This chapter explores the significant impact that explainability methods have within the clinical diagnosis field, focusing on how the advanced application of LIME and GraphLIME can enhance the predictability and performance of models. We investigate how the extraction and analysis of relevant features through an attention mechanism directly influence the accuracy of predictions in diagnosing thoracic diseases. A clear distinction is made between LIME and GraphLIME, highlighting the advantages of each method and their applicability in the context of analyzing tabular datasets and images. By implementing these explainability techniques, the chapter demonstrates a notable improvement in model performance, achieving at least a 25% increase in accuracy, aided by the integration of attention mechanisms. The outcomes are illustrated through ROC curves, confusion matrices, and supplementary plots, providing a detailed insight into the mechanisms and architecture underlying the enhancement of predictions. This exploration emphasizes the importance of adopting explainability methods in thoracic diagnosis, suggesting that a deep understanding of data and the prediction process can lead to significant advancements in diagnostic precision. Additionally, this chapter proposes a new perspective on model interpretability, suggesting that such insights could pave the way for more personalized and effective medical treatments, further underscoring the potential of explainability in transforming healthcare analytics.