Figure 2 | Scientific Reports

Figure 2

From: Training data size and predication errors in the use of machine-learning assisted intraocular lens power calculation

Figure 2

Effect of the dataset sized from full (4800 eyes) to 75 eyes on the proportions within ± 0.25 D, ± 0.50 D, and ± 1.00 D errors. Compared with the use of overall data (N = 4800), the proportions within ± 0.50 D error for the dataset of 75 eyes were significantly less (P = 0.016, * in Fig.). For errors within ± 0.25 D, the use of datasets of 75 and 150 eyes resulted in a lower proportion (P < 0.030, * in Fig.).

Back to article page