This work investigates the identification of Colorectal Carcinogenesis (CRC) through the analysis of textural alterations in rat liver microphotographs. Prior approaches using features extracted using Convolutional Neural Networks (CNNs) achieved a promising accuracy of 92.5% but indicated room for optimization. To address this, the present study proposes an in-depth, systematic analysis of the Discrete Wavelet Transform (DWT), a classic texture descriptor capable of multi-resolution analysis. The methodology explores the impact of mother wavelet selection, decomposition level, and the statistical features extracted from the resulting sub-bands. The optimized pipeline, utilizing the mother wavelet Daubechies with five vanishing moments (db5) and with three decomposition levels and features of mean, standard deviation, energy, and entropy, achieved a state-of-the-art accuracy of 98.75%. This result provides a new and robust performance baseline for this task, demonstrating the value of applying specialized and interpretable texture descriptors to this specific histopathological application.