Advanced Multimodal Deep Learning Enhances Risk Prediction for Cervical Cancer Radiotherapy

Recent advancements in artificial intelligence have led to the development of a novel multimodal deep learning model aimed at improving risk stratification in cervical cancer radiotherapy. Currently, standard concurrent chemoradiotherapy (CCRT) achieves disease-free survival (DFS) in about 70% of patients with locally advanced disease, but nearly 30% still face recurrence or metastasis. While intensified treatment approaches may boost survival rates, they often come with increased toxicity and costs, underscoring the importance of accurately identifying patients who truly require aggressive therapy.

A multidisciplinary team led by Associate Professor Liang Xiaokun from the Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, alongside Prof. Hu Ke and Prof. Hou Xiaorong from Peking Union Medical College Hospital, introduced CerviPro. This deep learning-based prognostic model integrates diverse data types, including pre- and post-radiotherapy CT imaging, radiomics features, and clinical information, to provide personalized risk assessments.

CerviPro employs advanced automatic segmentation techniques to precisely delineate tumor regions, leveraging a pre-trained CT foundation model to extract high-dimensional features. These features are then fused using principal component analysis (PCA) to reduce dimensionality and enhance feature relevance. To validate its clinical utility, researchers compiled data from over 1,000 cervical cancer patients across multiple Chinese hospitals. The model demonstrated high accuracy in predicting prognosis across various datasets, outperforming traditional models like Cox proportional hazards and DeepSurv.

Importantly, CerviPro effectively stratifies patients into high-risk and low-risk groups, guiding clinicians in tailoring treatment strategies—potentially escalating therapy for high-risk cases while de-escalating for low-risk ones. This innovative approach provides a reliable decision-support tool, advancing personalized medicine in cervical cancer care.

For more details, see the original study published in npj Digital Medicine.