BME Frontiers / 2022 / Article / Fig 8

Research Article

Weakly- and Semisupervised Probabilistic Segmentation and Quantification of Reverberation Artifacts

Figure 8

(a) Left to right: input images, reverberation artifact masks, results from U-Net, results from partial U-Net, and ground truths. The partial U-Net uses the reverberation artifact masks to mask the convolution and does not generate false positive where needles are present like U-Net. (b) Quantitative comparison between Partial U-Net and U-Net.