I found a temporary solutions, producing a ESRI shapefile from classic roi and using it. Howover this is not a solution for me. In fact, I need to: verify if a ESRI Shapefile exists. if yes, use it for the ML classification vector statistic computation, following the example I have linked before. if not, convert the roi classic file into a vector and then re-try to execute the ML classification I need to execute the workflow described above in batch mode through the idlpy bridge, so definitively using envi api, envi classic, idl or a combination of that. Any suggestion?

You can get the statistics from an ROI using the method shown in this example: http://www.harrisgeospatial.com/docs/... Once you calculate the stats for the classification method you want to use, you can pass them to the classification task.

It seems an interesting solution. I will try soon and I will report any problems, consideration. Thank you.