A Clinical-Radiomic Nomogram Based on Unenhanced Computed Tomography for Predicting the Risk of Aldosterone-Producing Adenoma

Abstract

Purpose: To develop and validate a clinical-radiomic nomogram for the preoperative prediction of the aldosterone-producing adenoma (APA) risk in patients with unilateral adrenal adenoma.

Patients and methods: Ninety consecutive primary aldosteronism (PA) patients with unilateral adrenal adenoma who underwent adrenal venous sampling (AVS) were randomly separated into training (n = 62) and validation cohorts (n = 28) (7:3 ratio) by a computer algorithm. Data were collected from October 2017 to June 2020. The prediction model was developed in the training cohort. Radiomic features were extracted from unenhanced computed tomography (CT) images of unilateral adrenal adenoma. The least absolute shrinkage and selection operator (LASSO) regression model was used to reduce data dimensions, select features, and establish a radiomic signature. Multivariable logistic regression analysis was used for the predictive model development, the radiomic signature and clinical risk factors integration, and the model was displayed as a clinical-radiomic nomogram. The nomogram performance was evaluated by its calibration, discrimination, and clinical practicability. Internal validation was performed.

Results: Six potential predictors were selected from 358 texture features by using the LASSO regression model. These features were included in the Radscore. The predictors included in the individualized prediction nomogram were the Radscore, age, sex, serum potassium level, and aldosterone-to-renin ratio (ARR). The model showed good discrimination, with an area under the receiver operating characteristic curve (AUC) of 0.900 [95% confidence interval (CI), 0.807 to 0.993], and good calibration. The nomogram still showed good discrimination [AUC, 0.912 (95% CI, 0.761 to 1.000)] and good calibration in the validation cohort. Decision curve analysis presented that the nomogram was useful in clinical practice.

Conclusions: A clinical-radiomic nomogram was constructed by integrating a radiomic signature and clinical factors. The nomogram facilitated accurate prediction of the probability of APA in patients with unilateral adrenal nodules and could be helpful for clinical decision making.

Keywords: adenoma; nomogram; precision medicine; primary aldosteronism; radiomics.

Figure 1

The patient enrollment pathway, along with the inclusion and exclusion criteria.

Figure 2

APA candidate variable selection using LASSO regression. (A) Binomial deviation graph of the optimal tuning parameter (λ) in the LASSO model. (B) LASSO coefficient profiles of the nine possible influencing factors.

Figure 3

Receiver operating characteristic (ROC) curve analysis based on the model prediction. The best cutoff values are indicated on the curves. (A) ROC curve of the training cohort. (B) ROC curve of the validation cohort.

Figure 4

The clinical-radiomic nomogram that was developed.

Figure 5

Calibration curves of the clinical-radiomic nomogram in the training cohort (A) and validation cohort (B).

Figure 6

DCA of the nomogram model. The y-axis represents the net benefit. The red line represents the predictive APA nomogram model. The gray line represents the assumption that all patients have APA. The black line represents the assumption that no patients have APA.