An ensemble deep learning for automatic prediction of papillary thyroid carcinoma using fine needle aspiration cytology

•Deep learning for identifying papillary thyroid carcinomas and computing certainty.•Ensemble technique combines multiple deep CNNs to significantly increase accuracies.•GRAD-CAM method highlights dominant tissues for the discriminative decision-making.•Effectiveness of stain normalization methods on predictive performances is evaluated.•Proposed ensemble learning is superior to current models for automatic diagnosis.