Figure 2 | Scientific Reports

Figure 2

From: Detecting Retinal Nerve Fibre Layer Segmentation Errors on Spectral Domain-Optical Coherence Tomography with a Deep Learning Algorithm

Figure 2

Spectral-domain optical coherence tomography (SDOCT) B-scan with no segmentation error, as labelled by human graders and the deep learning algorithm. Class activation map (heatmap) involves the whole B-scan, rather than being concentrated on a particular area. The probability of segmentation error given by the deep learning algorithm was below 1%.

Back to article page