A Prospective Study of the Diagnostic Performance of Photon Counting CT Compared to MRI in the Characterization of Renal Masses

Abstract

Objectives:

To assess the inter-reader reliability and per-RCC sensitivity of high-resolution Photon Counting CT (PCCT) in the detection and characterization of renal masses in comparison to MRI.

Materials and Methods:

This prospective study included 24 adult patients (mean age 52±14 years; 14 females) who underwent PCCT (using an investigational whole-body CT scanner) and abdominal MRI within a 3-month time interval and underwent surgical resection (partial or radical nephrectomy) with histopathology (n=70 lesions). Of the 24 patients, 17 had a germline mutation and the remainder were sporadic cases. Two radiologists (R1 and R2) assessed the PCCT and corresponding MRI studies with a 3-week washout period between reviews. Readers recorded the number of lesions in each patient and graded each targeted lesion’s characteristic features, dimensions, and location. Data were analyzed using a two-sample t-test, Fisher’s exact test, and weighted kappa.

Results:

In patients with von-Hippel-Lindau mutation, R1 identified a similar number of lesions suspicious for neoplasm on both modalities (51 vs 50; p=0.94), while R2 identified more suspicious lesions on PCCT scans as compared to MRI studies (80 vs 56, p=0.12). R1 and R2 characterized more lesions as predominantly solid in MRI images (R1:58/70 in MRI vs. 52/70 in PCCT, p<0.001; R2:60/70 in MRI vs. 55/70 in PCCT, p<0.001). R1 and R2 performed similarly in detecting neoplastic lesions on PCCT and MRI studies (R1:94% vs 90%, p=0.5; R2:73% vs. 79%, p=0.13).

Conclusions:

The inter-reader reliability and per-RCC sensitivity of PCCT scans acquired on an investigational whole-body PCCT were comparable to MRI scans in detecting and characterizing renal masses.

Introduction

Photon-counting CT (PCCT) is a promising multi-energy CT modality that addresses many of the limitations of conventional CT, which uses energy-integrating detectors (EIDs) with inherent limits to the spatial resolution and signal-to-noise ratio of an image [1, 2]. Rather than summing the energy-weighted interactions with x-ray photons as with EIDs, photon-counting detectors (PCDs) record individual photons with energy above pre-specified thresholds. Compared to conventional CT, this results in a greater contrast-to-noise ratio, improved spatial resolution, and greater radiation dose efficiency [1, 3]. Inherent image acquisition at two or more energy levels allows the calculation of material-specific concentration maps based on the energy absorption profiles [4].

Initial clinical experience in abdominal imaging using an investigational whole-body clinical PCCT scanner (not approved at the time for clinical use) showed similar performance to EID-CT regarding image quality, noise, and artifacts, while iodine concentration maps had a 32% improvement in the contrast-to-noise ratio [5]. Both ex-vivo data and in-vivo studies in patients with renal stones demonstrated improved detection of small stones (≤3 mm) and characterization of internal composition with PCCT compared to dual-energy CT [6, 7]. The benefit of reduced noise and thus overlap in attenuation values for different components of a renal cyst with PCCT-derived images at discrete energy bins was seen in a renal phantom study which demonstrated a linear increase in attenuation from saline, blood, and iodine [4]. Thus, PCCT may allow for improved elemental analysis and characterization of complex renal masses.

National guidelines recommend either contrast-enhanced CT or MRI to assess renal masses [8, 9]. Our patient population predominantly consists of individuals with familial renal cancer, necessitating lifelong surveillance imaging. In light of various considerations including claustrophobia, patient size limitations, and the need for non-invasive options, MRI has been our preferred choice due to its lack of ionizing radiation. However, MRI is resource-intensive, with the average abdominal MR lasting approximately 45 minutes. Given these constraints, identifying imaging modalities with performance comparable to MRI is essential for surveillance populations. Given the potential benefits of PCCT over conventional EID-CT [10, 11], PCCT may offer performance similar to MRI, while also enabling superior characterization of the internal composition of renal masses through material decomposition analysis, such as iodine. Additional advantages of PCCT include enhanced spatial resolution, the absence of downweighting of low energy photons (yielding a better iodine signal), and the capability to produce multi-energy images. These features suggest that PCCT could provide a comprehensive assessment of renal masses. Therefore, our study aimed to evaluate whether high-resolution PCCT matches MRI in the detection and characterization of renal masses, thus offering a viable alternative for surveillance imaging in patients with familial renal cancer.

Materials and Methods

Approvals and Disclosures

This prospective study was conducted under a protocol approved by the Institutional Review Board and was compliant with Health Insurance Portability and Accountability Act (HIPPA). Written informed consent was obtained from each research participant prior to enrollment in this protocol (NCT00026884).

Patient selection

In this study, we specifically targeted a unique patient cohort from [institution], recognized as a tertiary referral center with extensive expertise in renal neoplasia and cancer syndromes. We recruited 30 consecutive adult patients presenting with renal masses, all of whom consented to participate in this study. These patients underwent PCCT scanning at [institution] between January and June of 2021. Only patients who had an MRI scan within 3 months of their PCCT study and underwent surgical resection of the lesion(s) for histopathologic confirmation were included (Figure 1). Six patients were excluded as they did not have an MRI (n=1) or did not undergo partial/radical nephrectomy (n=5). The latter group was used for a pilot study to optimize inter-reader reliability. The final study comprised 24 patients (mean age 52 ± 14 years; 14 female and 10 male) with 70 renal lesions with pathologic diagnosis (Table 1).

Figure 1.

Flow diagram of included patients in both full study and pilot study. (PCCT: photon-counting computed tomography)

Table 1.

Patients’ demographic data (VHL: Von Hippel-Lindau syndrome; BHD: Birt-Hogg-Dubé syndrome; HLRCC: hereditary leiomyomatosis and renal cell cancer; TSC: tuberous sclerosis complex; RCC: renal cell carcinoma).

	Total	VHL	Sporadic	BHD	HLRCC	TSC
Number of patients	24	12	7	2	2	1
Number of lesions	70	35	21	7	3	4
Age (Mean ± SD)	52 ± 14	49 ± 12	58 ± 11	58 ± 15	43 ± 24	55
Gender (F/M)	14/10	8/4	3/4	2/0	0/2	1/0
Number of prior ipsilateral partial nephrectomies	12	8	0	1	2	1
Clear cell RCC	49	35	14	0	0	0
Chromophobe RCC	5	0	1	0	0	4
Hybrid	8	0	4	4	0	0
Oncocytoma	3	0	0	3	0	0
Angiomyolipoma	1	0	1	0	0	0
HLRCC	3	0	0	0	3	0
Reninoma	1	0	1	0	0	0

All the removed tumors were labeled and numbered in the images prior to surgery. The excised tumor was then sent for pathology evaluation, tagged with the identical number used in the imaging, to ensure accurate correlation between the pathological and radiological findings.

Image acquisition protocols

The study used an investigational whole-body PCCT (SOMATOM CounT, Siemens Healthineers), which was based upon a modified dual-source dual-energy scanner (SOMATOM Definition Flash, Siemens Healthineers). The scanner contained a subsystem with an EID and a subsystem with a Cd-Te PCD. Participants in this clinical trial underwent imaging with both subsystems. This study only analyzed data acquired with the PCD subsystem. Under the protocol, a conventional EID study was performed first. Iopamidol (Isovue-370 mg/mL, Bracco Diagnostics) was administered intravenously via antecubital vein (2 mL/kg, up to a maximum of 130 mL) at a rate of 2 mL/sec. Scanning was triggered in the arterial phase using a bolus tracking technique (Care Bolus, Siemens Healthineers) at 20 seconds with the EID system. Portal venous phase images were acquired 80 seconds after the start of the arterial phase. PCCT images were obtained 2.5 minutes after the arterial phase. PCCT was performed with the following parameters: collimation, 64 x 0.5 mm; gantry rotation time, 0.5 sec; pitch, 0.4; z-length, 10 cm; total scan time, 9.8 sec; 140 kV; and 688 mAs. Two energy thresholds were used: one at 25 keV to function as a noise floor and one at 65 keV. Data from these two energy bins was used to generate three image sets: low (25-65 keV: Bin1), high (65-140 keV: Bin2), and full spectrum (25-140 keV: Threshold1-Th1) energy representations.

All patients underwent standard-of-care and clinically indicated MRI examinations (1.5T MAGNETOM Aera; Siemens Healthcare) per our institution’s protocol for renal masses using intravenous administration of 0.1 mmol/kg IV gadobutrol (Gadavist, Bayer HealthCare Pharmaceuticals) followed by a 20 mL saline flush. Imaging sequences included T2-weighted images, in- and out-of-phase, diffusion weighted images with corresponding ADC maps, and pre- and post-contrast T1-weighted images during corticomedullary phase (20-second), nephrographic phase (70-second), and excretory phase (3-minute).

PCCT Reconstruction Techniques and Quantitative Iodine Calculation

Images were reconstructed using filtered back projection (ReconCT, Siemens Healthineers) at a 1.5 mm section thickness and 1.5 mm increment using the D30f kernel. Bin datasets were imported for iodine quantification. A mixed image (0.6 Bin1:0.4 Bin2 ratio) was used as the background image. A region of interest (ROI) was drawn in three axial sections for each lesion, the mid-lesion axial section and the most superior and inferior sections that contained the lesion, to include at least two-thirds of the lesion’s area, from which the mean iodine value was calculated to determine the average iodine concentration (AIC).

Qualitative Image Assessment

Two fellowship-trained abdominal radiologists (XXX and XXX with 14 and 8 years of experience, respectively) reviewed the PCCT and MR images of the patients who did not undergo surgery (5 patients with 8 lesions) to optimize inter-reader reliability by reviewing and discussing discrepancies between the radiologists’ findings and establishing agreement on feature definitions. Then, patients (n = 24) included in the analysis were randomly divided into 2 equal groups (batch 1 and batch 2). Radiologist 1 (R1) assessed PCCT images of batch 1 and MRI images of batch 2 while radiologist 2 (R2) assessed PCCT images of batch 2 and MRI images of batch 1. R1 and R2 then reviewed the respective complementary image sets after a 3-week washout period to prevent recall bias in the interpretation of different scans from each patient (Figure 2).

Figure 2.

Schematic image of the study design. A, diagram illustrates photon counting detector absorbing x-ray photons and generating different level of electrical pulses, assigned to 25 keV and/or 65 keV energy bins. B, reconstructed images for different thresholds (25 keV threshold: 25-140 keV; 65 keV threshold: 65-140 keV) and bins (Bin1: 25-65 keV; Bin2> 65 keV) derived from the photon-counting CT data. C, Workflow representations of the readout sessions for 2 readers, where the 24 patients were randomly assigned to batch 1 or batch 2 for review.

For patients with Von Hippel-Lindau syndrome (VHL), radiologists identified renal lesions larger than 1 cm and suspicious of being a neoplasm. For hereditary leiomyomatosis and renal cell carcinoma (HLRCC), any lesion of any size was identified. For all other patients, all lesions larger than 1 cm were identified. The likelihood of being a neoplasm was scored based on a 5-point Likert scale for each identified lesion in patients with non-VHL mutation (Figure A. Supplementary data). Apart from germline mutation (Figure 3), radiologists were otherwise blinded to patients’ clinical information. For each annotated lesion (70 lesions with histopathology), radiologists answered questions regarding size, location, and internal tumor composition (Table 2). For lesions with cystic components, septations and nodules were defined based on the 2019 Bosniak Classification System [12]. The distance of the lesion from the sinus line and polar line were assessed in the coronal plane. A line was drawn perpendicular to the sinus line and polar line as defined by PADUA and RENAL classification systems [13, 14]. The craniocaudal distance from the mass to each of these lines was denoted as the distance of the lesion from the sinus line and polar line, respectively. Distance to the closest segmental vessel was measured in axial sections as the linear distance of the lesion to the closest identifiable vessel seen in the renal pelvis.

Figure 3.

Axial PCCT iodine images (A, D, G, I, K, M), MRI 3 minutes delay images (B, E, H, J, L, N), and MRI apparent diffusion coefficient (ADC) images (C, F) of patients with different germline mutations. The relevant lesion for each figure is indicated by a green arrow. A-F: A 66-year-old man had 2 renal cell carcinoma lesions consistent with hereditary leiomyomatosis and renal cell cancer. R1 missed identifying one lesion (A-C) in both PCCT and MRI scans, while R2 was indeterminate in calling the same lesion as neoplasm. Both readers were confident (>75%) in calling the second lesion (D-F) as neoplasm. G, H: A predominantly solid clear cell renal carcinoma grade II lesion from a 33-year-old woman with von Hippel-Lindau syndrome. I, J: Complex cystic clear cell renal cell carcinoma of a 67-year-old male with no germline mutation. R1 was quite confident (>75%) in calling this lesion as neoplasm in both modalities, while R2 identified this lesion as a neoplasm with >50% confidence on PCCT and indeterminate on MRI. K, L: A predominantly solid hybrid tumor in a 43-year-old female with Birt-Hogg-Dubé syndrome. M, N: A chromophobe renal cell carcinoma in a 55-year-old female with tuberous sclerosis complex.

Table 2.

Questionnaire for the readout sessions. (VHL: Von Hippel-Lindau syndrome; HLRCC: hereditary leiomyomatosis and renal cell cancer)

Question	Answer Choices
Each case (24 patients)
Quality of image:	Optimal Suboptimal
Number of lesions:	VHL: suspicious lesions ≥ 1cm HLRCC: any lesion Others: any lesions ≥ 1cm
Likelihood of being neoplasm for each identified lesion in non-VHL patients:	Very likely benign (>75% confident) Likely benign (>50% confident) Indeterminate (<50% confident) Likely neoplasm (>50% confident) Very likely neoplasm (>75% confident)
Each Annotated lesion (70 Lesions with pathology)
Maximum dimension of the lesion:	In cm
Location of the tumor:	>50% exophytic <50% exophytic Entirely endophytic
Internal tumor composition:	Predominantly solid (>75% solid) Mixed solid (50-75% solid) Mixed cystic (25-50% solid) Predominantly cystic (<25% solid)
If cystic, type of fluid:	Serous or hypodense fluid Hemorrhagic Indeterminate
If cystic, any septations?	Number of septations No
If cystic, any nodules?	Number of nodules No
Percentage of the lesion that enhances compared to unenhanced phase:	0-25% 25-50% 50-75% 75-100%
Proximity to sinus line:	In mm
Proximity to polar line:	In mm
Proximity to the closest segmental vessel:	In mm

Statistical analysis

Continuous data was compared by the two-sample t test, while categorical proportions were assessed using Fisher’s exact test. Inter-reader and intra-reader reliability were assessed with weighted kappa (κ) [15]. Measurements of tumor dimension and distance to sinus line, polar line, and closest segmental vessel were compared between readers and modalities using Bland-Altman plots [16]. McNemar’s test was used to compare the sensitivities of radiologists between imaging modalities in predicting malignancy. Receiver operating characteristic analysis was used to determine the ability of AIC to predict lesion subtype. All analysis was done using Microsoft Excel (Microsoft Inc) and SPSS statistical software (version 28.0.1.0, IBM). All tests were two-tailed, unless otherwise specified.

A mixed-effects model was employed to evaluate the impact of reader and imaging modality on the assessed likelihood of malignancy of renal masses. This approach was taken to account for both fixed effects (reader and modality) and random effects (variability between patients).

Results

Number of Detected Renal Lesions

In the 24 patients (Table 1), R1 detected a similar number of renal lesions on PCCT scans as compared to MRI studies (143 vs. 137, p=0.9), while R2 noted more renal lesions on PCCT images (174 vs. 147, p=0.5). For VHL patients, R1 identified a similar number of suspicious lesions on both modalities (51 vs 50; p=0.94), while R2 identified more suspicious lesions on PCCT scans as compared to MRI studies (80 vs 56, respectively, p=0.12). For non-VHL patients, both radiologists detected a similar number of lesions on PCCT and MRI images (R1: 92 vs. 87, p=0.89; R2: 94 vs. 91, p=0.94).

Qualitative Analysis of Annotated Lesions (70 lesions with histopathology)

Inter-reader reliability was significant for exophytic nature of tumors, characterization of internal lesion composition (i.e., solid/cystic), and percentage of enhancement of lesion for both PCCT and MRI (K=0.34 – 0.72, p<0.01). For each radiologist, agreement on whether a tumor was exophytic and solid/cystic as well as degree of enhancement was at least moderate (K=0.47 – 0.74, p<0.001). While R1 and R2 defined a similar number of lesions (8-10) as predominantly cystic (<25% solid portion) in both modalities, they categorized more lesions as mixed cystic (25-50% solid portion) in PCCT as compared to MRI images [R1: 6% (4/70) vs 1% (1/70); R2: 4% (3/70) vs 0% (0/70)] (Figure 4). For both radiologists, the largest discrepancies in the characterization of mass composition between imaging modalities were seen in VHL or bilateral multifocal clear cell renal cell carcinoma (ccRCC) patients. The distribution of responses for percent enhancement of lesions mirrored those of internal mass composition.

Figure 4.

A-B: Clustered column plots demonstrating distribution of scored internal composition (A) and percent enhancement of lesions (B) by radiologists across imaging modalities. C: Box and whisker plots showing average iodine concentration (in ROIs of upper, mid, and lower boundaries of lesion as measured on PCCT scan) based on scored internal composition for radiologists across both imaging modalities (ROI: region of interest; PCCT: photon-counting CT; R1: Radiologist 1; R2: Radiologist 2)

Readers’ confidence and accuracy in non-VHL lesion classification

In the 73 lesions that were identified in non-VHL patients by both radiologists across imaging modalities, R1 was equally confident in classifying lesions on PCCT scans (68/73) and MRI studies (63/73) as very likely benign or very likely neoplastic (p=1). In contrast, R2 classified more lesions on MRI images as very likely benign or very likely neoplastic (65/73 for MRI vs. 59/73 for PCCT, p=0.04).

Pathological confirmation was available for 48 lesions identified by the readers in 12 non-VHL patients. Of these, 45 lesions showed neoplastic features in their pathology report, with the remaining 3 being benign (two renal cysts and one specimen showing renal parenchyma). R1 and R2 had similar intra-reader accuracy in detecting neoplastic lesions on PCCT and MRI studies (R1: 94% vs 90%, p=0.5; R2: 73% vs. 79%, p=0.13). Both radiologists failed to identify a hybrid oncocytic tumor in a Birt-Hogg-Dubé syndrome (BHD) patient in both imaging modalities. R1 missed a hereditary leiomyomatosis and renal cell cancer (HLRCC) in both PCCT and MRI scans, while R2 classified the lesion as indeterminate for both imaging modalities. R1 missed 2 sporadic clear cell renal tumors in 2 different patients on MRI images that were identified and classified as neoplasm on PCCT scan. Figure 5 demonstrates the proportion of readers’ confidence in each modality for identification of these lesions.

Figure 5.

A. Clustered column plot demonstrating radiologists’ confidence in classification of annotated lesions across imaging modalities. B. Box and whisker plots showing average iodine concentration (in ROIs of upper, mid, and lower boundaries of lesion as calculated from PCCT images) based on radiologists’ classification of annotated lesions across imaging modalities. Only those non-VHL lesions with known pathological diagnoses were included in these plots (A and B). C. Mean average iodine concentration for sporadic and hereditary kidney cancer syndromes (ROI: region of interest; VHL: Von Hippel-Lindau syndrome; PCCT: photon-counting CT; R1: Radiologist 1; R2: Radiologist 2; BHD: Birt-Hogg-Dubé syndrome; HLRCC: hereditary leiomyomatosis and renal cell cancer; TSC: tuberous sclerosis complex)

Quantitative Analysis of Annotated Lesions

For both readers, there was a positive relationship between AIC and the degree of enhancement as well as the solid component of a lesion in PCCT scans (Figure 4). However, no meaningful relationship was noted between the AIC on PCCT and internal lesion composition or the percentage of lesion enhancement reported on MRI studies. The calculated AIC did not differentiate ccRCC from non-ccRCC lesions (AUC: 0.55, 95% CI: 0.41-0.70) or low grade ccRCC from high grade ccRCC (AUC: 0.57, 95% CI: 0.40-0.74). Across the different hereditary cancer syndromes, AIC was highest in tuberous sclerosis complex (3.4 ± 0.6) and VHL (3.4 ± 1.1) lesions followed by BHD (2.7 ± 0.6) and HLRCC (2.3 ± 0.9), while sporadic lesions had the lowest AIC (2 ± 1.3) (Figure 5). Figure 5 shows that in non-VHL patients R2 identified more lesions as neoplasm with higher AIC and missed or identified lesions as non-neoplastic with low AIC (p=0.001).

Mixed-Effects Model Analysis of Likelihood of Malignancy

baseline likelihood of malignancy (Intercept) was estimated at 4.187 (Std. Err. = 0.414, p < 0.001). When comparing the reader effect, the likelihood of malignancy assessed by R 2 was slightly, but not significantly, lower than that assessed by R 1 (Coef. = −0.458, Std. Err. = 0.406, p = 0.259). Regarding the modality, there was a marginal, non-significant difference in the likelihood of malignancy between Photon CT and MRI (Coef. = 0.125, Std. Err. = 0.382, p = 0.744). Notably, the analysis revealed variability between patient groups (Group Var = 0.737) (Table 3).

Table 3.

Mixed-Effects Model Analysis of the Impact of Reader and Modality on the Likelihood of Renal Mass Malignancy Assessment

Parameter	Coefficient	Standard Error	z-value	P-value	95% Confidence Interval
Intercept	4.187	0.414	10.127	< 0.001	3.377 to 4.998
Reader (R1 vs. R2)	−0.458	0.406	−1.130	0.259	−1.254 to 0.337
Modality (Photon CT vs. MRI)	0.125	0.382	0.327	0.744	−0.624 to 0.874
Group Variance	0.737	-	-	-	-
Group x Reader R1 Covariance	−0.315
Group x Reader R2 Covariance	0.092	-	-	-	-
Reader	0.223	-	-	-	-

Discussion

Here we have provided, to our knowledge, the first comparison of PCCT scan with MRI studies in characterizing renal lesions. Previous studies assessing the utility of PCCT scans in renal imaging have focused on complex renal cysts and/or renal stones. Furthermore, the standard of care comparison for PCCT studies has been EID-CT as opposed to MRI [10,11]. Given the large population of patients with familial renal cancer seen at our institution, MRI is currently our preferred modality for imaging renal masses. However, the trade-off for the well-known advantages of MRI scans in depiction of soft tissue is lengthy study duration (45-60 minutes). Given the putative benefit of PCCT over conventional EID-CT in improving image quality [10,11], this modality may have comparable performance to MRI studies, while allowing for improved characterization of internal composition of renal masses due to material decomposition (i.e., iodine) analysis.

In this study, both radiologists performed similarly with PCCT and MRI studies in identifying basic qualitative and quantitative characteristics associated with renal lesions, such as exophytic/endophytic location, lesion size, and proximity to the sinus line, polar line, and closest segmental vessel. Furthermore, the overall per RCC sensitivity of PCCT was comparable to that of MRI.

Confidence in characterizing lesions was similar between PCCT and MRI scans for R1; however, R2 was more confident in characterizing lesions on MRI. Differences in confidence may be related to differential level of experience of R1 as compared to R2, as R1 was overall more confident in lesion classification than R2. Nevertheless, despite no prior experience with PCCT scans, accounting for decreased confidence, R2 still had similar level of performance in lesion identification and classification between MRI and PCCT studies. It is important to note that MRI did not outperform PCCT in classifying lesions despite the presence of additional sequences unique to MRI, e.g., diffusion weighted imaging, that allow assessment of subtle differences in tumor biology. That being said however, there were few tumors included where there is more reliance on diffusion weighted imaging for classification, such as HLRCC.

More lesions were characterized as mixed solid or cystic on PCCT scans than MRI. This discrepancy could be attributed to the fact that the lesions were only observed during the early excretory phase in PCCT scans. As the degree of contrast decreased, the readers may have interpreted the lesions as having a lower density (cystic). A predominantly cystic lesion on MRI images may thus be classified as mixed cystic on PCCT scans, with the radiologist therefore being more likely to classify it correctly as neoplastic. Determining internal composition of tumor on PCCT is facilitated by quantitative assessment of iodine composition. Given the availability of iodine quantification with PCCT scans, we may expect improved confidence and accuracy in characterizing small, otherwise indeterminate lesions compared to MRI. This is particularly relevant in syndromes such as HLRCC which has an aggressive tumor biology with size-independent metastasis. However, in this study, we had 2 HLRCC patients with 3 renal lesions, making it difficult to draw meaningful conclusions. R1 did miss 1 sub-centimeter HLRCC malignancy on both PCCT and MRI scans, while R2 characterized this lesion as indeterminate on both modalities. Future studies utilizing PCCT images on this patient population are needed.

The differences in AIC among the hereditary kidney cancer syndromes were reflected in differential patterns of enhancement. ccRCCs are highly vascular, with hyperenhancement following iodinated contrast. In contrast, HLRCC is often associated with papillary histology and tends to be comparatively hypoenhancing. Sporadic tumors were a heterogenous group (angiomyolipoma, bilateral multifocal ccRCC, a d juxtaglomerular cell tumor), precluding generalizations about the average ROI within these tumors.

There were limitations to the study. The research protocol required acquisition of the diagnostic EID scan prior to the PCCT study. PCCT images were therefore acquired 2.5 minutes following contrast administration; thus, the iodine maps may not have been optimal for per RCC sensitivity. Acquiring PCCT images earlier (i.e., 20- or 80-seconds post-injection) may allow more robust and consistent predictions of mass composition as well as pathology using iodine mapping. Additionally, the majority of assessed lesions were greater than 1 cm, making it difficult to draw conclusions on the efficacy of PCCT in small lesions. The clinical relevance of neoplastic detection rate in small lesions would apply to renal masses associated with syndromes such as HLRCC, BRCA associated protein 1 (BAP1), and succinate dehydrogenase (SDH).

Finally, while pathologic correlation was included, there was no standard reference measure for tumor composition and distance measurements.

In conclusion, this prospective study on the feasibility and per RCC sensitivity of PCCT scans showed that it was comparable to MRI scans in identifying and qualitatively characterizing a wide range of renal cancer subtypes. Further advantages in PCCT over MRI studies may be realized with optimization of scan timing and imaging parameters along with quantitative approach towards lesion characterization through radiomics.

Figure 6.

Bland-Altman plots showing the mean difference between PCCT and MRI studies in terms of size (cm) and location (mm) measurements of lesions, averaged between radiologists. (PCCT: photon-counting CT; LoA: limits of agreement)

Clinical relevance statement.

PCCT scans have comparable performance to MRI studies while allowing for improved characterization of the internal composition of lesions due to material decomposition analysis. Future generations of this imaging modality may reveal additional advantages of PCCT over MRI.

Key results:

1. PCCT scans were comparable to MRI scans in identifying and qualitatively characterizing renal lesions.

2. Radiologists had a similar level of performance in classifying renal lesions in PCCT and MRI scans, despite the presence of additional sequences allowing tissue characterization in MRI.

3. Both readers had similar intra-reader per-RCC sensitivity in detecting neoplastic renal lesions on PCCT and MRI studies (R1: 94% vs 90%, p=0.5; R2: 73% vs. 79%, p=0.13).

Abbreviations:

AIC:

average iodine concentration

BHD:

Birt-Hogg-Dubé syndrome

ccRCC:

clear cell renal cell carcinoma

EID:

energy integrating detectors

HLRCC:

hereditary leiomyomatosis and renal cell carcinoma

PCCT:

photon-counting computed tomography

ROI:

region of interest

TSC:

tuberous sclerosis complex

VHL:

Von-Hippel-Lindau syndrome