The field of medical image mining has garnered significant attention from researchers and professionals alike. This paper delves into the challenges and issues associated with medical images, such as low accuracy, poor quality, and false features. In response, we propose a prototype framework that utilizes image processing and data mining to enhance diagnostic decision-making through the extraction of relevant features from medical images. Firstly, the framework implements image processing algorithms to address problems related to brightness and imaging environment, thereby improving the quality of targeted medical images. Secondly, image mining techniques, such as segmentation and clustering, are employed on the processed images to identify and extract pertinent indicators. The model is trained iteratively using reference images, and classification techniques are utilized to identify features in test medical images. The prototype, developed using MATLAB, was tested on medical images of patients suspected to have leukemia. Results demonstrate that the proposed framework outperforms many comparable models using the same dataset, with a maximum accuracy of 98% achieved using K-mean segmentation and Super vector machine (SVM) clustering, compared to the 85-95% accuracy of commonly used frameworks for leukemia diagnosis. Validation of the proposed model confirms its adequacy and highlights the value added by incorporating image mining after preprocessing medical images using typical image enhancement techniques.