Evaluation of CT-based radiomics signature and nomogram as prognostic markers in patients with laryngeal squamous cell carcinoma

Abstract

Background: The aim of this study was to evaluate the prognostic value of radiomics signature and nomogram based on contrast-enhanced computed tomography (CT) in patients after surgical resection of laryngeal squamous cell carcinoma (LSCC).

Methods: All patients (n = 136) were divided into the training cohort (n = 96) and validation cohort (n = 40). The LASSO regression method was performed to construct radiomics signature from CT texture features. Then a radiomics nomogram incorporating the radiomics signature and clinicopathologic factors was established to predict overall survival (OS). The validation of nomogram was evaluated by calibration curve, concordance index (C-index) and decision curve.

Results: Based on three selected texture features, the radiomics signature showed high C-indexes of 0.782 (95%CI: 0.656-0.909) and 0.752 (95%CI, 0.614-0.891) in the two cohorts. The radiomics nomogram had significantly better discrimination capability than cancer staging in the training cohort (C-index, 0.817 vs. 0.682; P = 0.009) and validation cohort (C-index, 0.913 vs. 0.699; P = 0.019), as well as a good agreement between predicted and actual survival in calibration curves. Decision curve analysis also suggested improved clinical utility of radiomics nomogram.

Conclusions: Radiomics signature and nomogram showed favorable prediction accuracy for OS, which might facilitate the individualized risk stratification and clinical decision-making in LSCC patients.

Keywords: Head and neck cancer; Laryngeal cancer; Nomograms; Prognosis.

Fig. 1

Texture feature selection using LASSO Cox regression. a Selection of tuning parameter (λ) in the LASSO model using 10-foldncross-validation with minimum criteria. Partial likelihood deviance was plotted versus log (λ). Vertical line of the optimal values (log (λ) = − 2.573) were drawn based on the minimum criteria and the 1-standard error of the minimum criteria. b LASSO coefficient profiles of 36 texture features. Vertical line was plotted at the selected value via 10-fold cross-validation, where optimal λ resulted in 3 nonzero coefficients

Fig. 2

Development of the prognostic index based on radiomics score (Rad-score). Distribution of prognostic index in training cohort (a) and validation cohort (b). Patients were sorted in numerical order according to Rad-score on the x-axes, and divided into high-risk and low-risk groups. Survival status in training cohort (c) and validation cohort (d) showed higher proportion of death patients in high-risk group. Heatmap of texture features values in training cohort (e) and validation cohort (f) tended to be higher in high-risk patients

Fig. 3

Kaplan-Meier curves with number of risk and censoring for OS in the training cohort (a) and validation cohort (b). Elevated radiomics scores were significantly associated with poorer OS

Fig. 4

Radiomics nomogram for the prediction of 1-year and 3-year OS based on the training cohort

Fig. 5

Calibration curves of radiomics nomogram in the training cohort (a) and validation cohort (b). The diagonal dotted line represented a perfect prediction by an ideal model, and solid line represented performance of the nomogram. The closer fit between the diagonal dotted lines and solid lines showed good prediction of 1-year and 3-year OS

Fig. 6

Decision curve analysis for each model based on the validation dataset. The risk probability of death was recorded as Pi. When Pi reached a certain threshold probability (Pt), it was defined as positive, and some intervention were taken. The net benefit was calculated by summing the benefits of intervention in true positive proportion (a) and loss benefit of unnecessary treatment in false positive proportion (b). Net benefit = a-b [Pt/(1-Pt)]. The horizontal black line represented all negative samples and no intervention. The gray dotted oblique represented intervention of all patients. The radiomics nomogram had the highest net benefit compared with all-treat scheme or non-treat scheme and other models across the range of 15–55% in Pt. For example, if the Pt reached 30%, the net benefit was about 0.10 when using the radiomics nomogram to determine whether to perform therapies