I think I have found the answer from the manual: "The Image Processing Toolbox software follows the convention that double-precision L*a*b* arrays contain 1976 CIE L*a*b* values. L*a*b* arrays that are uint8 or uint16 follow the convention in the ICC profile specification (ICC.1:2001-4, www.color.org) for representing L*a*b* values as unsigned 8-bit or 16-bit integers".
So If i convert the LAB image in uint format to double format with "lab2double", adapthisteq works well.