Applicability of quantitative CT image features in predicting the clinical course of adrenocortical carcinoma

BACKGROUND: Adrenocortical carcinoma (ACC) accounts for 0.4–4% of adrenal incidentalomas. The high risk of aggressive clinical behavior and the limited therapeutic options for ACC underscore the importance of early diagnosis and effective prediction of the clinical course of this malignancy. Classification based on the Ki-67 proliferation index has prognostic significance, while computed tomography (CT) provides superior identification of structural tumor characteristics compared to other imaging modalities. Radiomics, as a comprehensive approach to medical image analysis, has the potential to improve diagnostic accuracy and contribute to more effective management of patients with ACC.

AIM: To develop a methodology for the quantitative analysis of contrast-enhanced CT (CECT) images of ACC to assess the risk of tumor progression, taking into account the Ki-67 proliferation index.

METHODS: A retrospective analysis of preoperative four-phase CECT images was conducted. Exclusion criteria included artifacts in the adrenal region on CT, lack of histopathological verification, absence of any CT phases (unenhanced, arterial, venous, or delayed), and absence of immunohistochemical evaluation for the Ki-67 index. CECT results underwent comprehensive analysis, including tumor segmentation using 3D Slicer 5.6.2 software, extraction of quantitative features using the PyRadiomics 3.1.0 module in Python 3.9.21, post-processing of extracted features, dimensionality reduction, and cluster analysis to evaluate the effectiveness of the proposed methodology for quantitative CT image analysis.

RESULTS: Quantitative analysis of CECT images was performed for 24 patients with histologically verified ACC. Patients were divided into groups based on the Ki-67 index threshold value of 10% (15 cases with Ki-67 > 10% and 9 cases with Ki-67 ≤ 10%). Comparison of patient groups following clustering into two groups using the fuzzy K-means method and Fisher’s exact test (p-value = 0.015) revealed a statistically significant difference between the clusters at a significance level of 0.05 when classified according to the Ki-67 index.

CONCLUSION: The application of radiomics demonstrates the potential to improve diagnostic accuracy through quantitative analysis of medical images.