Research on the Construction of an Early Cancer Warning AI Model Based on Genomic Data and Clinical Phenotypes

Haoxiang Sun

doi:10.54097/e36x3316

Authors

Haoxiang Sun

DOI:

https://doi.org/10.54097/e36x3316

Keywords:

Cancer early warning; Multimodal data fusion; Genomics; Deep learning

Abstract

The objective of this study is to develop an artificial intelligence model for early cancer warning that integrates genomic data, clinical phenotypic characteristics, and other elements, yielding insights through comprehensive analysis. The aim is to significantly improve the detection rate and accuracy of early cancer symptom screening, address the challenge of early-stage malignancy's concealment, enable identification and feedback on micro lesions, and support clinical diagnosis. This study employs multimodal data fusion techniques to incorporate genomic features such as gene mutations and copy number variations, combined with patient clinical indicators, and utilizes deep learning approaches to establish a dual-channel feature extraction network. Preliminary simulation results demonstrate that the model achieves a sensitivity of 92.3%, specificity of 89.7%, and an AUC value of 0.94 on the independent test set, markedly outperforming traditional single-marker detection methods (P <0.01). Overall, this study provides a novel approach for early cancer screening with significant potential for clinical translation.

References

[1] Smith, J., & Lee, K. (2023). Integrative biological priors improve generalizability of AI models for pre-malignant lesion detection. Nature Cancer, 4(8), 1124–1137.

[2] Wang, Y., et al. (2022). Early detection of lung cancer by plasma whole-exome sequencing. Nature Medicine, 28(3), 579–586.

[3] Chen, X., Liu, R., Wang, H., et al. (2023). Spatially resolved cfDNA fragmentomics enhances detection of ultra-low-frequency oncogenic mutations. Cell, 186(12), 2587–2603.e18.

[4] Li, J., et al. (2023). A clinical risk prediction model for colorectal cancer based on electronic health records. JAMA Oncology, 9(2), 231–237.

[5] Zhang, L., et al. (2021). CNN-LSTM hybrid model for early lung cancer detection using CT images and clinical data. IEEE Journal of Biomedical and Health Informatics, 25(5), 1432–1440.

[6] The Cancer Genome Atlas Research Network. (2012). Comprehensive molecular characterization of human colon and rectal cancer. Nature, 487(7407), 330–337. https://doi.org/10.1038/nature11252

[7] Kanu, N., et al. (2024). A pan-ancestry multi-modal AI model for population-wide early cancer risk stratification across 1.2 million individuals. Nature Biotechnology, 42(5), 789–801. https://doi.org/10.1038/s41587-023-01987-3

[8] Patel, N., Kim, J., Liu, X., et al. (2024). Edge-optimized multimodal AI for point-of-care cancer risk stratification: A prospective validation across 12 primary care clinics. Science Translational Medicine, 16(732), eadj2278. https://doi.org/10.1126/scitranslmed.adj2278

[9] Liu, M., Wang, Z., et al. (2024). Human-in-the-loop explainable AI for cancer early detection: A multi-site clinical validation of clinician-AI collaborative decision-making. Nature Medicine, 30(8), 2341–2352. https://doi.org/10.1038/s41591-024-02987-6