Yes, but you have to use two commands. Here is the complete code (confirmed to work with the current DataLad release 0.12):
datalad clone ///adhd200/RawData/Brown
datalad -C Brown get 0026001/session_1/anat_1
Instead of install --get-data, first clone the dataset (you can also keep using install, but strip the -g option). This will obtain the “empty” dataset only (which is fast and lean). Then follow it up with a get for the files you are interested in. The -C changes the working directory to the root of the Brown dataset – just like Git’s -C option.