Table 2 (Related to Fig. 5): table showing the mean error difference and their standard errors between ABRA-detected wave 1 latency and amplitude and corresponding ground truth values detected by human reviewers. Two sample t-tests found that mean error differences were not significant for wave 1 latency nor for wave 1 amplitude estimates at the 95% significance level after bonferroni correction.
From: An open-source deep learning-based toolbox for automated auditory brainstem response analyses (ABRA)
Wave 1 Latency (ms) | Wave 1 Amplitude (µV) | ||
|---|---|---|---|
Mean Difference between prediction and ground truth (± S.E.M.) | Lab A (nwaveforms=80, nmice=8) | -0.046 (± 0.021) | -0.115 (± 0.043) |
Lab B (nwaveforms=2,247, nmice=21) | -0.003 (± 0.002) | -0.009 (± 0.005) | |
Overall Test Set | -0.004 (± 0.002) | -0.013 (± 0.005) | |
RMSE (95% CI) | Overall Test Set | 0.092 (0.088-0.095) | 0.245(0.235-0.255) |
