Background: Lung cancer surgery often involves complex decision-making, where accurate and interpretable predictive models are crucial for assessing postoperative risks and optimizing outcomes. This study presents XplainLungSHAP, a novel framework combining SHAP (SHapley Additive exPlanations) and attention mechanisms to enhance both predictive accuracy and transparency. The aim is to support clinicians in preoperative evaluations by identifying and prioritizing key clinical features. Methods: The framework was developed using data from 470 patients undergoing lung cancer surgery. Key clinical features were identified through SHAP, ensuring alignment with medical expertise. These features were dynamically weighted using an attention mechanism in a neural network, enhancing their impact on survival predictions. The model’s performance was evaluated through accuracy, confusion matrices, and ROC analysis, demonstrating its reliability and interpretability. Results: The XplainLungSHAP model achieved an accuracy of 91.49%, outperforming traditional machine learning models. SHAP analysis identified critical predictors, including pulmonary function, comorbidities, and age, while the attention mechanism prioritized these features dynamically. The combined approach ensured high accuracy and offered actionable insights into survival predictions. Conclusions: XplainLungSHAP addresses the limitations of black-box models by integrating explainability with state-of-the-art predictive techniques. This framework provides a transparent and clinically relevant tool for guiding surgical decisions, supporting personalized care, and advancing AI applications in thoracic oncology.

Understanding the decision of large deep learning models is a critical challenge for building transparent and trustworthy systems. Although the current post hoc explanation methods offer valuable insights into feature importance, they are inherently disconnected from the model training process, limiting their faithfulness and utility. In this work, we introduce a novel differentiable approach to global explainability by design, integrating feature importance estimation directly into model training. Central to our method is the ScoresActivation function, a feature-ranking mechanism embedded within the learning pipeline. This integration enables models to prioritize features according to their contribution to predictive performance in a differentiable and end-to-end trainable manner. Evaluations across benchmark datasets show that our approach yields globally faithful, stable feature rankings aligned with SHAP values and ground-truth feature importance, while maintaining high predictive performance. Moreover, feature scoring is 150 times faster than the classical SHAP method, requiring only 2 seconds during training compared to SHAP's 300 seconds for feature ranking in the same configuration. Our method also improves classification accuracy by 11.24% with 10 features (5 relevant) and 29.33% with 16 features (5 relevant, 11 irrelevant), demonstrating robustness to irrelevant inputs. This work bridges the gap between model accuracy and interpretability, offering a scalable framework for inherently explainable machine learning.