Spectroscopic and deep learning-based approaches to identify and quantify cerebral microhemorrhages

Abstract

Cerebral microhemorrhages (CMHs) are associated with cerebrovascular disease, cognitive impairment, and normal aging. One method to study CMHs is to analyze histological sections (5-40 μm) stained with Prussian blue. Currently, users manually and subjectively identify and quantify Prussian blue-stained regions of interest, which is prone to inter-individual variability and can lead to significant delays in data analysis. To improve this labor-intensive process, we developed and compared three digital pathology approaches to identify and quantify CMHs from Prussian blue-stained brain sections: (1) ratiometric analysis of RGB pixel values, (2) phasor analysis of RGB images, and (3) deep learning using a mask region-based convolutional neural network. We applied these approaches to a preclinical mouse model of inflammation-induced CMHs. One-hundred CMHs were imaged using a 20 × objective and RGB color camera. To determine the ground truth, four users independently annotated Prussian blue-labeled CMHs. The deep learning and ratiometric approaches performed better than the phasor analysis approach compared to the ground truth. The deep learning approach had the most precision of the three methods. The ratiometric approach has the most versatility and maintained accuracy, albeit with less precision. Our data suggest that implementing these methods to analyze CMH images can drastically increase the processing speed while maintaining precision and accuracy.

Figure 1

Overarching experimental timeline and data processing scheme. (A) Experimental timeline to induce cerebral microhemorrhages (CMHs), followed by tissue sectioning, staining, imaging, analysis, and quantitation. (B) Data preprocessing steps include: (1) acquiring raw color images of each CMH, (2) manually annotating color images by four users, and (3) determining the ground truth through majority voting. Three segmentation approaches (ratiometric, phasor analysis, and deep learning) were developed and compared. After the segmentation approaches, each set of results went through a series of image processing, data quality check, and quantitation steps (see the Image Processing section for details regarding each approach).

Figure 2

Receiver operating characteristics (ROC) curve for each approach to localize CMHs. (A) The ratiometric approach had an AUC of 0.973 and a final post-processed sensitivity and specificity of 0.835 and 0.997, respectively. (B) The phasor analysis approach had an AUC of 0.960 and a final post-processed sensitivity and specificity of 0.768 and 0.998, respectively. (C) The deep learning approach had an AUC of 0.932 and a final post-processed sensitivity and specificity were 0.708 and 0.998, respectively. The red circles represent the sensitivity and specificity for each independent user (relative to the ground truth), and the green squares represent the post-processed sensitivity and specificity for each approach.

Figure 3

Representative examples (n = 7) of raw RGB images, ground truth data, and three segmentation approaches. The sensitivity is shown in the upper left of each image. The specificity of each image is greater than 0.990. The scale bar in the raw images is 50 µm.

Figure 4

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.

Figure 5

CMH area quantification between the segmentation approaches and ground truth data with the Bland–Altman plot. (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) Comparison between the ground truth and the segmentation approaches (D) ratiometric, (E) phasor analysis, and (F) deep learning for areas less than 1500 µm². In each graph (A–F), the y-axis is the percent difference between the areas from the selected segmentation approach (ratiometric, phasor, or deep learning) and the ground truth, and the x-axis is the mean area between the selected segmentation approach and ground truth. The solid black line is the mean difference and the dashed black lines are the 95% upper and lower limits of agreement.