Hi Shaimaa,<div><br></div><div><br><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Q2. If I draw ROI in specific location eg. Left cerebral peduncle and get a ROI statistics, and if I repeat the same step in each patient manually eg in left cerebral peduncle too, if I compare the results among subjects, will the results be reliable?<br>
</blockquote><div><br></div><div>For brain image analysis, manual delineation of structures is considered as the gold standard. It is recognized as ACCURATE. </div><div>However, it is known as low PRECISION.</div><div><br>
</div><div>Accuracy is also called bias and precision is related to variability or reproducibility. </div><div><br></div><div>I don't know the definition of "reliability", though (somebody can explain it?).</div>
<div><br></div><div>For example, if you have a rifle and shoot at a bull's eye. You expect your shots are scattered around the center. The amount of the scattering is precision (variability or reproducibility). It tells how good you are.</div>
<div><br></div><div>On the other hand, if your rifle's scope is misaligned, you see the center of your scattered shots is not aligned to the center of the bull's eye. This is accuracy (bias).</div><div><br></div><div>
If you use automated analysis (e.g. voxel-based analysis or automated segmentation), you get 100% precision (do 10 times and get 10 identical answers). However the accuracy is unknown. Therefore, people compare the automated analysis results with manual delineation to measure the accuracy. People want to use automated methods because manual delineation has low precision. There is a bit of circular logic here and there is no single method that is the best.</div>
<div><br></div><div>So, your approach is perfectly fine. In Fig. 1A of the attached paper, there is a description about different ROI placement approaches. Each has pros and cons. Hope this will help you understanding the art of ROI.</div>
<div><br></div><div>Susumu</div></div><br></div>