Quantitative colour Doppler and greyscale ultrasound for evaluating prostate cancer

Abstract

Introduction

Although transrectal ultrasound is routinely performed for imaging prostate lesions, colour Doppler imaging visualizing vascularity is not commonly used for diagnosis. The goal of this study was to measure vascular and echogenic differences between malignant and benign lesions of the prostate by quantitative colour Doppler and greyscale transrectal ultrasound.

Methods

Greyscale and colour Doppler ultrasound images of the prostate were acquired in 16 subjects with biopsy-proven malignant or benign lesions. Echogenicity and microvascular flow velocity of each lesion were measured by quantitative image analysis. Flow velocity was measured over several cardiac cycles and the velocity–time waveform was used to determine microvascular pulsatility index and microvascular resistivity index. The Wilcoxon rank sum test was used to compare the malignant and benign groups.

Results

Median microvascular flow velocity of the malignant lesions was 1.25 cm/s compared to 0.36 cm/s for the benign lesions. Median pulsatility and resistive indices of the malignant lesions were 1.55 and 0.68, respectively versus 6.38 and 1.0 for the benign lesions. Malignant lesions were more hypoechoic relative to the surrounding tissue, with median echogenicity of 0.24 compared to 0.76 for the benign lesions. The differences between the malignant and benign groups for each measurement were significant (p < 0.01).

Conclusion

Marked differences were observed in flow velocity, microvascular pulsatility, microvascular resistance, and echogenicity of prostate cancer measured with quantitative colour Doppler and greyscale ultrasound imaging. Vascular differences measured together with echogenicity have the combined potential to characterize malignant and benign prostate lesions.

Introduction

Prostate carcinoma is the second most diagnosed cancer in men and the fifth leading cause of cancer death among men worldwide, and in the United States it is the second leading cause of death after lung cancer.^1,2 A fundamental step for the continued growth of tumours and the development of metastases is the induction of angiogenesis. Tumours smaller than 0.5 mm in diameter receive oxygen and nutrients by diffusion, but, in larger tumours, vascular endothelial cells proliferate to form new blood vessels.^3–5 Slow-growing benign neoplasms are generally sparsely vascularized, whereas malignant neoplasms are fast growing and highly vascular.^3,4,6 Prostate cancer growth and metastasis depend highly on angiogenesis, but the mechanisms underlying these changes are still not fully understood.^7,8

Doppler ultrasound imaging provides direct visualization and quantitation of small blood vessels (∼0.5 mm) in the prostate.^9,10 The study of blood flow through these micro blood vessels is important in increasing our understanding of the role it plays in the pathogenesis of prostate diseases. Accompanying Doppler changes, prostate lesions also exhibit echogenic changes in greyscale ultrasound images. Hypoechoic lesions are generally cancers and are easily detected by transrectal ultrasound (TRUS), whereas isoechoic lesions, which are less likely to be cancerous, are hard to detect, especially when they are small in diameter.^11–13 However, several studies have reported a significant overlap in the echogenicity of the lesions, so the use of echogenicity alone for characterizing lesion is often not recommended.^14,15 The overall objective of this study is to improve prostate lesion characterization by ultrasound imaging. Since Doppler blood flow and tissue echogenicity provide independent information, the study explores the potential of using Doppler blood flow with echogenicity to characterize prostate lesions.

We hypothesize that characteristics of blood flow in hypoechoic prostate tumours are different than in suspicious isoechoic regions in the surrounding tissues, and these differences can be exploited to characterize malignant lesions. Towards this goal, mean colour Doppler velocities and their pulsation during the cardiac cycle due to pressure changes are measured in hypoechoic and isoechoic regions of the prostate. Vascular pulsations of microvessels are often visible on the colour Doppler images but rarely used for diagnosis. We explore whether the pulsatility of microvessels can be reliably measured for characterizing malignant prostate cancers. Doppler measurements of the two regions are compared with the biopsy results of the corresponding regions to assess the difference in flow patterns of the two groups. Echogenicity of the lesion is also measured quantitatively and used with Doppler measurements to assess lesion discrimination.

Methods

With approval from the institutional review board (IORG0000029), the study was performed on consecutive subjects recommended for biopsy based on prostate abnormalities, such as abnormal digital rectal examination and/or serum prostate specific antigen (PSA) level greater than 4.0 ng/ml. A total of 200 biopsies, 12–16 per subject as per routine clinical practice, were performed in 16 patients. Of the 200 biopsies, 50 were adenocarcinoma, and the remaining 150 were benign. In 14 patients, biopsies were guided by TRUS. In the remaining two cases where an echogenic lesion could not be seen on TRUS, biopsies were performed using MR–ultrasound fusion. Of the 16 patients, 4 had palpable nodules and 12 had an elevated PSA level (>4 ng/ml).

Image acquisition

All patients had a TRUS examination with greyscale and colour Doppler imaging performed by a board-certified sonographer. Video clips of colour Doppler images were acquired for 3–10 seconds in a fixed plane of the prostate for assessing haemodynamics over the cardiac cycle. The prostate measurements on TRUS were performed 2–24 hours before biopsy. Additionally, greyscale images were recorded to measure regional echogenicity.

Quantitative image analysis

Colour Doppler and greyscale images were analysed offline for vascularity and echogenicity using software written in IDL programming language (Harris Corporation, Melbourne, FL). Guided by the location where the biopsy was performed, two suspicious hypoechoic and two isoechoic regions of interest (ROIs) were manually outlined in the first frame of the video clip of the prostate ultrasound images (Figure 1). The user was blinded to the results of the biopsy at the time of region selection. The ROIs of the first frame were replicated on the remaining frames of the video clip to assess vascular changes over the cardiac cycle.

Figure 1.

Selection of ROIs. Colour Doppler image of a hypoechoic lesion in the right posterior base/mid in the coronal plane. Four ROIs were identified: two in the hypoechoic lesion (white) and two in the normal tissue (yellow).

Colour Doppler flow velocities within the ROIs were measured from the images using the approach described previously.^16,17 The process consisted of two steps. First, the colour scale on the Doppler image was measured by dividing each colour of the bidirectional colour bar on the Doppler image into equal parts and assigning each part a velocity from zero to maximum velocity based on its location in the colour bar. The second step involved detecting the coloured pixels representing microvessels (mv) within the region of interest, and using their colour levels and the colour scale to calculate mean flow velocity $\left(V_{mv}\right)$. Automated vascular measurements were made for each image in the video clip to generate a time–velocity waveform depicting changes during cardiac cycles. From the time–velocity waveform of Doppler microvessels, mean velocity $\left(V_{mv}\right)$, peak systolic velocity (${\mathit{PSV}}_{mv})$, and end-diastolic velocity (${\mathit{EDV}}_{mv})$ were measured (Figure 4) to determine pulsatility index (${PI}_{mv}$) and resistive index (${RI}_{mv}$) of the microvessels, defined as

$${PI}_{mv}=\frac{{\mathit{PSV}}_{mv}-{\mathit{EDV}}_{mv}}{V_{mv}}$$

$${RI}_{mv}=\frac{{\mathit{PSV}}_{mv}-{\mathit{EDV}}_{mv}}{{\mathit{PSV}}_{mv}}$$

Figure 4.

Change in microvascular flow velocity of the prostate over cardiac cycles. The upper and lower dotted lines represent peak systole and end diastole, respectively. The middle-dotted line is the mean velocity over cardiac cycle. (a) and (b) show microvascular velocity measurements in malignant and benign regions of the prostate, respectively.

In addition to Doppler measurements, the mean brightness of the suspicious regions was determined by measuring the average grey level of the pixels enclosed in the ROIs defining the lesion. In the area surrounding the lesion, an isoechoic region was identified and its mean brightness was measured. The ratio of lesion brightness to that of the surrounding tissue was used to define echogenicity, E, quantitatively, where E < 1, E = 1, and E > 1, represent hypo-, iso-, and hyperechoic lesions, respectively.

Statistical analysis

The two vascular and greyscale measurements obtained from each patient were averaged for comparison between malignant and benign groups. The paired Wilcoxon rank sum test was used to establish statistical significance (MedCalc Software, Belgium). p ≤ 0.05 was considered to be significant. Unless specified otherwise, median of each measurement with 95% confidence interval (CI) is reported.

Results

The patient ages ranged from 54 to 82 years (mean ± SD, 67 ± 7 years), PSA levels ranged from 3.4 to 17.2 ng/ml (8.8 ± 5.9 ng/ml), Gleason score ranged from 6 to 9, and prostate volumes ranged from 20 to 96.8 ml (47.9 ± 22.2 ml). Based on visually observed echogenicity, TRUS did not detect abnormalities in the prostate in 2 out of 16 cases even though the pathology report stated that malignant cells were identified in these areas.

Figure 2 shows the distributions of pulsatility, velocity, and echogenicity of the malignant and benign lesions. The measurements show malignant lesions to be less pulsatile and less echogenic than the benign lesions. Mean flow velocity, on the other hand, was higher for the malignant lesions compared to benign tissues. A comparison of colour Doppler velocity with echogenicity shows a good separation with a small overlap between the malignant and benign lesions (Figure 3). Both malignant and benign regions were less echogenic (<1) than the surrounding tissue (Figure 3). The median echogenicity of the malignant tissue was 0.24 (CI: 0.16–0.37) versus 0.76 (CI: 0.73–0.80) for the benign regions. The difference was significant (p < 0.01).

Figure 2.

Relationship between representative ultrasound measurements and prostate lesion types. Squares and dots represent benign and malignant lesions. PI_mv, V_m, and E in the graph represent microvascular pulsatility index, mean velocity, and echogenicity, respectively.

Figure 3.

Mean velocity measured in the regions of the prostate with varying echogenicity. The red dots correspond to the regions identified by biopsy to be malignant. Blue dots represent measurements in the benign regions.

Microvascular waveforms for the malignant and benign prostate regions demonstrated a cyclic change over the cardiac cycle (Figure 4). Although both waveforms showed regular oscillations, individual cycles were different for the two groups, especially at end diastole. In malignant lesions, the flow was continuous and phasic (Figure 4(a)). The waveforms measured in benign regions, on the other hand, were irregular and noisier, possibly due to lower Doppler shift at low velocities (Figure 4(b)). The zero velocity (runoff) or no-flow at end diastole for the benign group lasted for a prolonged period of the cardiac cycle before rapidly increasing over the next cycle (Figure 4(b)).

Microvascular measurements of the malignant and the benign groups are summarized in Table 1. The microvascular median velocity for the malignant regions was 1.25 cm/s (CI: 1.09–1.86), and the microvascular median velocity of 0.36 cm/s (CI: 0.26–0.56) observed for the benign regions. The peak systolic velocity for malignant regions was 2.81 (CI: 2.22–3.10), and peak systolic velocity of 1.89 (CI: 1.52–2.16) was observed for the benign regions. Based on median value, end-diastolic velocity for malignant regions and benign regions was 0.74 (CI: 0.24–1.15) and 0.00004 (CI: 0.00004–0.15), respectively. Microvascular PI for the malignant and benign regions was 1.55 (CI: 0.89–1.86) and 6.38 (CI: 2.95–9.82), respectively. The microvascular RI for the malignant and benign regions was 0.68 (CI: 0.53–0.87) and 1.0 (CI: 0.92–1.00). The differences in median, peak systolic, and end-diastolic velocities and PI and RI indices were highly significant (p < 0.01).

Table 1.

Comparison of vascular and echogenic properties of benign and malignant tissue of the prostate. Median values of each group with 95% confidence intervals (CI) are reported.

Description	Malignant	Benign	Two-sided p values
Echogenicity	0.24 (CI: 0.16–0.37)	0.76 (CI: 0.73–0.80)	<0.01
Mean velocity $\left(V_{mv}\right)$ (cm/s)	1.25 (CI: 1.09–1.86)	0.36 (CI: 0.26–0.56)	<0.01
Peak systolic velocity (${\mathit{PSV}}_{mv})$ (cm/s)	2.81 (CI: 2.22–3.10)	1.89 (CI: 1.52–2.16)	<0.01
End-diastolic velocity (${\mathit{EDV}}_{mv})$ (cm/s)	0.74 (CI: 0.24–1.15)	0.00004 (CI: 0.00004–0.15)	<0.01
Microvascular PI $\left({PI}_{mv}\right)$	1.55 (CI: 0.89–1.86)	6.38 (CI: 2.95–9.82)	<0.01
Microvascular RI $\left({RI}_{mv}\right)$	0.68 (CI: 0.53–0.87)	1.00 (CI: 0.92–1.00)	<0.01

PI: Pulsatility Index; RI: Resistive Index.

Discussion

TRUS-guided biopsy is a widely practised method for histological diagnosis in men with suspected prostate cancer.^12,18,19 Although the use of TRUS has improved significantly over the years and has contributed to an increase in the diagnosis of prostate cancer, the technique is known for low specificity and sensitivity.^20,21 The fact that TRUS did not detect abnormalities in 2 of 16 cases is consistent with a prior study^21,22 in which five of seven prostate cancers were missed by greyscale TRUS but were detected with the help of colour Doppler ultrasound.

In this study, we report microvascular velocity measurements made in biopsy-proven malignant and benign regions of the prostate. Suspicious regions in the prostate were identified on the TRUS images, and in two cases where TRUS did not detect suspicious lesions, MR-ultrasound fusion imaging was used. Colour Doppler flow velocities in the suspicious regions found to be malignant were in the order of 1.25 cm/s, compared to 0.36 cm/s observed in suspicious regions that proved to be benign. These velocities correspond to mean velocities determined by measuring the mean frequency shift of the Doppler spectrum. At clinical ultrasound frequencies of 5–15 MHz, the colour Doppler images visualize blood vessels of size ≥ 0.2–0.5 mm.^23,24 Thus, the measurements reported in this study correspond to networks of blood vessels of submillimetre size.

Tumours induce angiogenesis.^3,4,25 The increased flow velocity in the malignant microvasculature is possibly the result of a greater demand for blood from the downstream angiogenic vessels. The presence of a larger number of angiogenic blood vessels also indicates reduced resistance to flow in the upstream microvessels. Furthermore, neovascularized vessels are primitive avascular channels that lack smooth muscle and often consist only of an endothelial layer and connective tissue.²⁶ Due to the lack of smooth muscle, particularly the muscular medial layer, low resistance to flow in such vessels can be expected. The observation that the microvascular resistivity index for malignant regions of the prostate was lower, 0.71 compared to 0.94, than that estimated for the benign regions (Table 1) is consistent with theoretical expectations and has been noted in malignant tumours such as renal cell carcinoma.²⁷

In addition to angiogenesis effects, increasing cell density in the extravascular tissue could also influence flow velocity in the microvasculature. The increased packing of cells per unit volume due to cancer-related growth could potentially constrict the blood vessels and cause flow velocities to increase. Tighter packing of the cells within the prostate tissue volume also restricts the vasomotion of the microvessels and diminishes the pulsatility of the vasculature. Furthermore, increased cell density also influences echogenicity. In a medium with sparsely distributed cells, ultrasound backscatter increases with the cell density. As the cells become more closely packed together by rapid cell multiplication during tumour growth, the aggregate of cells behaves as a continuous medium,²⁸ thereby reducing ultrasound scatter or echogenicity in the images. Studies performed in blood show that backscatter increases with haematocrit initially but, after peaking at approximately 15%, decreases continuously to zero as the haematocrit is increased to 90%.²⁹

The results of this study show that malignant regions with high microvascular velocity are also hypoechoic (Figure 3). The median echogenicity of the malignant regions was 0.26 compared to 0.77 for the benign regions, suggesting greater cell density for the former. The median microvascular pulsatility for malignant regions was found to be 1.55, which was lower than for benign regions with median value of 6.38 (Figure 4 and Table 1). In short, increased flow velocity accompanied by lower vascular pulsatility and tissue hypo-echogenicity is consistent with the expected appearance of closer packing of the cancer cells.

It has been suggested that colour Doppler ultrasound is useful for determining the aggressiveness of cancer and that higher blood flow corresponds to a higher Gleason score and higher incidence of relapse.³⁰ The observation that malignant lesions have higher flow velocity compared to the benign tissues is consistent with the previous study.³⁰

Conclusion

This study demonstrates the use of quantitative Doppler ultrasound imaging to measure microvascular flow velocities over cardiac cycles and to assess microvascular pulsatility and resistivity indices. The results confirm that vascular characteristics of the malignant and benign lesions are markedly different and consistent with vascular and tissue remodelling known to be associated with tumour growth. Our investigation also shows that computer analysis of pixel colour intensities provides robust measurements of microvascular blood flow velocities, vascular pulsatility, and microvascular flow resistance for differentiation of malignant and benign lesions. Since colour Doppler imaging is routinely available in modern ultrasound scanners, it could play an important role in prostate cancer diagnosis. The number of cases in this study is small. A larger clinical study is warranted to determine the sensitivity and specificity of the combined use of vascular and greyscale features derived from Doppler and B-mode images.