This research investigates the optimization of ovarian cancer diagnosis through integrating multi-modal data with the hybrid evolutionary deep learning model. Employing CT and MRI scan images, as well as textual radiology reports, and patient history, this study utilizes Convolutional Neural Networks for spatial feature extraction and Gated Recurrent Unit for processing sequential data. The data used in the analysis include images obtained from the GDC portal for 587 ovarian cancer patients. To analyze the data, the dataset was split into training and testing subsets, with the training subset representing 70% of the data. The results show the CNN model’s effectiveness in spatial feature extraction when analyzing medical images, as it demonstrates the accuracy of 93.85%. As for the GRU model, the results present a satisfactory outcome as well, as it processes sequential data with the accuracy of 90.35%. When integrating both CNN and GRU models, the results suggest the combined accuracy of 97.45%, which is higher than the accuracy obtained when analyzed separately. Lastly, the performance of each model is evaluated through precision, recall, F1 score, and AUC ROC metrics, which emphasize the implications of the integrated approach to achieving higher diagnostic accuracy. It should be noted that the results also detect the superior predictive abilities of the suggested hybrid evolutionary deep learning model. Overall, the findings can contribute to clinical practice since healthcare professionals are provided with an effective tool for earlier detection and more accurate diagnosis of ovarian cancer, which, consequently, enhances the quality of patient care and their outcomes.