I cannot see how you could do it. Take for example the normal distribution; mean and median are equal. And if a distribution is skewed, the difference between the median and the mean does not inform you about the spread of the data. However, for small samples you can get an estimate of SD from the range (max minus min) and the sample size.

Thank you Dear Ballabeni

Yes, in the case of Normal distribution we can estimate SD from range divided by 4.  However, in the above mentioned case, I encountered a problem in data abstraction for estimation of pooled effect size across several studies. I need mean and SD or other measure of distribution such as SE or range. 

what is your recommendation in this case; exclusion of the study or roughly estimation of SD from other similar studies?

Check out this website https://influentialpoints.com/Training/standard_error_of_median.htm