Near infrared transmission spectroscopy of Whole blood are investigated with different thickness(0.5, 1, 2, 4 mm)in order to explore the feasibility of detecting alanine aminotransferase rapidly by near-infrared spectra. The results show that the whole blood sample with 0.5 mm thickness is more suitable for spectral analysis. And then Near infrared spectroscopy of 176 samples were collected. Multiplicative scatter correction and second-order differential method have been used to spectral pretreatment. Stepwise multiple linear regression method and partial least squares regression method have been employed to establish quantitative detection model to predict content of alanine aminotransferase in whole blood. The alanine aminotransferase measured presents best result in calibration and prediction by Near-Infrared Spectroscopy with partial least squares regression calibration model, and the calibration correlation coefficient, the standard error of calibration and the standard error of prediction are 0.98, 2.42 and 7.22 respectively.