Data-Driven Multi-Modal Cancer Prognosis Using Machine Learning and Deep Learning Techniques
DOI:
https://doi.org/10.63949/Keywords:
- Machine learning,
- Deep learning,
- Stacking ensemble,
- Survival analysis,
- SHAP interpretability,
- Cancer prognosis
Abstract
Proper prognosis of cancer is essential to enable individual treatment and enhance survival of cancer patients. Cox regression, and other more traditional statistical models, are frequently inadequate to address complex nonlinear interactions between clinical, genomic and imaging features. This paper presents a hypothesized data prognosis model to combine multi-modal clinical and genomic data utilizing conventional ML, DL, and ensemble methods. The stacking ensemble model was created, which integrates Support Vector Machine, Random Forest, XGBoost, and ANN and tested through stratified 10-fold cross-validation with SMOTE as a solution to the issue of class imbalance. Performance measures, ROC-AUC and Concordance Index of survival models. It has been found that the stacking ensemble obtained 94.1% accuracy and 0.97 ROC-AUC which is much higher compared to the performance of the individual models and DeepSurv obtained the best performance on survival prediction (C-index = 0.83). The feature importance analysis using SHAP placed the top prognostic factors as tumour stage, age, lymph node involvement, and key gene expressions. Learning curves showed low levels of overfitting and Kaplan-Meier survival curves showed effective risk stratification. The suggested framework provides a powerful, interpretable, and clinically useful approach to predicting early cancer risks to support individual treatment planning and distribution of resources.
Downloads
References
[1] A. Mahmoud, M. Alhussein, K. Aurangzeb and E. Takaoka, Breast Cancer Survival Prediction Modeling on Genomic Data: an Improved Prognosis-driven Deep Learning, in IEEE Access, vol. 12, pp.119502-119519, 2024, doi: 10.1109/ACCESS.2024.3449814.
[2]. Oncology Iqbal, M.J., Javed, Z., Sadia, H. et al. Clinical applications of artificial intelligence and machine learning in cancer diagnosis: looking into the future. Cancer Cell Int 21, 270 (2021). https: doi.org/10.1186/s12935-021-01981-01.
[3]. Kumar, Y., Gupta, S., Singla, R. et al. Artificial Intelligence in Cancer Prediction and Diagnosis: A Systematic Review of Techniques. Gergis et al., 2022 (2022). https://doi.org/10.1007/s11831-021-09648-w. Arch Computation Methods Eng 29, 20432070.
[4]. Lobato-Delgado, B., Priego-Torres, B., and Sanchez-Morillo, D. (2022). Integrating Molecular, Imaging and Clinical Data Analysis to predict cancer prognosis. Cancers, 14(13), 3215.
[5]. Mihaylov, I., Kańdulla, M., Krachunov, M. et al. A new structure of horizontal and vertical data integration in cancer research with implementation to models of survival time prediction. Biol Direct 14, 22 (2019). https://doi.org/10.1186/s13062-019-0249-6
[6]. Nikolaou, N., Salazar, D. RaviPrakash, H. et al. A multimodal data fusion machine learning method to predict survival in cancer patients. npj Precis. Onc. 9, 128 (2025). https://doi.org/10.1038/s41698-025-00917-6
[7]. Kundu, T. K., Das, S., & Nidhya, R. (2025). Detection of peripheral blood smear malarial parasitic microscopic images utilizing convolutional neural network. Generative Artificial Intelligence: Concepts and Applications, 121-142.
[8]. Othman, N. A., Abdel-Fattah, M. A., and Ali, A. T. (2023). Deep Hybrid Layer-based Breast Cancer Survival Prediction at the Decision Level. Big Data and Cognitive Computing, 7(1), 50. doi.org/10.3390/bdcc7010050.
[9]. Sharma, D., Deepali, Garg, V.K. et al. Deep learning-based integrative survival time prediction of head and neck squamous cell carcinoma patients. Neural ComputApplic 34, 2135321365 (2022). https:doi.org/10.1007/s00521-022-07615-5.
[10]. Pavithra, D., Nidhya, R., Vinothini, C., & Murugaiyan, M. (2023). A Lightweight Attention based MobileNetv2 Model for Brain Tumor Segmentation and Severity of Tumor Classification using Support Vector Machine.
[11]. Tiwari, A., Mishra, S. &Kuo, TR. Existing AI systems in cancer diagnosis and treatment. Mol Cancer 24, 159 (2025). https://doi.org/10.1186/s12943-025-02369-9
[12]. Wekesa JS and Kimwele M (2023) A review of disease diagnosis, prognosis, and treatment Multi-omics data integration using deep learning models. Front. Genet. 14:1199087. doi: 10.3389/fgene.2023.1199087
[13]. Yongfei Hu, Xinyu Li, Ying Yi, Yan Huang, Guangyu Wang, Dong Wang, Deep learning-based survival prediction in pan-cancer studies through the combination of multimodal histology-genomic data, Briefings in Bioinformatics, Volume 26, Issue 2, March 2025, bbaf121, https:/doi.org/10.1093/bib/bbaf121.
[14]. Zehua Wang, Ruichong Lin, Yanchun Li, Jin Zeng, Yongjian Chen, Wenhao Ouyang, Han Li, XueyanJia, Zijia Lai, Yunfang Yu, Herui Yao, Weifeng Su, Deep learning-based multi-modal data integration enhancing breast cancer disease-free survival prediction, Precision Clinical Medicine, Volume 7, Issue 2, June 2024, pbae012, p. doi.org/10.1093/pcmedi/
