Figure 4
From: Spectroscopic and deep learning-based approaches to identify and quantify cerebral microhemorrhages
CMH area quantification between segmentation approaches and ground truth data with intraclass correlation coefficient (ICC). (A–C) Comparison between the ground truth and the segmentation approaches (A) ratiometric, (B) phasor analysis, and (C) deep learning for all calculated areas. (D, E) Magnified data of the shaded gray box shown in (A–C). Comparison between the ground truth and the segmentation approaches (D) ratiometric, (E) phasor analysis, and (F) deep learning for areas less than 1500 µm2. Intraclass correlation coefficients (ICC) and 95% confidence intervals are located in the lower right of each figure. In each graph (A–F), the solid black line is unity, and the dashed lines are ± 20% error.
