Nonlinear optical imaging techniques based e.g. on coherent anti-Stokes Raman scattering (CARS) or second-harmonic
generation (SHG) show great potential for in-vivo investigations of tissue. While the microspectroscopic imaging tools
are established, automized data evaluation, i.e. image pattern recognition and automized image classification, of
nonlinear optical images still bares great possibilities for future developments towards an objective clinical diagnosis.
This contribution details the capability of nonlinear microscopy for both 3D visualization of human tissues and
automated discrimination between healthy and diseased patterns using ex-vivo human skin samples. By means of CARS
image alignment we show how to obtain a quasi-3D model of a skin biopsy, which allows us to trace the tissue structure
in different projections. Furthermore, the potential of automated pattern and organization recognition to distinguish
between healthy and keloidal skin tissue is discussed. A first classification algorithm employs the intrinsic geometrical
features of collagen, which can be efficiently visualized by SHG microscopy. The shape of the collagen pattern allows
conclusions about the physiological state of the skin, as the typical wavy collagen structure of healthy skin is disturbed
e.g. in keloid formation. Based on the different collagen patterns a quantitative score characterizing the collagen
waviness - and hence reflecting the physiological state of the tissue - is obtained. Further, two additional scoring
methods for collagen organization, respectively based on a statistical analysis of the mutual organization of fibers and on
FFT, are presented.