External validation of automated CT volumetry in renal tumor volume measurement and its surgical implications

Abstract

Background

Accurate renal tumor anatomy description is crucial for improving surgical and oncological outcomes. CT advancements assist in surgical planning by determining renal tumor volume.

Objectives

To examine the accuracy and precision of CT in renal tumor volume measurement and to assess the impact of tumor volume on intra- and post-operative outcomes.

Methods

Data from 45 patients with renal mass who underwent partial or radical nephrectomy at our institution between December 2019 and January 2024 were analyzed. Automated CT volumetry software was used to detect and measure tumor volume from the pre-operative CT DICOM files. The volume of the retrieved specimens was re-measured using the water immersion method.

Results

The agreement between CT and the water immersion method was tested using Lin's concordance correlation coefficient (CCC), which yielded a value of 0.856. The bias correction factors (C_b) and Pearson's correlation coefficient (ρ) were 0.928 and 0.922, respectively. CCC was better for the partial nephrectomy group (0.943) than the radical nephrectomy group (0.799). The Bland-Altman plot indicated that CT overestimated large-sized tumors. A positive correlation was detected between tumor volume and each of the PADUA score (ρ = 0.494, p value = 0.001) and ischemia time (ρ = 0.048, p value = 0.821).

Conclusions

CT volumetry is an accurate tool for renal tumor volume measurement that could serve as a valuable parameter for optimizing surgical decisions.

Introduction

Renal tumors represent a serious health problem that is considered the 14th most common cancer worldwide, with more than 434,840 new cases (4.4% of all new cancers) and 155,953 deaths (1.5% of all cancer mortality). ¹ As the incidence of renal cell carcinoma (RCC) is increasing owing to the wide use of modern imaging techniques such as computed tomography (CT) and magnetic resonance imaging (MRI), various new management protocols have emerged, such as tumor ablation or partial nephrectomy (PN). These techniques require a good description of tumor criteria to reach optimal surgical and oncological outcomes. ²

The kidney, unlike other organs, doesn't have a fixed segmental anatomy and cannot be classified based on constant vascular or collecting system features. Therefore, the description of tumor complexity cannot be measured using one constant feature of the tumor anatomy. Recent recommendations did not even consider the renowned TNM staging system as the sole criterion for clinical decision-making. ³ Various descriptive strategies have been developed to describe the tumor's relation to kidney tissue, incorporating multiple parameters combined into different nephrometry scores. These protocols have yet to furnish reliable features that guarantee optimal surgery with minimal intra- or postoperative complications. Thus, it appears imperative to augment these scoring systems with additional descriptive features to enhance tumor characterization. ⁴

The renal tumor itself has traditionally been staged based on its maximal dimensions. However, this approach assumes uniform spherical shape and growth in all directions, which is unrealistic. Therefore, there is growing interest in using tumor volume rather than size as a predictor of prognosis. This shift allows for a better representation of tumor anatomy. Reliance solely on size, which can vary among different radiologists, may lead to discrepancies in tumor staging, potentially influencing treatment decisions and prognosis. ⁵

In the initial stage, it is essential to assess the accuracy and precision of CT in discriminating tumors from normal tissue and measuring their volume. The majority of published studies on CT volumetry primarily focused on comparing the pre- and post-operative CT measurements.^6,7 In contrast, our study takes a different approach by comparing CT-derived measurements with a ground-truth method: the water immersion technique. This comparison allows us to assess the reliability of automated CT in measuring renal tumors. Additionally, we aim to evaluate how these CT measurements influence both intraoperative and postoperative outcomes.

Materials and methods

We enrolled 45 patients who presented with renal mass, diagnosed by contrast CT, in our hospital from December 2019 to January 2024.

The estimated glomerular filtration rate (eGFR) was measured preoperatively, on postoperative day one, and approximately 6 months after surgery according to the Modification of Diet in Renal Disease (MDRD) formula, ⁸ except for children for whom this formula could not be applied. Tumor complexity was measured based on the Preoperative Aspect and Dimension Used for an Anatomical Classification (PADUA) score to help in determining the surgical decision. ⁹ Operative complications were classified according to the Uro-Clavien-Dindo system. ¹⁰

Tumor histopathology and the surgical margin for the malignant lesions were recorded, while pathological TNM staging (pTNM) ¹¹ and grading according to the WHO/ISUP histological grading system were noted only for applicable tumor types. ¹²

Automated tumor volume measurement

Contrast CT was done through the following scanning parameters: 120 kVp, 280 mAs, and a gantry rotation speed of 0.75 s/revolution. The collimation used was 64×6.25 mm, with a table speed of 5 mm per revolution, an image reconstruction interval of 0.8 mm, and a slice thickness of 1 mm.

The automated tumor volume measurement was conducted through a 3 Dimensional (3D) workstation, specifically a PHILIPS Extended Brilliance V4.5.450030-EBW: V4.5.4.50015 system. Through this system, the radiologist recalls all CT data and sums contributions from each voxel along a line from any viewing angle through the dataset. Each voxel is assigned opacity, expressed as a percentage from 0% to 100%, which determines its contribution to the sum. The use of this volume rendering technique incorporates the entirety of the dataset, does not require preliminary editing, and enhances visualization by allowing for the selective removal of overlying structures, thereby providing detailed views of the tumor tissue, as shown in (Figure 1).

Figure 1.

Automated computed tomography (CT) method for renal tumor volume measurement. a contrast CT of right midpole renal mass, b automated CT method for segmentation of tumor, c Digital subtraction of tumor with 3 dimensional tumor volume.

In vitro tumor volume measurement

In the radical nephrectomy (RN) patient, the retrieved specimen was dissected to separate the tumor tissue from normal parenchymal tissue, while in the PN group, the retrieved specimen required no additional manipulations as we used the tumor enucleation technique, which removes only the tumor without safety margin. These specimens were immersed in a graded jar partially full of water, and then the volume of displaced fluid was measured by subtracting the pre- and post-immersion fluid levels (Figure 2).

Figure 2.

Water immersion method for renal tumor volume measurement. a graded jar partially filled with water up to 500 mL, b specimen of renal tumor after enucleation, c graded jar after immersion of the tumor the fluid level reach 700 mL which mean that the tumor volume is 200 mL.

Statistical analysis

Intermethod agreement analyses were done using Lin's concordance correlation coefficient (CCC) and its components, namely, the bias correction factor (C_b) indicating accuracy (systematic error) and Pearson's correlation coefficient (ρ) indicating precision (random noise). The corresponding Bland-Altman limits-of-agreement plot was used to further investigate the agreement for different zones of the measured quantity. Paired replicate analyses were done by Wilcoxon's signed-rank sum test. Before-and-after dot-and-line diagrams were used to illustrate the location shift corresponding to selected tests. The association between two quantitative variables was analyzed by Spearman's rank correlation. Two-tailed p values less than the traditional 0.05 were deemed statistically significant. Pairwise deletion was used to deal with missing data. Additionally, some analyses were conducted for certain subgroups. Therefore, the sample size varied between different analyses. Accordingly, the sample size for each analysis was reported alongside. The statistical analysis was performed using MedCalc^® 15.8 (MedCalc Software, Ostend, Belgium).

Ethical considerations:

The study protocol was approved by the Ethical Review Board of our institution (no. 17200472) and was registered in clinical trials registry (no. NCT04244136). All patients signed a written informed consent for PN or RN.

Results

Forty-five patients met the inclusion criteria of our study; PN was done in 25 patients, while RN was done in 20 patients. Of these patients, 27 were male and 18 were female. Apart from two children aged six and seven years, respectively, the age range of the patients was from 33 to 73 years. Seven patients reported hematuria, 21 reported pain, and the remainder was accidentally discovered. All procedures were conducted using the open technique, employing either flank supracostal or anterior subcostal approaches. In the PN group, zero ischemia was utilized in 4 cases, while warm ischemia was employed in the remaining cases, with a median (range) of 15 (5–30) minutes. The median (range) pre-operative eGFR was 86.9 (46.0–186.9) mL/min/1.73 m², and the immediate post-operative eGFR was 66.7 (38.4–150.6), while the late eGFR was 71.7 (45.4–240.6). Complications of various grades occurred in 9 (20%) of cases. Further details regarding patient characteristics are presented in (Table 1).

Table 1.

Patient characteristics.

Variable	n a	Value
Median age in year (IQR)	45	48 (40–60.25)
Gender (%)	45
Male		27 (60)
Female		18 (40)
Diabetes mellitus n (%)	45	5 (11.1)
Hypertension n (%)	45	13 (28.9)
Complaint n (%)	45
Pain		21 (46.7)
Accidentally discovered		17 (37.8)
Hematuria		7 (15.6)
Surgery n (%)	45
PN		25 (55.6)
RN		20 (44.4)
Ischemia type in PN group n (%)	25
Warm		21 (84)
Zero		4 (16)
Median ischemia time in minutes (IQR) in PN group	25	15 (7.8–17.8)
Median operative time in hours (IQR)	45	3 (3–4)
Median hemoglobin drop in gm/dL (IQR)	43	1.1 (0.175–2)
Median blood transfusion units (IQR)	45	1 (0–1)
Median hospital stay in days (IQR)	45	4 (3–4)
Median eGFR by MDRD formula in mL/min/1.73m2 (IQR)
Pre-operative eGFR	37	86.9 (72.9–104.2)
Immediate post-operative eGFR	36	66.7 (58.0–85.7)
Late post-operative eGFR	43	71.7 (56.5–83.8)
Median immediate postoperative eGFR drop by mL/min/1.73m2 (IQR)	36	11.2 (−1.375–35.9)
Median late postoperative eGFR drop by in mL/min/1.73m2 (IQR)	33	13.8 (0.96–25.1)
Complications according to The Uro-Clavien—Dindo classification system n (%)	45
No		36 (80)
I		1 (2.2)
II		2 (4.4)
IIi		1 (2.2)
IIIbi		4 (8.9)
IIIbid		1 (2.2)

^a

number of patients on which the analysis was done.

PN = Partial Nephrectomy; RN = Radical Nephrectomy; eGFR = estimated glomerular filtration rate; MDRD = Modification of Diet in Renal Disease; IQR = interquartile range.

Tumor surgical complexity was assessed using the PADUA score. Within the PN group, three cases were categorized as having low complexity, 10 as having medium complexity, and 12 as having high complexity. Conversely, in the RN group, three cases exhibited medium complexity, while 17 cases were classified as highly complex. Histopathological analysis revealed the presence of RCC in 37 cases, while 4 cases had benign pathology. All cases had a negative surgical margin, and only three cases in the RN group experienced a distant recurrence during follow-up. A summary of tumor characteristics is provided in (Table 2).

Table 2.

Tumor characteristics.

Variable	n a	Value
Laterality n (%)	45
Right		24 (53.3)
Left		19 (42.2)
Bilateral (operated left side only)		2 (4.4)
Clinical T stage n (%)	45
T1		24 (53.3)
T2		20 (44.4)
T3		1 (2.2)
Median PADUA score (IQR)	45	11 (9–12)
Low complexity n (%)		3 (6.7)
Medium complexity n (%)		13 (28.9)
High complexity n (%)		29 (64.4)
Median CT volume of the tumor in cm3 (IQR)	45
The whole cohort		100.9 (53.75–344.5)
PN group		58 (36–108.4)
RN group		304.8 (135–609)
Median immersion volume of the tumor in cm3 (IQR)	45
The whole cohort		100 (50–325)
PN group		50 (45–200)
RN group		275 (90–550)
Tumor histopathology n (%)	45
RCC or other malignancy		41 (91.1)
Benign kidney pathology		4 (8.9)
Pathological T stage n (%)	26
T1		16 (61.5)
T2		8 (30.8)
T3		2 (7.7)
WHO/ISUP grade of applicable cases n (%)	40
1		10 (33.3)
2		17 (56.7)
3		3 (10)

^a

number of patients on which the analysis was done.

RCC = renal cell carcinoma; WHO/ISUP = world health organization /international society of urologic pathology; PADUA = preoperative aspect and dimension used for an anatomical classification; CT = computed tomography; RN = radical nephrectomy; PN = partial nephrectomy; IQR = interquartile range.

The median (range) of the CT volume of the tumor was 100.9 (5–1703) cm³, while the same measurement by graded jar was 100 (5–1200) cm³, but in general, the jar tumor volumes were not significantly less than their CT-measured counterparts (p = 0.056). Nonetheless, when the analysis was stratified, only in the RN subgroup were the immersion-measured volumes significantly less than their CT-measured counterparts (p = 0.0095) (Table 3, Figure 3).

Figure 3.

Before-and-after dot-and-line diagram comparing between preoperative computed tomography and postoperative immersion for measurement of tumor volume; p value by Wilcoxon's signed-rank sum test. a for the whole cohort. b for partial nephrectomy group. c for radical nephrectomy group.

Table 3.

Difference between CT and immersion for measuring tumor volume.

Samples	n a	Median	95% confidence interval	Interquartile range	Range	Count of positive differences	Count of negative differences	p value b
Pre-operative CT volume of tumor (cm3)		100.9	59.8—221	53.8—344.5	5—1703
Postoperative jar volume of tumor (cm3)		100	50—227	50—325	5—1200
(Whole cohort)	45					16	27	0.056
Pre-operative CT volume of tumor (cm3)		304.8	147.8—605.5	135—609	20—1703
Postoperative jar volume of tumor (cm3)		275	108.5—483	90—550	20—1200
(Radical nephrectomy)	20					5	14	0.0095
Pre-operative CT volume of tumor (cm3)		58	40.6—99	36—108.4	5—717
Postoperative jar volume of tumor (cm3)		50	50—97.1	45—200	5—800
(Partial nephrectomy)	25					11	13	0.99

^a

Size of the relevant subgroup, which is indicated between parentheses for each analysis in the first column.

^b

p value by the Wilcoxon's signed-rank sum test; values less than 0.05 are indicated in bold and underlined.

CT = computed tomography.

The agreement between the two measurement methods was evaluated using CCC yielding a value of 0.856 with C_b of 0.928 and ρ of 0.922. Upon further stratification, CCC was better for the PN group (0.943) than for the RN group (0.799), mainly due to the nearly perfect Cb (0.9998) in the former group compared to only (0.880) in the latter group (Table 4, Figure 4). The Bland-Altman plot indicates overestimation of the volume by CT for large-sized tumors (Figure 5).

Figure 4.

Scatter plot of preoperative and postoperative tumor volume as measured by computed tomography and immersion, respectively; CCC = lin's concordance correlation coefficient; C_b = bias correction factor (systematic error component); ρ = Pearson's correlation coefficient (random noise component); dashed line = line of equality; solid line = reduced major axis line. a for the whole cohort. b for partial nephrectomy group. c for radical nephrectomy group.

Figure 5.

Bland-Altman plot for agreement between preoperative and postoperative tumor volume as measured by computed tomography and immersion, respectively; short-dashed line = line of equality; solid line = mean of differences; long-dashed lines = limits of agreement; sd = standard deviation. a for the whole cohort. b for partial nephrectomy group. c for radical nephrectomy group.

Table 4.

Summary of statistics of intermethod agreement analyses.

Methods	n a	Lin's concordance correlation				Bland-Altman limits of agreement analysis
		CCC		Components b		Mean of differences			SD	Lower limit		Upper limit
			95% CI	Cb	ρ		95% CI	p c			95% CI		95% CI
Pre-operative CT volume of tumorPost-operative jar volume of tumor(Whole cohort)	45	0.856	0.784—0.905	0.928	0.922	53.8	0.6—107.1	0.048	177.3	−293.7	−385.4 — −202	401.4	309.6—493.1
Pre-operative CT volume of tumorPost-operative jar volume of tumor(Radical nephrectomy)	20	0.799	0.641—0.892	0.88	0.908	124.8	11.1—238.4	0.033	242.8	−351.2	−548.7 — −153.6	600.7	403.2—798.3
Pre-operative CT volume of tumorPost-operative jar volume of tumor(Partial nephrectomy)	25	0.943	0.875—0.974	1	0.943	−2.9	−27—21.2	0.8	58.3	−117.3	−159 — −75.5	111.4	69.7—153.1

^a

Size of the relevant subgroup, which is indicated between parentheses for each analysis in the first column.

^b

C_b = bias correction factor (systematic error component, a measure of accuracy); ρ = Pearson's correlation coefficient (random noise component, a measure of precision).

^c

p value for the null hypothesis of a zero mean; values less than 0.05 are indicated in bold and underlined.

CCC = concordance correlation coefficient; CI = confidence interval; CT = computed tomography; SD = standard deviation of the differences.

Our analysis revealed a significant positive correlation between CT tumor volume and the PADUA score (Spearman's rho = 0.494, p value = 0.0006), suggesting a possible association between tumor size and surgical complexity. The effect of CT tumor volume on the preoperative eGFR showed that there was a negative correlation (Spearman's rho = −0.04, p value = 0.81). Also, when examining the impact of CT tumor volume on the immediate drop in eGFR postoperatively, there was a positive correlation (Spearman's rho = 0.121, p = 0.48).

Additionally, we observed a positive correlation between tumor volume and ischemia time (Spearman's rho = 0.048, P value = 0.82), suggesting that larger tumors may necessitate longer periods of dissection during surgery. Furthermore, a weak positive correlation was presented between tumor volume and the decrease in hemoglobin levels postoperatively (Spearman's rho = 0.0685, P = 0.6626), indicating a potential association between tumor size and intraoperative blood loss.

Discussion

Renal tumors come in many different and often unpredictable shapes, leading to a constant gap between the anatomical imaging and the actual tumor appearance, which makes their description based on width and length from subjective clinical assessment or imaging studies an inaccurate and unreliable estimation. ¹³ Our study evaluated the rule of automated CT volumetry in providing a proper evaluation of these tumors and how these measured volumes could affect renal tumor surgery.

In a study conducted by Tann et al., helical CT imaging was employed to examine twelve patients with renal tumors. Two radiologists utilized the summation of area approach to evaluate the tumor volume on the CT scan and subsequently compared these findings with the water displacement volume of the specimen. A notable overestimation was detected between the mean CT measurements of the two radiologists and the water displacement volume, amounting to 48 cc, with a percentage difference in volume of 52.87%. ¹⁴ In our study, we observed good agreement between CT and immersion volume measurements; the CCC value was 0.856.

Chen et al. utilized automated 3D segmentation software to determine tumor volumes in twenty-seven patients who underwent RN or PN. They compared these measurements with the volumes estimated using the conventional ellipsoid equation, based on the parameters from radiologic imaging and pathologic specimens. Their findings indicated that 3D segmentation provides a rapid and accessible method for volume calculation, differing significantly from traditional ellipsoid equation calculations. Moreover, the 3D segmentation method correlated more closely with histopathological volume. However, their analysis revealed an increase in discrepancy with the pathologic method as tumor size increased for both measurement approaches. ¹⁵ In another study by Choic et al., comparing the radiologic and pathologic tumor volumes, they noted only accurate estimation in small tumors, but as the tumor size increased more than 7 cm, the radiologic volume showed overestimation. ¹⁶ This observation aligns with our study, as we similarly noted a greater discrepancy in volume measurement for larger tumor sizes, which may be due to improper tumor tissue discrimination.

Another study done by Tiwari et al. tried to determine the complexity of renal masses in PN by using the intraparenchymal tumor volume (IPV) measured by CT. A high IPV score (IPV >27.26 cc) was significantly correlated to the absolute change in creatinine (P = 0.018), the percentage change in creatinine (P = 0.004), and the RENAL score (P < 0.001). ¹⁷ In our investigation, we analyzed the correlation with the whole tumor volume and discovered a significant association with the PADUA score, along with a positive correlation with the postoperative drop in eGFR.

Moreover, tumor volume could provide additional prognostic information. In a study done by Jorns et al. on 955 patients treated with RN or PN for unilateral, sporadic, non-cystic pT1 clear cell RCC, tumor volume was estimated by the ellipsoid formula, and it was found that increasing tumor volume is associated with a greater risk of RCC-specific death. In addition, there was evidence that tumor volume might provide more accurate prognostic information than tumor size alone in pT1a patients. ¹⁸ Another study by Secil et al. retrospectively assessed 46 patients with RCC and found a significant relationship between tumor volume and patient survival; however, it was not an independent parameter. Additionally, there was a relationship between the increased tumor volume and the invasion rates in the perinephric fat, adrenal vein, or renal vein. ¹⁹ In our study, we observed a strong positive correlation between tumor volume and T stage progression (Spearman's rho = 0.908, P = 0.0001). Notably, tumor recurrence occurred in three cases, all of which were associated with large tumor volumes (610 cc, 1120 cc, and 1643 cc). Additionally, we found that two patients who experienced high-grade complications (Grade IIIbi) had tumors with significantly larger volumes (1643 cc and 1703 cc).

The limitation of our study was the small sample size. Future studies with larger cohorts and multicenter collaboration are needed to validate our findings and further explore the impact of tumor volume on surgical outcomes.

In conclusion, our findings highlight the possibility of using renal tumor volume as a prognostic parameter for intraoperative tumor complexity and postoperative renal function. The high agreement between CT and graded jar volume measurements supports the use of these methods in clinical practice, with careful consideration of potential discrepancies in larger tumors. This factor, along with other indicators of tumor complexity, should be combined to reach the ideal surgical and oncological outcome.