As usual for ViewForum burnt CDROMs, the Endianness switches from Little Endian
to Big Endian, and then back to Little Endian within an 'Explicit VR' 
private DICOM Sequence (the length of which is *not* 0xffffffff).
The heuristics to guess the endianness looking at the 'delimitors'
(whether they start with 0xfffe or 0xfeff) works for a while, and then breaks.
in SQ 2001|e05f, the first Item starter (at offset 0x1102) tells us 
the endianess swaps to BigEndian (we find FFFE  E000) ;
The entry of the first SQItem (2001|0010) is BigEndian, but its length (1600 on
disk is LittleEndian)  
It seems to be a serious bug !

The only way to read the image is to ask the reader (if you can) to skip the
'private sequences'.
-> use gdcm::File::SetLoadMode( NO_SHADOWSEQ );