PERSONALIZED MEDICINE RECOMMENDATION SYSTEM FOR VARIOUS LUNG CANCER TYPES USING FDHSCNN AND LIME-FWPIMS

Personalized medicine plays a vital role in improving accuracy in diagnosis and treatment planning for lung cancer by patient-specific imaging and computational intelligence. This paper proposes a Personalized Medicine Recommendation System for Various Lung Cancer Types based on a Feature-Driven Hierarchical Spatial Convolutional Neural Network (FDHSCNN) combined with an explainable framework named Local Interpretable Model-agnostic Explanations (LIME)based Feature-Weighted Personalized Inference and Mapping System (LIME-FWPIMS). The proposed system utilizes multimodal PET–CT images to capture both functional and anatomical tumor characteristics, allowing for precise classification of multiple lung cancer subtypes.

The FDHSCNN architecture hierarchically learns discriminative spatial and deep semantic features, improving robustness and classification performance while mitigating overfitting. To enhance transparency and reliable clinical outputs, LIME-FWPIMS is employed to generate localized, feature-weighted explanations, supporting interpretable and personalized clinical decision-making. Experimental evaluation on PET–CT datasets demonstrates that the proposed method achieves a high classification accuracy of a 99.07%, outperforming existing deep learning–based approaches.

The results indicate that integrating advanced deep feature learning with explainable artificial intelligence, significantly enhances diagnostic reliability and personalized treatment recommendation. The proposed framework shows strong potential for deployment in precision oncology applications, offering accurate, interpretable, and patient-centric lung cancer diagnosis.