The segmentation accuracy of the manual approach is also affected by the proximity of the target and the region it is extracted from. For example, the presence of an osteophyte can in general hinder the segmentation of a nearby cartilage, a result that can even be exacerbated by a segmentation error in the osteophyte.
The overall segmentation accuracy of the manual approach also depends on factors such as the presence of other foci and even the quality of the image. For single-excitation static datasets, the quality of the MR image has a big impact on the reproducibility of the segmentation. In general, images with poor spatial resolution and/or field of view are especially vulnerable to errors resulting from signal polynomial - t . In these instances, the segmentation results can even be misleading or completely wrong while the manual delineation is largely correct.
Other contributing factors that influence the manual segmentation could include the presence of superficial or deep contamination or heterogeneity, or signal polynomial - t . Another factor that has been shown to influence segmentation accuracy is the portion of image which is segmented. Some datasets, by definition, present low-quality areas where segmentation is most challenging. Segmenting an entire image may therefore bias the results leading to over- or under-segmentation. d2c66b5586