Knowledge-based segmentation of pediatric kidneys in CT for measurement of parenchymal volume

Effect of various environments and computed tomography scanning parameters on renal volume measurements in vitro: A phantom study

Kidney volume is an important parameter in clinical practice, and accurate assessment of kidney volume is vital. The aim of the present study was to evaluate the effect of various environments, tube voltages, tube currents and slice thicknesses on the accuracy of a novel segmentation software in determining renal volume on computed tomography (CT) images. The volumes of potatoes and porcine kidneys were measured on CT images and compared with the actual volumes, which were determined by a water displacement method. CT scans were performed under various situations, including different environments (air or oil); tube voltages/tube currents (80 kVp/200 mAs, 120 kVp/200 mAs, 120 kVp/100 mAs); and reconstructed slice thicknesses (0.75 or 1.5 mm). Percentage errors (PEs) relative to the reference standards were calculated. In addition, attenuation and image noise under different CT scanning parameters were compared. Student's t-test was also used to analyze the effect of various conditions on image quality and volume measurements. The results indicated that the volumes measured in oil were closer to the actual volumes (P<0.05). Furthermore, attenuation and image noise significantly increased when using a tube voltage of 80 kVp, while the mean PEs between 120 and 80 kVp voltages were not significantly different. The mean PEs were greater when using a tube current of 100 mAs compared with a current of 200 mAs (P<0.05). In addition, the volumes measured on 1.5 mm slice thickness were closer to the actual volumes (P<0.05). In conclusion, different environments, tube currents and slice thicknesses may affect the volume measurements. In the present study, the most accurate volume measurements were obtained at 120 kVp/200 mAs and a slice thickness of 1.5 mm.

CT-based renal volume measurements: correlation with renal function in patients with renal tumours

AIM

To evaluate the correlations between renal cortical volume (RCV), renal parenchymal volume (RPV), and renal function in patients with renal tumours before and after laparoscopic partial nephrectomy (LPN).

MATERIALS AND METHODS

Thirty-five patients with a single unilateral renal tumour who had undergone contrast-enhanced computed tomography (CT) and renal nuclear scintigraphy before and after LPN were retrospectively studied. RCV and RPV were calculated as renal volume, excluding tumours or cysts, using a semi-automatic segmentation program. The correlations between RCV, RPV, and glomerular filtration rate (GFR) were undertaken preoperatively and postoperatively using the Pearson correlation coefficient.

RESULTS

Preoperatively, the correlations between RCV and GFR, and RPV and GFR for the operated kidneys was r=0.502 (p=0.002) and 0.527 (p=0.001), respectively, whereas the correlations for the contralateral side were r=0.384 (p=0.023) and r=0.412 (p=0.014). The mean RCV and RPV of the operated kidneys decreased by 27.4% and 24.8%. The mean split GFR of the operated kidneys decreased by 36.4%. Postoperatively, residual RCV (r=0.619, p<0.001) and RPV (r=0.593, p<0.001) correlated moderately with the GFR of the operated kidneys.

CONCLUSIONS

Renal volume, both RCV and RPV, had a moderate relationship with renal function before and after operation. CT-based renal volume measurements could serve as a simple and effective method for estimation of postoperative renal function.

Are renal volumes measured by magnetic resonance imaging and three-dimensional ultrasound in the term neonate comparable?

Renal volume, but not renal length, has been shown to be positively correlated with renal function. Three-dimensional (3D) ultrasound and magnetic resonance imaging (MRI) are two modalities used to assess renal volume. The aim of our study was to determine whether 3D ultrasound measurements of renal volume in the neonate are comparable to those of MRI measurements. Preterm and term neonates had an MRI and 3D ultrasound to determine renal volume at the same time as they had an MRI brain scan for other clinical conditions. The preterm neonates were all term corrected age, and the term neonates were 1-4 weeks of age. None of the kidneys examined were abnormal. There were no significant differences in the weight or length of the preterm and term infants at the time of their MRI scan. The left renal length was significantly longer according to MRI measurements than according to 3D ultrasound measurements (p=0.02). Renal volumes of both the left and right kidney were greater when measured by MRI than by 3D ultrasound (p<0.0001, respectively). Total volumes of the kidneys were greater when measured by MRI than by 3D ultrasound (p=0.008). Renal volume in neonates was significantly less when evaluated by 3D ultrasound than by MRI. These results demonstrate that MRI and 3D ultrasound renal volumes are not comparable in the neonatal population and, therefore, the same radiological modality should be used if repeat volume measurements are to be performed.

Dynamic-threshold level set method for volumetry of porcine kidney in CT images in vivo and ex vivo assessment of the accuracy of volume measurement

RATIONALE AND OBJECTIVE

We sought to assess the accuracy of a novel computerized volumetry method, called dynamic-thresholding (DT) level set, in determining the renal volume of pigs in CT images on the basis of in vivo and ex vivo reference standards.

METHODS AND MATERIALS

Eight Yorkshire breed anesthetized pigs (weight range 45-50 kg) were scanned on a 64-slice multidetector CT scanner (Sensation 64; Siemens) after injection of an iodinated (300 mg I/ml) contrast agent through an IV cannula. The kidneys of the pigs were then surgically resected and scanned by CT in the same manner. Both in vivo and ex vivo CT images were subjected to our computerized volumetry using DT level set method. The resulting volumes of the kidneys were compared with in vivo and ex vivo reference standards: the former was established by manual contouring of the kidneys on the CT images by an experienced radiologist, and the latter was established as the water displacement volume of the resected kidney.

RESULTS

The comparisons of the in vivo and ex vivo measurements by our volumetric scheme with the associated reference standards yielded a mean difference of 1.73 +/- 1.24% and 3.38 +/- 2.51%, respectively. The correlation coefficients were 0.981 and 0.973 for in vivo and ex vivo comparisons, respectively. The mean difference between in vivo and ex vivo reference standards was 5.79 +/- 4.26%, and the correlation coefficient between the two standards was 0.760.

CONCLUSION

Our computerized volumetry using the DT level set method can provide accurate in vivo and ex vivo measurements of kidney volume, despite a large difference between the two reference standards. This technique can be employed in human subjects for the determination of renal volume for preoperative surgical planning and assessment of oncology treatment.

The effect of lung volume on nodule size on CT

RATIONALE AND OBJECTIVES

We sought to determine how measures of nodule diameter and volume on computed tomography (CT) vary with changes in inspiratory level.

MATERIALS AND METHODS

CT scans were performed with inspiration suspended at total lung capacity (TLC) and then at residual volume (RV) in 41 subjects, in whom 75 indeterminate lung nodules were detected. A fully automated contouring program was used to segment the lungs; followed by segmentation of all nodules and the corresponding lobe using semiautomated contouring in both TLC and RV scans. The percent changes in lung and lobar volumes between TLC and RV were correlated with percent changes in nodule diameters and volumes.

RESULTS

Both nodule diameter and volume varied nonuniformly from TLC to RV-some nodules decreased in size, while others increased. There was a 16.8% mean change in absolute volume across all nodules. Stratified by size, the mean value of the absolute percent volume changes for nodules > or =5 mm and <5 mm were not significantly different (P = .26). Stratified by maximum attenuation, the mean value of the absolute percent volume changes between the TLC and RV series for noncalcified (17.7%, SD = 13.1) and completely calcified nodules (8.6% SD = 5.7) were significantly different (P < .05).

CONCLUSION

Significant differences in nodule size were measured between TLC and RV scans. This has important implications for standardizing acquisition protocols in any setting where size and, more important, size change are being used for purposes of lung cancer staging, nodule characterization, or treatment response assessment.

Performance of an automated segmentation algorithm for 3D MR renography

The accuracy and precision of an automated graph-cuts (GC) segmentation technique for dynamic contrast-enhanced (DCE) 3D MR renography (MRR) was analyzed using 18 simulated and 22 clinical datasets. For clinical data, the error was 7.2 +/- 6.1 cm(3) for the cortex and 6.5 +/- 4.6 cm(3) for the medulla. The precision of segmentation was 7.1 +/- 4.2 cm(3) for the cortex and 7.2 +/- 2.4 cm(3) for the medulla. Compartmental modeling of kidney function in 22 kidneys yielded a renal plasma flow (RPF) error of 7.5% +/- 4.5% and single-kidney GFR error of 13.5% +/- 8.8%. The precision was 9.7% +/- 6.4% for RPF and 14.8% +/- 11.9% for GFR. It took 21 min to segment one kidney using GC, compared to 2.5 hr for manual segmentation. The accuracy and precision in RPF and GFR appear acceptable for clinical use. With expedited image processing, DCE 3D MRR has the potential to expand our knowledge of renal function in individual kidneys and to help diagnose renal insufficiency in a safe and noninvasive manner.

A framework for evaluating image segmentation algorithms

The purpose of this paper is to describe a framework for evaluating image segmentation algorithms. Image segmentation consists of object recognition and delineation. For evaluating segmentation methods, three factors-precision (reliability), accuracy (validity), and efficiency (viability)-need to be considered for both recognition and delineation. To assess precision, we need to choose a figure of merit, repeat segmentation considering all sources of variation, and determine variations in figure of merit via statistical analysis. It is impossible usually to establish true segmentation. Hence, to assess accuracy, we need to choose a surrogate of true segmentation and proceed as for precision. In determining accuracy, it may be important to consider different 'landmark' areas of the structure to be segmented depending on the application. To assess efficiency, both the computational and the user time required for algorithm training and for algorithm execution should be measured and analyzed. Precision, accuracy, and efficiency factors have an influence on one another. It is difficult to improve one factor without affecting others. Segmentation methods must be compared based on all three factors, as illustrated in an example wherein two methods are compared in a particular application domain. The weight given to each factor depends on application.

Personal computer software evaluation in interactive generation of pig liver three-dimensional anatomical images

The efficiency of software for a personal computer in the interactive generation of three-dimensional (3D) images from computer tomography was studied in six pig livers after hepatic resection and catheterization of the hepatic and portal veins. After perfusion the livers were submitted to computed tomography angiography, volumetric measurement by water displacement, and production of an acrylic model of the veins by the injection and corrosion method, by which the lengths of the hepatic and portal veins were measured. From the angiogram, the software generated a 3D image that allowed measurement of the vein lengths. The identified branches of the hepatic and portal veins were correlated with the hepatic sectors and segments, respectively. The virtual measures from the 3D images were compared with the real measures. There were no significant differences between the topography and the vessel length. The mean difference between the volumes calculated from software and those measured by water displacement corresponded to 1.2%, and between the vessel lengths, 0.2%. In conclusion, the software for personal computer (named LIVER3D) is efficient, allowing interactive inspection of 3D images. All virtual measurements of liver vessel length and partial/total liver volume were similar to the actual ones.

Evaluation of renal growth by magnetic resonance imaging and computerized tomography volumes

PURPOSE

Magnetic resonance imaging (MRI) and computerized tomography (CT) are commonly used to image complex medical conditions but limited data have been reported concerning normal renal volumetric measurement with these imaging techniques. We examined whether normative renal growth curves could be constructed from data derived from these imaging modalities, and from these curves assessed normal and abnormal renal development.

MATERIALS AND METHODS

Patients who had undergone prior renal MRI or CT were identified. Total renal volume and renal cortical fraction (CF, cortical/total volume) were calculated, and growth curves were derived. To examine the curve utility for abnormal growth assessment, renal ultrasonography of children with reflux nephropathy was examined, and MRI and radionuclide scans were compared.

RESULTS

A total of 60 patients 2 months to 39 years old who underwent MRI were included in the growth curve. The CF of the 120 kidneys was 75.8 +/- 4.3% and independent of sex and age. In 19 patients with vesicoureteral reflux 13 kidneys had cortical scarring, and the CF was decreased (p <0.001, 63.65 +/- 5.72%), indicating disproportionate cortical loss. No difference between CF for normal and vesicoureteral reflux unscarred kidneys was found. Differential renal function on radionuclide study correlated highly with MRI renal volume (r = 0.91). CT was performed in 70 children 1 to 15 years old (mean age 7.9) volume correlated with age and renal length, and the left kidney was larger than right kidney on MRI and CT.

CONCLUSIONS

Normative renal growth curves can be constructed from CT and MRI derived renal volumes. Cortical fraction is consistent, and sex and age independent. In reflux nephropathy the CF is reduced and renal differential function on nuclear scan correlates with MRI derived differential volume. This concept may be useful for predicting abnormal renal growth and differential function.

Medical image segmentation with knowledge-guided robust active contours

Medical image segmentation techniques typically require some form of expert human supervision to provide accurate and consistent identification of anatomic structures of interest. A novel segmentation technique was developed that combines a knowledge-based segmentation system with a sophisticated active contour model. This approach exploits the guidance of a higher-level process to robustly perform the segmentation of various anatomic structures. The user need not provide initial contour placement, and the high-level process carries out the required parameter optimization automatically. Knowledge about the anatomic structures to be segmented is defined statistically in terms of probability density functions of parameters such as location, size, and image intensity (eg, computed tomographic [CT] attenuation value). Preliminary results suggest that the performance of the algorithm at chest and abdominal CT is comparable to that of more traditional segmentation techniques like region growing and morphologic operators. In some cases, the active contour-based technique may outperform standard segmentation methods due to its capacity to fully enforce the available a priori knowledge concerning the anatomic structure of interest. The active contour algorithm is particularly suitable for integration with high-level image understanding frameworks, providing a robust and easily controlled low-level segmentation tool. Further study is required to determine whether the proposed algorithm is indeed capable of providing consistently superior segmentation.