Research Article
Deep Learning to Predict EGFR Mutation and PD-L1 Expression Status in Non-Small-Cell Lung Cancer on Computed Tomography Images
Table 2
Predictive performance in predicting EGFR mutation and PD-L1 expression status.
| | | EGFR | PD-L1 | ACC (95%CI) | AUC (95%CI) | Sensitivity (95%CI) | Specificity (95%CI) |
| | Training set | − | − | 0.92 (0.91–0.93) | 0.97 (0.96–0.97) | 0.87 (0.83–0.9) | 0.93 (0.92–0.95) | | − | + | 0.91 (0.89–0.93) | 0.96 (0.95–0.97) | 0.75 (0.7–0.78) | 0.97 (0.95–0.98) | | + | − | 0.87 (0.85–0.89) | 0.96 (0.95–0.97) | 0.96 (0.94–0.98) | 0.81 (0.79–0.84) | | + | + | 0.91 (0.90–0.92) | 0.95 (0.94–0.96) | 0.39 (0.33–0.46) | 1 (0.99–1) | | Average | 0.90 (0.86–0.93) | 0.96 (0.94–0.98) | 0.74 (0.31–1) | 0.93 (0.79–1) |
| | Validation set | − | − | 0.78 (0.72–0.72) | 0.82 (0.75–0.88) | 0.57 (0.44–0.7) | 0.85 (0.79–0.9) | | − | + | 0.76 (0.71–0.71) | 0.78 (0.71–0.84) | 0.45 (0.33–0.57) | 0.86 (0.81–0.91) | | + | − | 0.74 (0.68–0.68) | 0.85 (0.79–0.89) | 0.78 (0.69–0.87) | 0.71 (0.64–0.78) | | + | + | 0.84 (0.79–0.79) | 0.75 (0.66–0.82) | 0.13 (0.03–0.26) | 0.95 (0.92–0.98) | | Average | 0.78 (0.70–0.86) | 0.80 (0.72–0.88) | 0.48 (0.01–0.95) | 0.84 (0.66–1) |
| | Test set | − | − | 0.74 (0.7–0.7) | 0.76 (0.72–0.81) | 0.45 (0.36–0.54) | 0.82 (0.79–0.86) | | − | + | 0.68 (0.64–0.64) | 0.66 (0.61–0.71) | 0.28 (0.21–0.36) | 0.83 (0.79–0.87) | | + | − | 0.73 (0.69–0.69) | 0.79 (0.75–0.83) | 0.82 (0.77–0.88) | 0.68 (0.63–0.73) | | + | + | 0.83 (0.79–0.79) | 0.69 (0.63–0.75) | 0.15 (0.07–0.23) | 0.95 (0.93–0.97) | | Average | 0.75 (0.65–0.85) | 0.73 (0.63–0.83) | 0.43 (0–0.92) | 0.82 (0.62–1) |
|
|
Abbreviation: ACC: accuracy; AUC: area under the ROC curve.
|
