Computer Vision News - July 2021

2 Summary AI Rese ch 8 Based on the results acquired from this model, a semi-automated workflow was developed to correct for the bias of readers. This semi-automated approach may create a few problems. By adjusting the threshold value (from a human reader) variability will be seen, which may even be stronger than the one observed! In Figure 4 a visualization can be seen of the super-imposed segmentation with different thresholds. In Figure 5, you can see the pre and post-biological treatment and the corrected super-imposed segmentation by the second reader (plus the automated algorithm). Even though the deep learning network can provide a correction, it does not eliminate the variability. This is partially due to the variation between abnormal and normal water content due to the cellularity/expansion of the extracellular space but also from the variation between the water content of (e.g. from sclerosis). Wrapping up! Thanks for staying through this article! I would like to thank C. Hepburn and the authors for writing this great paper Stay tuned for more news and many interesting articles from all the colleagues! Figure 5: Here are the MRI slices with the STIR using pre, post-biological treatment (left/top respectively) and with super-imposed corrected by the second reader automatic segmentations