Fig. 2 | Scientific Reports

Fig. 2

From: Lossy DICOM conversion may affect AI performance

Fig. 2

Quantitative differences between original and DICOM-converted images. Structural similarity index (SSIM) was calculated between original and DICOM-converted tiles extracted from WSIs of bladder, ovarian, and prostate tissue using two conversion tools: SlideMaster (blue) and wsidicomizer (SlideMaster settings (q = 80): orange, default settings (q = 90): beige). (a) Heatmap depicting the dataset based SSIM values against original MRXS tiles, given numbers are mean and standard deviation. (b) Boxplot depicting the patient-based variation between tiles for bladder tissue tiles at 0.25 mpp (top) and at 1.0 mpp (bottom).

Back to article page