The study introduces a Natural Language Processing (NLP) approach to obtain critical information from pathological reports and classify tumor stages, with a focus on enhancing diagnostic accuracy and treatment planning. This NLP approach significantly improves patient care by enabling more precise and timely interventions. Additionally, NLP minimizes human errors in diagnostics by automatically analyzing and learning from thousands of medical reports, ensuring that critical terms are not overlooked—a common issue when humans handle large volumes of data. In this research, special attention is given to extracting detailed information regarding the T4 stage of the tumor, a stage often inadequately detailed in previous studies. Our approach successfully identifies the distinguishing factors of the T4 stage directly from the pathological reports. The performance of this proposed model was evaluated and compared with various Machine Learning (ML) models. The Support Vector Machine (SVM) model outperformed the other models for stages T1 and T2, with accuracy rates of 98% and 97.2%, respectively. On the other hand, the Naive Bayes (NB) model outperformed the other models for stages T3 and T4, with accuracy rates of 94% and 99%, respectively.