Table 2 Performance metrics obtained through evaluation of the hybrid neural network algorithm ‘GraftIQ’ in predicting diagnosis categories

From: GraftIQ: Hybrid multi-class neural network integrating clinical insight for multi-outcome prediction in liver transplant recipients

Category

Sensitivity

Specificity

PPV

NPV

AUC only multiclass NN based ML model

AUC ML model + clinical insight (GraftIQ)

ACR

0.814

0.730

0.642

0.865

0.774

0.801

AIH

0.895

0.942

0.815

0.965

0.924

0.937

BO

0.902

0.912

0.826

0.915

0.902

0.933

Congestion

0.891

0.935

0.837

0.938

0.922

0.942

HCV

0.875

0.778

0.752

0.889

0.859

0.866

MASH

0.885

0.921

0.859

0.938

0.929

0.930

  1. Performance metrics obtained using only the multiclass NN as well as the hybrid neural network algorithm (multiclass NN + clinical insight) for each complication group. Comprehensive analysis in terms of sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and area under the curve (AUC) is represented for each category. The higher each metric, the better the classification is. Note the improvement in AUC values as observed upon integrating clinician expertise into the neural network model for each diagnosis category.
Back to article page