Fig. 4: Performance of DeepNeo and human experts. | Communications Medicine

Fig. 4: Performance of DeepNeo and human experts.

From: Deep learning model DeepNeo predicts neointimal tissue characterization using optical coherence tomography

Fig. 4

Confusion matrices for the performance of DeepNeo and experts B and C with labels by expert A taken as ground truth (a). Note that automated analysis by DeepNeo is similar to the inter-expert variability. N = 420 (not analyzable: 23, homogenous: 186, heterogenous: 117, neoatherosclerosis: 94). Calibration of DeepNeo (b): the probability of predicted class (x-axis) vs. true probability (y-axis). To calculate true probability, the samples are split into 10 equally sized bins according to the predicted probability of a sample. The true probability for a bin is then calculated by dividing the number of true predictions by the number of samples.

Back to article page