be performed for validation of these encouraging results.





radical prostatectomy


frozen section


extracapsular extension


native radio frequency


area under the curve


The ideal trifecta for radical prostatectomy (RP) is oncological control, continence and potency recovery. A nerve-sparing (NS)RP is performed whenever possible, to maximize potency and to reduce the possibility of compromise of cancer control.

Frozen section (FS) procedures are routinely performed during RP [1] in the Martini-Clinic. The FS is used to assess the posterior lateral margin of the excised prostate to determine whether there is any cancer at the margin of the resection line. If the margin is cancer-free then the primarily preserved neurovascular bundle can be left in situ without compromising the removal of all cancerous tissues. This approach leads to a high rate of patients harbouring pT2 tumours in which a NS RP could have been safely performed [1, 2]. If a positive surgical margin is found in the FS, the ipsilateral neurovascular bundle and Denonvilliers (DE) fascia will be removed in the same operation.

An intraoperative FS procedure can be time- and cost-intensive and some centres do not perform this at all. Alternative strategies are therefore required to provide the best method of ensuring oncological safety. In addition to nomograms that predict, to some extent, side-specific extracapsular extension (ECE), the precise visualization of tumour foci would allow preoperative treatment planning to ensure a safe NS procedure or a primary non-NS procedure.

There is no established imaging method in routine clinical practice that can demonstrate the presence of prostate cancer within the prostate and reliably demonstrate ECE and assess its Gleason score. MRI, with its multiparametric approach for example, has a local staging accuracy of 50–92% [3, 4] and has not yet become a routine imaging method for ECE prediction. Knowledge of these variables is essential, however, for decisions regarding treatment or surveillance.

Recently, innovative TRUS-based techniques, e.g. elastography [5], computerized TRUS [6], and contrast-enhanced ultrasonography [7] have been proposed for the non-invasive detection of prostate cancer. Another innovative technique is HistoScanning™ ([HS] Advanced Medical Diagnostics, Waterloo, Belgium) [8]. HS technology uses information from backscattered ultrasonographic raw data (i.e. the native radio frequency [NRF] data) that are processed by a set of ‘tissue-characterization algorithms’. The first studies on patients with peri-operative TRUS and whole-mounted gland histological examination suggested high sensitivity and specificity for the detection of tumour foci ≥0.5 mL in size [8, 9]. These studies reported on small numbers of patients, however, and their results have been questioned [10].

Initial data on HS suggested that this technology can exclude the presence of foci ≥0.5 mL with 100% accuracy [8]. If such a high negative predictive value holds true, HS may help in the prediction of a negative surgical margin, which may reduce the number of intraoperative FSs needed and allow complete sparing of the neurovascular bundle. We investigated the ability of the HS technique to exclude the presence of cancerous tissue in the vicinity of neurovascular bundles.

Materials and Methods

Between October 2009 and December 2009, 80 consecutive patients with biopsy-proven prostate cancer scheduled for intrafascial NSRP [2] underwent TRUS on the day of their hospital admission. In line with the Martini-Clinic's standard procedures, patients consented to use of their data for scientific purposes.

The TRUS was performed with a Pro Focus Ultraview machine (B-K-Medical, Copenhagen, Denmark) and included a three-dimensional examination of the prostate using a motorized three-dimensional mover in the transversal plane. NRF ultrasonographic data were acquired to the HS system hard drive after a US examination that lasted 5–10 min. If the size of the prostate exceeded 50 mL, two overlapping scans were acquired to ensure that the whole prostate had been captured. The analysis of the NRF data took place after surgery and so HS analyses had no influence on surgical practice. Three urologists performed the TRUS, one urologist performed the HS analysis and five ‘high-volume’ surgeons performed the RP.

Nerve-sparing RP [2, 11] was performed for all patients [2] In the Martini-Clinic, the decision to perform a NS procedure is made in almost every patient with potentially localized prostate cancer, with the exception of those with Gleason 5 tumours.

After gland removal, 3–4-cm slices were cut tangentially to the gland surface adjacent to the neurovascular bundle (Fig. 1). The external surface and margins were stained with different ink colours and immediately sent to the pathology department. Perpendicular to the colour-marked surfaces, 3–5-mm thick slices were cut and frozen. From each of these slices, two 5-μm thick sections were cut and mounted on microscopic glass. After that, sections were rapidly stained with haematoxylin and eosin and examined. Evidence for prostate margin negative for cancer was established if no microscopic contact between the cancer and the surgical margin was observed in any of the sections. Any tumour adjacent to the margin resulted in resection of the primarily spared neurovascular bundle on the side of the positive FS.

Figure 1.

A slice of 3–4 cm diameter was cut tangentially to the gland surface adjacent to the neurovascular bundle. The iatrogenic margin at the side of the prostate was inked yellow. The outer side of the slice was inked green (right side of the prostate).

The HS analysis was performed using semi-automatic HS software (Advanced Medical Diagnostics). Using drawing tools embedded in the HS machine, the urologist first grossly defined the contours of the gland in the sagittal and transversal planes. The prostate outer limits were then manually refined resulting in an accurate delineation of the gland margins. The HS analysis was performed and areas found to be suspicious for cancer were displayed in red on the video screen (Fig. 2). The video screen also provided volume data for each focus. This full process, culminating in HS analysis, took 10–12 min.

Figure 2.

Suspicious HS areas are marked red in the transversal ultrasonography picture. A virtual cutting line was drawn on both sides of the prostate corresponding to areas where slices for FSs were taken. a) A slice of 3–4 cm diameter was cut tangentially to the gland surface adjacent to the nerve bundle. b) Iatrogenic margin at the side of the prostate was inked yellow. The outer side of the slice was inked green (right side of the prostate).

Virtual cutting lines, adjacent to each of the left and right neurovascular bundles, and representative of the inner extremity of the removed for FS, were drawn onto the three-dimensional video display (Fig. 2). The location and size of the suspicious lesions within these volumes corresponding to the tissue removed for FS analysis was recorded.

The data analysis focused on the ability of the HS analysis to predict whether the FS was negative, so that a NS procedure could have been performed on that side. A prediction was scored as positive if FS confirmed the absence of cancer at the resected margin. Two different volume thresholds for potentially cancerous foci at HS, 0.2 mL and 0.5 mL, were selected for calculations of FS prediction. As the volume of foci detected at HS departed from normal distribution, with a strongly left skewed distribution (i.e. small cancers were the majority), for data description we used medians and interquartile ranges, for comparisons between groups we used non-parametric statistical tests, and HS volumes were log-transformed before use as continuous variables. We compared proportions using the uncorrected chi-squared test. To evaluate whether HS side-specific volumes added information beyond side-specific biopsy results alone for predicting the possibility of side-specific NSRP, we fitted a logistic regression model, with the endpoint being the absence of side-specific NSRP. Independent variables included in the model were the ipsilateral biopsy result (negative or positive), the highest side-specific Gleason grade found in biopsy cores, the pre-biopsy PSA level and the ipsilateral HS volumes. A positive biopsy result was automatically associated with a Gleason score, which implied collinearity. To overcome this limitation, we created a three-level category variable that had a value ‘1’ when all side-specific biopsies were negative, a value ‘2’ when at least one side-specific biopsy was positive and a dominant Gleason score of 3 was recorded on all biopsies taken on that side, and ‘3’ when at least one side-specific biopsy was positive and a dominant Gleason score of 4 or 5 was recorded on all biopsies taken on that side. For assessing the predictive accuracy of the variables included in the model, we calculated the receiver–operator characteristic area under the curve (AUC) using the c-statistics procedure in MEDCALC (MedCalc Software, Mariakerke, Belgium). The predictive accuracy of the AUC describes the extent to which the values predicted by the logistic model match the observed values, thus providing an estimation of the uncertainty attributable to statistical fluctuations and ‘noise’ in the input data values.


Table 1 summarizes the main patient characteristics. Eighty patients were included in the study, representing 160 sides where intraoperative FSs were taken. FS analysis showed negative surgical margins in 122 (76%) sides and positive surgical margins in 38 (24%). There were a total of 38 resected neurovascular bundles. Postoperative histology showed that in 62 (78%) of the prostate glands, cancer was organ-confined. ECE and seminal vesicle invasion was found in 12 and in six specimens, respectively. Positive surgical margins were found in eight (10%) specimens.

Table 1. Characteristics of 80 patients who underwent RP.
Preoperative data 
Mean (range) age64 (48–76)
Mean (range) PSA, μg/mL6.0 (1.3–33.8)
Clinical stage, n (%) 
T1c63 (79)
T2a14 (18)
T2b3 (3)
NSRP, n (%) 
On both sides47 (59)
On right or left only28 (35)
On neither side5 (6)
Histology of the prostate gland 
Stage, n (%) 
pT2 (OC)62 (78)
pT3a (ECE)12 (15)
pT3b (seminal vesicle invasion)6 (8)
Positive surgical margins (R)8 (10)
Gleason score, n (%) 
3 + 313 (16)
3 + 448 (60)
4 + 314 (18)
4 + 41 (1)
4 + 54 (5)
Lymph node status, n (%) 
N056 (70)
N15 (6)
NX19 (24)

No focus or a focus <0.20 mL was found by HS in 38 (38%) left sides and 35 (44%) right sides (Table 2). Of these 30 left sides, 28 (93%) had undergone NSRP on the left side and of the 35 right sides, 31 (89%) had undergone NSRP on the right side. After pooling results for both sides in Table 2, the sensitivity of HS volume <0.2 cc for performing NSRP was 48%, the specificity was 84%, the negative predictive value was 34% and the positive predictive value was 91%. Hence, a threshold of 0.2 mL was seldom associated with the performance of NSRP when in fact the neurovascular bundle was invaded by cancer. This high specificity was at the expense of a relatively low sensitivity.

Table 2. Volume of foci estimated by HS and performance of NSRP.
 Right side of the prostate glandLeft side of the prostate gland
YesNoTotal% with NSRPYesNoTotal% with NSRP
All patients6317800.795921800.74
HSS analysis        
No focus – focus <0.20 cc282300.93314350.89
Focus 0.20–0.49 cc144180.7898170.53
Focus ≥0.50 cc21′11320.66199280.68

Univariable logistic regression in Table 3 shows the association between three factors and non-performance of NSRP. The pre-biopsy serum PSA level was not associated with performance or non-performance of NSRP, whereas side-specific biopsy, suggesting the presence of Gleason 4 cancerous tissue or side-specific detection of cancerous lesion ≥0.2 mL in volume, was strongly associated with non-performance of NSRP. The multivariable logistic regression model shows the independent contribution of each factor on non-performance of NSRP. A side-specific dominant Gleason score of 4 was associated with a nearly fourfold increase in the impossibility of performing NSRP on that side. In addition, the presence of a HS volume ≥0.2 cc in a specific side was associated with a 3.7-fold increased risk that NSRP could not be performed on that side. Because of the adjustment for other factors in the model, this association with HS is independent of that of the side-specific biopsy result and of that of the side-specific dominant Gleason grade. Of note, a difference of 8.03 existed between the −2*log likelihood of a model including biopsy Gleason grade and PSA levels, and a model including the two former variables plus HS volume 0.2 to 0.49 or ≥5 cc (2 degrees of freedom, P = 0.018).

Table 3. Factors associated with non-performance of side-specific NSRP.
 OR95% CIP
  1. *Odds ratios estimate the likelihood that each factor, considered individually, is associated with NSRPnot being performed. Adjusted odds ratios estimate the likelihood that a factor is associated with NSRP not being performed and the adjustment takes into account the influence of other factors included in the model. GS, Gleason score.
Univariate analysis*   
Side-specific positive GS = 3 vs negative biopsy result0.600.23; 1.610.312
Side-specific positive GS = 4 vs negative biopsy result5.122.05; 12.800.001
Serum PSA level, risk per ng/mL increase1.020.95; 1.100.556
HS volume ≥0.2 and <0.5 cc vs <0.2 cc4.561.54; 13.520.006
HS volume ≥0.5 cc vs <0.2 cc4.751.75: 12.880.002
Log (HS volume, continuous variable)2.481.36; 4.520.003
Multivariate analysis: HS volume in categories   
Side-specific positive GS = 3 vs negative biopsy result0.530.19; 1.460.218
Side-specific positive GS = 4 vs negative biopsy result3.881.50; 10.100.005
Serum PSA level, risk per ng/mL increase1.040.96; 1.130.369
HS volume ≥0.2 and <0.5 cc vs <0.2 cc3.661.16; 11.60.027
HS volume ≥0.5 cc vs <0.2 cc3.841.35; 11.00.012
Multivariate analysis: HS volume as a continuous variable (log-transformed)
Side-specific positive GS = 3 vs negative biopsy result0.330.08: 1.360.124
Side-specific positive GS = 4 vs negative biopsy result5.291.84; 15.20.002
Serum PSA level, risk per ng/mL increase1.150.99; 1.330.051
Log (Histoscaning volume)2.371.15; 4.880.019

An actual value for each side-specific HS volume was available for 68 patients. Incorporating the logarithm of the actual HS volume as a variable instead of the threshold values ≤0.2, 0.2–0.49 and ≥0.5 cc, we fitted a similar model to that shown in Table 3. For each doubling of the logarithm of the side-specific HS volume (corresponding to an increase from 0.25 to 0.5 cc or from 1.25 to 1.60 cc), the impossibility of performing NSRP increased by a factor of 2.39 (95% CI: 1.24–4.62; P = 0.009).

The predictive accuracy assessed by the AUC was 0.76 (96% CI: 0.67–0.83) for the model in Table 3 and 0.79 (95% CI: 0.71–0.85) for the model using the logarithm of the actual HS volume.

Preservation of the neurovascular bundle was far more frequent when pathological stage was pT2 than when it was pT3 (Table 4). The total HS volume was significantly increased in men with pathological stage pT3 vs pT2, and in men with dominant pathological Gleason grade 4 vs Gleason grade 3 (Table 5).

Table 4. Pathological prosate cancer stage in 80 patients and neurovascular bundle preservation.
 Preservation of the neurovascular bundle
On both sides, n (%)On one side, n (%)On neither side, n (%)Total
  1. *P < 0.001. Chi-squared test for heterogeneity.
TNM at histology*    
244 (94)17 (61)1 (20)62
33 (6)11 (39)4 (80)18
Table 5. Total volume of HS-detected suspicious foci for 68 patients treated with RP.
 Stage at pathologyPredominant Gleason grade at pathology
No. of patients54145117
Median total HS volume, cc0.421.830.361.70
Interquartile range0.24–1.380.32–4.700.21–1.360.47–3.20
Wilcoxon rank-sum testP = 0.045 P = 0.008 


The ideal candidate for a secure NS procedure is a patient with organ-confined prostate cancer. In patients with potential ECE a NS procedure should be considered whenever oncologically indicated [12]. Recent analysis of 11 069 patients who underwent RP in our clinic show that, in non-organ-confined cancers, intraoperative FSs allow the tailoring of the NS procedure according to the individual cancer without compromising cancer control [11].

The FS is to date the most reliable method for making an informed choice on leaving the neurovascular bundle in place or for removing it [1]. For this reason, use of FS at the Martini-Clinic is routinely performed in patients undergoing RP for clinically localized prostate cancer; however, this is currently not possible in every institution.

The present study suggests that, when no HS volume or a HS volume <0.2 cc is found in a prostate side, ECE can be excluded with 91% certainty. As a consequence, 40% of the FSs we performed during the course of the present study could have been avoided if the surgeon had had knowledge of HS results. The significant difference associated with addition of the HS volumes to a logistic model already including biopsy results (with Gleason score) and PSA levels (Table 3) indicates that HS seems to bring important additional clinical information.

A preoperative imaging method that can accurately predict the probability of a negative surgical margin has several advantages. Firstly, in centres that do not perform FS, HS analysis could predict whether a NS procedure may be securely performed. This would potentially increase the number of NS procedures, representing a real benefit to patients. Secondly, the rate of FS could be decreased in centres where this technique is the standard of care.

The histopathological characteristics of the surgically removed prostate specimens were consistent with the likelihood of NSRP: a bilateral NSRP was 4–5 times more frequent in pT2 than pT3 cancers. Histopathological characteristics were also consistent with HS analysis: the tumour volume estimated by HS was substantially greater in pT3 than in pT2 cancers, as well as in predominant Gleason grade 4 than Gleason grade 3. In this respect, our findings agree with those previously reported; these indicated a high correlation between cancer volume predicted by HS and volumes found by histological examination of whole-mounted prostate glands [8].

The predictive accuracy we obtained (AUC = 0.79) was intermediate to those associated with the base models developed in American subjects by Ohori et al. [13] (AUC = 0.72) and by Steuber et al. [14] in European subjects (AUC = 0.83). These models included clinical stage, PSA level, and biopsy Gleason score. These two studies involved larger numbers of patients, and the limited size of the present study did not permit greater in-depth exploration of the predictive accuracy.

HistoScanning™ has two main advantages on model-based predictions of cancer extension. First, HS visualized cancerous lesions and thus provided anatomic pictures useful for surgery planning. Second, as a consequence of the visualization, side-specific prediction is possible. Furthermore, an improvement in the HS technology is likely to increase the capacity to characterize prostate tissues in the vicinity of neurovascular bundles.

Although its preliminary data are encouraging, the present study has limitations. The sample size is relatively small. Also, no estimation was made of the distance from focus to posterior limit of the prostate. There is emerging evidence that this may be a marker of the quality of the US scan performed and thus of the quality of the subsequent HS analysis. This may explain some of the false-positive and false-negative results in the study. Another limitation of this technique might be the quality of scans in patients with prostatitis or calcifications; these factors might affect the quality of US. To determine if HS has the ability to securely predict ECE, will discriminate between significant and insignificant tumours, has the ability to differentiate between different Gleason scores as presented by Simmons et al. ([15]and how well HS correlates to tumour volume, will require larger prospective and multicentre studies. The real benefit of HS could ideally be evaluated by excluding patients with cancer or patients without cancer who are undergoing cystoprostatectomy using a saturation biopsy as the reference test.

In conclusion, FSs are recommended until further studies confirm the findings of the present study. Further new investigations involving a larger number of patients in a multicentre study setting and using a wider range of variables offered by HS analysis, such as the distance from the posterior limit of detected foci to the ipsilateral zone of the prostate gland in immediate contact with the neurovascular bundle, are thus desirable.

Conflict of Interest

Philippe Autier and Caroline Moore are Paid Consultants of Avanced Medical Diagnostics.