Figure 2 | Scientific Reports

Figure 2

From: Discriminative Scale Learning (DiScrn): Applications to Prostate Cancer Detection from MRI and Needle Biopsies

Figure 2

Visualization of multi-scale (b) Gabor, (c) Haralick, (d) LBP and (e) PHOW features. Each feature is extracted at the same three scales: 3 × 3, 5 × 5, and 7 × 7. As compared to the (a) cancer ground truth, we can see that (1) no scale pattern is obviously more discriminative than the others, and (2) no individual feature appears to perfectly distinguish cancer and benign regions. Therefore, it is necessary to combine all these features and learn the most discriminating scales for each of the features.

Back to article page