Medical diagnostics have shown strong potential in applying deep learning toward automating such complex tasks like interpreting medical images, laboratory results, and time series data. However, due to the inability of these models to be well interpreted, the acceptance of them in clinical setups is limited. This paper advances the development of explainable deep learning models to be applied toward medical diagnosis; the focus, however, remains on time series data, ECG, EEG, and patient vitals. The approach introduces temporal explainability, in which models explain the evolution of diagnostic decisions and highlight the time intervals that most contribute to predictions. This is achieved by integrating techniques such as attention mechanisms, saliency maps, and local interpretable model-agnostic explanations (LIME), thus providing both visual and textual justifications for model outputs. The proposed model contributes to the field of explainable AI by enabling clinicians to trace decision-making over time, thereby improving their understanding of patient-specific conditions. Experimental results from PhysioNet ECG and TUH EEG datasets demonstrate high diagnostic accuracy (92.5% and 89.7%, respectively) and interpretability scores (8.6/10 by clinicians). This approach provides both visual and textual explanations, enabling clinicians to trace the temporal evolution of decisions and encouraging greater trust in AI-based medical systems.