Mpox (Monkeypox) detection presents an ongoing challenge in clinical diagnostics, necessitating robust computational models capable of distinguishing Mpox from related conditions. This research introduces a motif-driven machine learning framework for classifying Mpox using clinical image datasets. Motif discovery is applied as a novel preprocessing step, taking advantage of the MEME algorithm to extract statistically significant patterns that guide feature engineering. These motif-based features inform the training of several machine learning models, including Support Vector Machines (SVM), Random Forests (RF), Decision Trees (DT), and Dense Neural Networks (DNN). Empirical results demonstrate that motif-informed features substantially enhance model generalization and resilience to noise. The SVM classifier, in particular, achieved a high recall of 93% for Mpox cases, despite an overall accuracy of 42%, highlighting its effectiveness in sensitivity-critical diagnostic settings. Meanwhile, RF and DT showed more balanced, though less optimal, performance, indicating the challenges of handling imbalanced datasets. To further improve robustness, the framework incorporates data augmentation, real-time preprocessing, and class-weighted loss functions, ensuring adaptability to diverse clinical scenarios. This study concludes that motif-based feature engineering is a powerful strategy for improving recall in high-priority disease detection, making it a valuable asset for global health diagnostics. Future directions include the development of ensemble approaches combining SVM and DNN, domain-specific motif embeddings, and cloud-based deployment for scalable, real-world application.