Medical Dosimetry
Volume 36, Issue 1 , Pages 85-90, Spring 2011

Factors Affecting Prostate Volume Estimation in Computed Tomography Images

  • Cheng-Hsiu Yang, M.Sc.

      Affiliations

    • Department of Power Mechanical Engineering, National Tsing Hua University, Hsinchu, Taiwan
    • ,
  • Shyh-Jen Wang, Ph.D.

      Affiliations

    • Divisions of Experimental Surgery, Department of Surgery, Taipei Veterans General Hospital, Taipei, Taiwan
    • Institute of Biomedical Engineering, National Yang Ming University, Taipei, Taiwan
    • ,
  • Alex Tong-Long Lin, M.D.

      Affiliations

    • Divisions of Urology, Department of Surgery, Taipei Veterans General Hospital, Taipei, Taiwan
    • ,
  • Chao-An Lin, Ph.D.

      Affiliations

    • Department of Power Mechanical Engineering, National Tsing Hua University, Hsinchu, Taiwan
    • Corresponding Author InformationReprint requests to: Chao-An Lin, Ph.D., Department of Power Mechanical Engineering, National Tsing Hua University, Hsinchu 30013, Taiwan

Received 26 February 2009; accepted 7 January 2010. published online 04 March 2010.

Article Outline

Abstract 

The aim of this study was to investigate how apex-localizing methods and the computed tomography (CT) slice thickness affected the CT-based prostate volume estimation. Twenty-eight volunteers underwent evaluations of prostate volume by CT, where the contour segmentations were performed by three observers. The bottom of ischial tuberosities (ITs) and the bulb of the penis were used as reference positions to locate the apex, and the distances to the apex were recorded as 1.3 and 2.0 cm, respectively. Interobserver variations to locate ITs and the bulb of the penis were, on average, 0.10 cm (range 0.03–0.38 cm) and 0.30 cm (range 0.00–0.98 cm), respectively. The range of CT slice thickness varied from 0.08–0.48 cm and was adopted to examine the influence of the variation on volume estimation. The volume deviation from the reference case (0.08 cm), which increases in tandem with the slice thickness, was within ± 3 cm3, regardless of the adopted apex-locating reference positions. In addition, the maximum error of apex identification was 1.5 times of slice thickness. Finally, based on the precise CT films and the methods of apex identification, there were strong positive correlation coefficients for the estimated prostate volume by CT and the transabdominal ultrasonography, as found in the present study (r > 0.87; p < 0.0001), and this was confirmed by Bland-Altman analysis. These results will help to identify factors that affect prostate volume calculation and to contribute to the improved estimation of the prostate volume based on CT images.

Key Words: Prostate volume, Computed tomography, Bottom of ischial tuberosities, Penile bulb

 

Back to Article Outline

Introduction 

Accurate determination of the volume and location of the prostate gland has a significant impact on the treatment regimen for patients. Prostatic volume (PV) estimation is necessary to determine the appropriate dose of radiation that is applied directly to the tumor site via seed implantation for brachytherapy. Also, PV is an important parameter in the diagnosis and management of both benign prostatic hyperplasia (BPH)1 and bladder outlet obstruction (BOO).2, 3

In contrast to ultrasonography, computed tomography (CT) is by far the most widely used technique to formulate a radiotherapy plan for prostate cancer4 and in postoperative dosimetry for brachytherapy. However, the PV as estimated by CT scans has been reported to be larger than PVs measured by 2D transrectal ultrasound,5, 6, 7 transrectal ultrasound step-section planimetry,6, 8, 9 and magnetic resonance images.9, 10, 11 The overestimation of volume estimation ranges from 30%–50%, although the causes of overestimation have not been ascertained. Prostate volume measurement using CT is not straightforward because it is based on contour segmentation drawn by a physician. Unexpected inclusion of the puborectalis muscle and the anterior venous plexus within the midprostate has also been a problem.10, 11 Badiozamani et al.12 indicated that overestimation was because of the inadequate contouring of the soft tissues surrounding the prostate gland.

In addition, delineation of the prostate using CT showed that the largest observer variations were in locating the apex of the prostate. The reason for these large variations was poor discrimination by the CT of the prostatic apex from the muscles of the pelvic floor.11, 13, 14 To locate the apex of the prostate, three procedures have been commonly adopted. One procedure is retrograde urethrography (RUG); researchers have advocated this to be the best method for defining the prostatic apex.15, 16, 17 RUG correlates with the prostatic apex but is an unnecessarily invasive procedure. Urethral strictures from infection, trauma, etc., will preclude an acceptable RUG. The other procedures adopt the bottom of ischial tuberosities (ITs) and the bulb of the penis as 2 noninvasive markers for locating the prostatic apex and have been investigated in recent studies.13, 14, 18, 19 Wilson et al.18 reported that 95.4% of prostate patients had ITs lying approximately 1.5 cm or more below the prostatic apex. Other studies have indicated that the distance was 2 cm or more for 83% of the patients13 and was, on average, 2.4 cm19 in all of the patients. On the other hand, the bulb of the penis lies approximately 0.3 cm below the peak of the urethrogram,14 which is 1 cm below the prostatic apex.16

Therefore, not only could the soft tissues surrounding the prostate gland and the identification method of location of the prostatic apex affect PV estimation with CT images, but also some technical aspects could affect the volume estimation, such as the slice thickness of the CT data. The aim of this study was to investigate the apex-locating methods and the slice thickness and how these methods affect PV estimation using CT. The estimated PV based on CT will be also contrasted with those volumes measured by transabdominal ultrasonography (TAUS) to assess their correlations.

Back to Article Outline

Methods and Materials 

Patients with lower urinary tract symptoms (LUTS) were admitted to Taipei Veterans General Hospital for surgical intervention. Informed consent was obtained from all 28 volunteers, including 24 adult men with LUTS and 4 symptom-free adult men.

All of the volunteers underwent evaluation of PV with CT and TAUS. For ultrasonography measurement, an ultrasound scanner (GE Logiq 9, GE Healthcare, Buckinghamshire, UK) was used. Furthermore, PV defined by ultrasonography was calculated using the ellipsoid formula (PV = π ÷ 6 × [width (cm) × thickness (cm) × length (cm)]), where width (right-left) and thickness (anterior-posterior) were measured on the transverse plane, and length (cranial-caudal) was measured on the sagittal plane.

CT examinations were performed with a conventional CT scanner (Toshiba Aquilion 64, Toshiba American Medical System, Inc., Tustin, CA). Axial CT images of the pelvis and prostate gland were obtained with men in a supine position. The CT slice thicknesses from the 28 subjects were not all the same (mean 0.1 cm, range 0.5–0.05 cm). To analyze the effect of CT slice thickness, 17 sets of CT data with 0.8-mm-thick slices were studied. The study of the slice thickness effect relies on eliminating slices to increase a thickness from 0.08 cm to 0.16, 0.24, 0.32, 0.4, and 0.48 cm; the procedure is illustrated in Fig. 1. The CT images were analyzed using a software package (AMIRA 3.1, Visage Imaging GmbH, Berlin, Germany), which included contour segmentation and volume calculation. Three well-trained observers, the first author and two graduate students studying medical physics and medical imaging engineering, independently contoured prostate glands from CT images, whereas only the first author made the TAUS measurements. To locate the apex of the prostate, two apex markers were adopted and discussed. One marker was the bottom of ITs, and the other marker was the bulb of the penis (Fig. 2). The ITs lie approximately 2 cm13, 18, 19 below the prostatic apex, and the bulb of the penis lies approximately 1.3 cm below the apex of prostate.14, 16 According to the apex markers and the slice thickness, the slice number locating the prostate apex could be identified.

  • View full-size image.
  • Fig. 1. 

    Schematic illustration of doubling the slice thickness by eliminating slices. (A) An original slice thickness model, i.e., 0.08 mm; (B) and (C) two scenarios of doubling the slice thickness, i.e., 0.16 mm, where the shaded areas cannot be included in the volume estimation.

  • View full-size image.
  • Fig. 2. 

    3D reconstructed image of the lower urinary tract system based on the CT images, showing the relative locations of the prostate apex with the bulb of the penis and the bottom of ischial tuberosities. ITs = the bottom of the ischial tuberosities; Bulb = bulb of penis.

The mean and standard deviation (SD) were calculated for volume defined by CT and TAUS. A paired two-tailed t-test was used to estimate the difference between the means of the two independent samples. Data were considered to be statistically significant with a p-value less than 0.05. The Pearson's correlation coefficient and the Bland-Altman analysis20 were used to analyze the correlation between CT and TAUS. Bland and Altman indicated that if the true volume of a variable is unknown, then it is more suitable to plot the difference between the two measurement methods against the mean of the two methods. Assuming a normal difference distribution, the upper and lower limits of agreement were thus calculated as the average difference ± 1.96 times the SD of the differences.

Back to Article Outline

Results 

Before estimating the PV from the CT data, the location of the prostatic apex should be defined. The bulb of the penis and ITs were adopted as apex markers to localize the prostatic apex (CTBulb represents the PV assessed by CT with the bulb of the penis, and CTITs represents the PV assessed by CT with ITs). The prostate volume, defined by CT with 2 kinds of apex-localizing methods, was calculated in each volunteer by all 3 observers. Interobserver variations to locate ITs and the bulb of the penis were, on average, 0.10 cm (range 0.03–0.38 cm) and 0.30 cm (range 0.00–0.98 cm), respectively.

The prostate volumes were estimated in 28 volunteers using CTBulb, CTITs and TAUS. The mean values and SD of PV are summarized in Table 1. For the definition of PV, the interobserver variation (1 SD) was, on average, 3.67 cm3 (range 0.88–6.79 cm3) for the CTBulb and 3.66 cm3 (range 0.69–8.55 cm3) for the CTITs. The TAUS volumes were found to be slightly smaller than the CT volumes. According to the average values of the 3 observer measurements, the mean prostate volume difference between CTBulb and TAUS was 1.50 cm3 (SD 10.86 cm3; p = 0.47), and that between CTITs and TAUS was 3.07 cm3 (SD 10.49 cm3; p = 0.13). The mean prostate volume percentage difference was 0.53% (SD 28.14%) between CTBulb and TAUS and was 4.36% (SD 26.2%) between CTITs and TAUS. The prostate volume obtained with CTBulb was 1.57 ± 1.99 cm3 (p < 0.001) smaller than the volume obtained by CTITs (volume percentage difference 3.83 ± 4.13%).

Table 1. The mean and SD of prostate volume (unit cm3) determined by TAUS and CT for three observers
ObserversTAUSCTBulbCTITs
A45.41±20.4747.85±22.3448.00±22.26
B—47.04±21.5149.16±21.72
C—45.86±22.0948.27±22.57
Average value45.41±20.4746.91±21.7548.48±21.94

Abbreviations: TAUS, volume assessed by transabdominal ultrasonography; CTBulb, volume assessed by CT with the bulb of the penis marker; CTITs, volume assessed by CT with the bottom of the ischial tuberosities marker.

There were strong positive correlation coefficients for the estimated prostate volume by CT and TAUS (Fig. 3). The Pearson correlation coefficient, r, between TAUS and CTBulb was 0.87 (p < 0.0001), whereas that between TAUS and CTITs was 0.88 (p < 0.0001). In addition, r between CTBulb and CTITs was 0.99 (p < 0.0001). Figure 4 shows good agreement between CT and TAUS, demonstrated by the Bland-Altman analysis. In this figure, the solid lines represent the mean difference, and the dashed lines represent the lower and upper limits of agreement in a range of ± 1.96 SD around the mean, where the thin lines and the thick lines collocate with the CTBulb and CTITs, respectively.

Seventeen of 28 CT data points, whose slice thicknesses were 0.08 cm, were adopted in the present study to discuss the influence of slice thickness on volume estimation. The slice thickness varied from 0.08–0.48 cm, and the effect of slice thickness on volume estimation is shown in Fig. 5, Fig. 6 using CTBulb and CTITs, respectively. The volume deviation from the reference case (0.08 cm), which increases in tandem with the slice thickness, was within ± 3 cm3, regardless of the adopted apex locating reference positions. In Fig. 5, Fig. 6, subjects are categorized into 5 volume ranges: <30 cm3, 30∼40 cm3, 40∼50 cm3, 50∼60 cm3, and >60 cm3. It was observed that the prostates with smaller volumes were more affected by CT slice thickness than the large prostate glands, and the difference even reached 10%.

  • View full-size image.
  • Fig. 5. 

    Effect of CT slice thickness on volume estimation based on the penile bulb, where the PVBulb,ST=0.08 represents the prostate volume estimation by CTBulb with 0.08-cm slice thickness.

  • View full-size image.
  • Fig. 6. 

    Effect of CT slice thickness on volume estimation based on the ITs, where the PVITs,ST=0.08 represents the prostate volume estimation by CTITs with 0.08-cm slice thickness.

Back to Article Outline

Discussion 

The overestimation of PV measured by CT has been previously reported in several articles,6, 8, 9, 10, 11, 12 and this was attributed to inadequate contouring.10, 11, 12 However, prostate volume measurement using CT is not straightforward, because the measurements are based on contour segmentation drawn by a physician. Moreover, the soft tissues surrounding the prostate gland are easily included, such as seminal vesicles, the bladder wall, the rectal wall, the puborectalis muscle, the anterior venous plexus, and other muscles of the pelvic floor.

Although the prostate and the seminal vesicles could be visualized on CT images, their color shades were very similar to each other. It is almost impossible to automatically contour the prostate in CT images, so the contours should be selected manually. Also, in the midprostate region, the puborectalis muscle and anterior venous plexus can be misjudged as the prostate gland in CT images; this issue was discussed by Badiozamani et al.12 In this study, the thickness of the puborectalis muscles was assumed to extend to the anterior side around the gland as a loop and was then eliminated. This practice can effectively reduce the estimated PV volume and is particularly important in contouring the midprostate.

The Effect of the Prostatic Apex Identification on Volume Estimation 

The difficulty in contouring the prostate in CT images has been previously reported5, 6, 8, 9, 10, 11 to be one of the reasons for overestimating PV. Localization of the apex of the prostate gland in the CT images has been the most difficult part of contouring. In this study, we adopted the bulb of the penis and ITs as landmarks to localize the apex of the prostate with 1.3 cm and 2.0 cm margins, respectively.

In this work, these 2 apex-localizing methods were calculated in each volunteer by 3 observers. Interobserver variations to locate ITs and the bulb of the penis were, on average, 0.10 cm (range 0.03–0.38 cm) and 0.30 cm (range 0.00–0.98 cm), respectively. This showed that ITs could be clearly observed on CT images, and there were consequently lower interobserver variations for contouring the apex of the prostate. However, the bulb of the penis is a relatively consistent soft-tissue landmark compared with ITs and is advantageous for identifying the prostatic apex.14

Although the locations of the apex may depend on the apex markers, there was only a modest influence observed on the estimated prostate volumes. According to the average prostate volumes measured by the 3 observers, the mean prostate volume difference between CTBulb and CTITs was 1.57 cm3 (SD 1.99 cm3). The difference is because of the cone shape of the prostate near the prostate apex (Fig. 2), and thus the variation of the truncated volume is relatively small. These results also demonstrated that the mean value of the volume difference between CTBulb and TAUS was only 1.50 cm3, and the difference between CTITs and TAUS was around 3.07 cm3. The results did not show a severe overestimation of the prostate volume.

The Effect of the CT Slice Thickness on Volume Estimation 

As shown in Fig. 5, Fig. 6, an increased slice thickness of the CT images usually causes the reduction of estimated prostate volumes. The reason for this is that the larger slice thickness cannot mimic the correct shape of the prostate in the base and apex. The volume deviation from the reference case (0.08 cm) was within ± 3 cm3, and this corresponds to 10% of the small prostate size. In previous studies,5, 6, 9, 12, 13 the CT slice thicknesses adopted were generally greater than 0.25 cm, and the reported results showed that the PV defined by CT was overestimated by over 30%. Therefore, based on the results shown in Fig. 5, Fig. 6, decreasing the slice thickness did not directly solve the problem of overestimation of PVs based on CT images.

The error in locating the apex can also be identified with the slice thickness. As shown in Fig. 7, it is clear that the maximum probability error in marking the peak of the bulb of the penis was 1 slice thickness. In addition, the corresponding maximum error in locating the ITs was also 1 slice thickness, but it was in the reverse direction. As indicated before, the distances to the apex are 1.3 cm and 2 cm, respectively, for the bulb of the penis and the ITs. In addition, because these distances would not be multiples of the adopted slice thickness, the errors can be induced. Depending on the choice of the CT slice, the maximum probability error will be ± 0.5 ST (Fig. 7). Based on the previous analysis, the maximum probability errors were computed. Therefore, the error range in locating the apex based on the bulb of the penis is from −1.5 ST to 0.5 ST and is −0.5 ST to 1.5 ST, as identified by the ITs. Thus, the absolute value of the maximum error will be 1.5 times the slice thickness. If ST = 0.5 cm is the commonly used value, the maximum error of the apex location will be 0.75 cm. Therefore, for identification of the apex, the slice thickness is a considerable factor.

  • View full-size image.
  • Fig. 7. 

    The maximum probability errors in locating the apex of the prostate identified by the slice thickness. ST = slice thickness; ITs = the bottom of the ischial tuberosities; Bulb = bulb of penis.

Back to Article Outline

Conclusions 

The present study explored influences of the apex location methods and the CT slice thickness on the estimated PV based on CT films. Although the interobserver variations of the apex location using the ITs were, on average, 3 times smaller than its penile bulb counterpart, only a modest influence (4%) was observed on the PV estimation. Also, the present CT-based estimated prostate volumes were compatible on average with those estimated by TAUS. On the other hand, the CT slice thickness provided a strong influence on the apex identification, and the maximum error was 1.5 times that of the slice thickness used. In addition, the prostates with smaller volumes were more affected by the CT slice thickness than large prostate glands, and the difference even reached 10%, where the variation of the slice thickness ranged from 0.08–0.48 cm. These findings might assist in the identification of factors that affect prostate volume calculation and improve estimation of the prostate volume based on CT images.

Back to Article Outline

Acknowledgements 

—The authors gratefully acknowledge the support of the Joint Research Program of Veterans General Hospitals and University System of Taiwan (VGHUST97-P5-18).

Back to Article Outline

References 

  1. Van Venrooij GE, Eckhardt MD, Boon TA. Noninvasive assessment of prostatic obstruction in elderly men with lower urinary tract symptoms associated with benign prostatic hyperplasia. Urology. 2004;63:476–480
  2. Lim KB, Ho H, Foo KT, et al. Comparison of intravesical prostatic protrusion, prostate volume and serum prostatic-specific antigen in the evaluation of bladder outlet obstruction. Int. J. Urol. 2006;13:1509–1513
  3. Yang CH, Lin CA, Wang SJ, et al. Computer tomography based noninvasive urodynamics diagnosis method as a predicting tool of bladder outlet obstruction. J. Urol. 2007;177(suppl.):574
  4. Gerber RL, Smith AR, Owen J, et al. Patterns of care survey results: Treatment planning for carcinoma of the prostate. Int. J. Radiat. Oncol. Biol. Phys. 1995;33:803–808
  5. Hoffelt SC, Marshall LM, Garzotto M, et al. A comparison of CT scan to transrectal ultrasound-measured prostate volume in untreated prostate cancer. Int. J. Radiat. Oncol. Biol. Phys. 2003;57:29–32
  6. Kälkner KM, Kubicek G, Nilsson J, et al. Prostate volume determination: Differential volume measurements comparing CT and TRUS. Radiother. Oncol. 2006;81:179–183
  7. Gloi A, McCourt S, Zuge C, et al. A Bland-Altman analysis of the bias between computed tomography and ultrasound prostate volume menasurements. Med. Dosim. 2008;33:234–238
  8. Narayana V, Roberson PL, Winfield RJ, et al. Impact of ultrasound and computed tomography prostate volume registration on evaluation of permanent prostate implants. Int. J. Radiat. Oncol. Biol. Phys. 1997;39:341–346
  9. Smith WL, Lewis C, Bauman G, et al. Prostate volume contouring: A 3D analysis of segmentation using 3DTRUS, CT, and MR. Int. J. Radiat. Oncol. Biol. Phys. 2007;67:1238–1247
  10. Roach M, Faillace-akazawa P, Malfatti C, et al. Prostate volumes defined by magnetic resonance imaging and computerized tomographic scans for three-dimensional conformal radiotherapy. Int. J. Radiat. Oncol. Biol. Phys. 1996;35:1011–1018
  11. Kagawa K, Lee WR, Schultheiss TE, et al. Initial clinical assessment of CT-MRI image fusion software in localization of the prostate for 3D conformal radiation therapy. Int. J. Radiat. Oncol. Biol. Phys. 1997;38:319–325
  12. Badiozamani KR, Wallner K, Cavanagh W, et al. Comparability of CT-based and TRUS-based prostate volumes. Int. J. Radiat. Oncol. Biol. Phys. 1999;43:375–378
  13. Milosevic M, Voruganti S, Blend R, et al. Magnetic resonance image (MRI) for localization of the prostate apex: Comparison to computed tomography (CT) and urethrography. Radiother. Oncol. 1998;47:277–284
  14. Plants BA, Chen DT, Fiveash JB, et al. Bulb of penis as a marker for prostatic apex in external beam radiotherapy of prostate cancer. Int. J. Radiat. Oncol. Biol. Phys. 2003;56:1079–1084
  15. Schild SE, Buskirk SJ, Robinow JS. Prostate cancer: Retrograde urethrography to improve treatment planning for radiation therapy. Radiology. 1991;181:885–887
  16. Roach III M, Pickett B, Holland J, et al. The role of the urethrogram during simulation for localized prostate cancer. Int. J. Radiat. Oncol. Biol. Phys. 1993;25:299–307
  17. Algan O, Hanks GE, Shaer AH. Localization of the prostatic apex for radiation treatment planning. Int. J. Radiat. Oncol. Biol. Phys. 1995;33:925–930
  18. Wilson LD, Ennis R, Percarpio B, et al. Location of the prostatic apex and its relationship to the ischial tuberosities. Int. J. Radiat. Oncol. Biol. Phys. 1994;29:1133–1138
  19. Wachter S, Wachter-Gerstner N, Bock T, et al. Interobserver comparison of CT and MRI-based prostate apex definition. Strahlenther. Onkol. 2002;178:263–268
  20. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;8:307–310

PII: S0958-3947(10)00011-7

doi:10.1016/j.meddos.2010.01.002

Medical Dosimetry
Volume 36, Issue 1 , Pages 85-90, Spring 2011

Article Tools

* Image Usage & Resolution