Good Morning,
I use a free utility called PhotoMe, that supports a wide range manufacturers. Running (actually dragging an image into) through PhotoMe, shows that the shutter count value is not standard field (variable) within the EXIF metadata structure, since it occurs in the "Manufacturer Notes" area of the EXIF. That tells me that each manufacturer places the count in different locations. PhotoMe has it labeled as a "undefined(4)", which indicates to me that it is indeed 4 bytes long, which probably should be interpreted as a long integer. That would give it a unsigned positive range of 0 to 4,294,967,295, more than sufficient to handle a function such as shutter count for the lifetime of a body - and if you exceed that it would just roll over to 0 and start again. Plus Pentax (er, as Ricoh) would probably like to hear about the occasion.
PhotoMe as a default comes up with a Nikon D80 image (as an example). It shows that Nikon places shutter count in two areas, one in a packed data structure (1156 bytes) that has a shutter count (with a displacement of 586 to 589) in deed a 4 byte integer, and a second field labeled shutter count that is a long integer (4 bytes). Each of these fields in the default example have an identical value in this case of 3598.
So just going on the values that you supplied, decimal 245, 27, 238, 164, would represent hex values of F5 1B EE A4 which as a long integer would have a value of 4112248484 (which I am guessing is a tad bit larger than what the actual shutter count value is being reported as). This is up to some conjecture since I don't know if the processor / compiler is big endian or little endian, byte/nibble swapped or not, signed or unsigned, etc., or even if RAW Therapee provided you with the correct values (from the right displacement).
Anyway, here is some information on the EXIF data structure...