New research suggests that an emerging diagnostic model, which incorporates computed tomography (CT) features, may enhance the detection of clear cell renal cell carcinoma (ccRCC) in small renal masses.
For the multicenter retrospective study, recently published in Academic Radiology, researchers compared adjunctive use of a classification and regression tree (CART) diagnostic model to radiologist assessments of small renal masses (SRMs) for ccRCC. The study authors reviewed data from 309 patients who had a total of 310 SRMs, including 220 ccRCCs.
The simplified CART algorithm for detecting ccRCC included the presence on CT of a heterogenous enhancement pattern and one or more of the following signs: an early dark cortical band (EDCB), the absence of a predominantly solid lesion enhancement pattern and a corticomedullary phase ratio of lesion to normal cortex attenuation (L/C) > 1.027, according to the study. The researchers noted that one of the reviewing radiologists had 18 years of experience and the other two had three years of experience.
The researchers found that the simplified CART model had a 92 percent area under the curve (AUC), 91 percent accuracy, 93 percent specificity and 91 percent sensitivity for diagnosing ccRCC.
For unassisted radiologist assessment of SRMs on CT for ccRCC, the average AUC, accuracy, specificity, and sensitivity were 70 percent, 66 percent, 61 percent, and 67 percent respectively. Adjunctive use of the CART model led to a 19 percent improvement in the average AUC (89 percent), a 20 percent improvement in accuracy (86 percent), a 14 percent increase in specificity (75 percent), and a 23 percent increase in sensitivity (90 percent), according to the study authors.
“The CART model provides better diagnostic results than experienced radiologists and can improve the diagnostic performance of radiologists. The CART model is easy to interpret and can be used by radiologists as an effective decision-making tool,” wrote lead study author Jiayue Han, M.D., who is affiliated with the Department of Radiology at the Fifth Affiliated Hospital of Sun Yat-Sen University at Guangdong, China, and colleagues.
Three Key Takeaways
- Improved diagnostic accuracy with the CART model. Incorporating a classification and regression tree (CART) diagnostic model alongside radiologist assessments significantly enhances the detection of clear cell renal cell carcinoma (ccRCC) in small renal masses (SRMs). The simplified CART algorithm, utilizing CT features like heterogeneous enhancement pattern, early dark cortical band (EDCB), and corticomedullary phase ratio, yielded substantially higher diagnostic performance compared to unassisted radiologist assessments.
- Key CT Features for ccRCC Detection: The presence of a heterogeneous enhancement pattern on CT scans was identified as highly associated with ccRCC. Additionally, the presence of an early dark cortical band (EDCB) on CT scans was found to be another significant distinguishing factor. These features showed notable prevalence in ccRCC tumors compared to indolent renal tumors, making them valuable indicators for diagnosis.
- Utility of corticomedullary phase ratio: The corticomedullary phase ratio of lesion to normal cortex attenuation (L/C) emerged as another important distinguishing feature that may be indicative of rapid enhancement in ccRCC. This finding underscores the significance of considering this ratio in CT assessment for identifying potential cases of ccRCC within small renal masses.
Based on a multivariate logistic regression analysis, the researchers noted that the heterogeneous enhancement pattern with an SRM was “highly associated” with ccRCC, and the presence of EDCB on CT was another significant correlating factor.
“In our study, EDCB was found in 50.5% (111/220) of ccRCC tumours, but only 12.2% of indolent renal tumours. To our knowledge, this is the first time the usefulness of EDCB in identifying ccRCC has been demonstrated in a study other than the one that defined this sign,” pointed out Han and colleagues.
The study authors added that corticomedullary phase L/C is another key distinguishing feature that may signal rapid enhancement of ccRCC.
(Editor’s note: For related content, see “Can a CT-Based Radiomics Model Enhance Risk Stratification for Clear Cell Renal Cell Carcinoma?,” “CT Study: AI Algorithm Comparable to Radiologists in Differentiating Small Renal Masses” and “Emerging PET/CT Agent May Enhance Diagnosis for Smaller Lesions of Clear Cell Renal Cell Carcinoma.”)
Beyond the inherent limitations of a retrospective study, the authors acknowledged variable CT scanning protocols with the participating centers in the study and noted variable cutoff points for the CART model in different samples.
