Early stage colorectal carcinoma (CRC) infiltrates the submucosa (pT1) and can spread to lymph nodes via lymphatic vessels. In pT1 CRC, there are no specific morphological characteristics capable of predicting the presence of lymph node involvement at the time of the endoscopic procedure, nor is a histological signature with predictive value obtained when they are combined. The objective of this study is to define a histological signature that combines several morphological parameters (degree of tumor differentiation, DGT, angiolymphatic, ALI, and perineural, PI, invasion, tumor budding, TB, poorly differentiated clusters, PDC, submucosal invasion, SMI, and, both, vertical and horizontal resection margins, VRM, HRM) to predict regional lymph node involvement at the time of endoscopic resection of pT1 CRC. Multivariate statistical analysis was used in combination with unsupervised techniques to identify the combination of pathological parameters that best stratifies lymph node involvement. Validation was done on a series of 223 pT1 CRC (94 women and 129 men) aged between, of which 89 had regional lymph node metastases at the time of diagnosis. In a univariate analysis, the distribution of DGT (p = 0.043), IAL (p = 0.0026), TB (p = 0.000), PDC (p = 0.013), and ISM (p = 0.0035) was significantly associated with lymph node involvement. The multivariate analysis showed that the combination of the classic parameters is a better indicator of metastasis than that obtained with the analysis of the parameters considered individually (p = 0.0066). However, when all parameters were combined to the most favorable value, 10% of patients with lymph node metastases were observed in this group (false negatives). This suggests further investigation on new technologies to identify a biomarker probably combining image analysis with morphological signs, in pT1 CRC that can be used to stratify the risk of lymph node involvement and determine the best treatment.