Ovarian cancer is a highly lethal gynecological malignancy, and early and accurate diagnosis is crucial. This study presents OvcaFinder, an interpretable model that integrates ultrasound images-based deep learning (DL) predictions, Ovarian-Adnexal Reporting and Data System (O-RADS) scores from radiologists, and routine clinical variables. OvcaFinder outperforms both the clinical model and the DL model, achieving area under the curves (AUCs) of 0.978 and 0.947 in internal and external test datasets, respectively. It significantly improves radiologists' diagnostic performance, with average AUCs increasing from 0.927 to 0.977 and false positive rates decreasing by 13.4% and 8.3% in internal and external datasets, respectively. OvcaFinder also enhances inter-reader agreement, demonstrating excellent agreement in both datasets. The model's interpretability is enhanced through heatmaps and Shapley values, highlighting key features contributing to its predictions. This study highlights the potential of OvcaFinder as a non-invasive tool to improve the accuracy and consistency of radiologists in diagnosing ovarian cancer.Ovarian cancer is a highly lethal gynecological malignancy, and early and accurate diagnosis is crucial. This study presents OvcaFinder, an interpretable model that integrates ultrasound images-based deep learning (DL) predictions, Ovarian-Adnexal Reporting and Data System (O-RADS) scores from radiologists, and routine clinical variables. OvcaFinder outperforms both the clinical model and the DL model, achieving area under the curves (AUCs) of 0.978 and 0.947 in internal and external test datasets, respectively. It significantly improves radiologists' diagnostic performance, with average AUCs increasing from 0.927 to 0.977 and false positive rates decreasing by 13.4% and 8.3% in internal and external datasets, respectively. OvcaFinder also enhances inter-reader agreement, demonstrating excellent agreement in both datasets. The model's interpretability is enhanced through heatmaps and Shapley values, highlighting key features contributing to its predictions. This study highlights the potential of OvcaFinder as a non-invasive tool to improve the accuracy and consistency of radiologists in diagnosing ovarian cancer.