Fig. 3

Challenges and solutions of multiomics combination. (1) The majority of current studies based on radiomics and other multiomics are small-scale, single-center, and retrospective, lacking adequate external validation. In the future, there is a need for large-scale prospective multicenter studies. (2) Presently, research on tumors predominantly focuses on single-omics studies using a singular data type. Despite the challenges in acquiring multiomics data, it is viable to collect such data through well-designed prospective studies, facilitating the integration of multimodal data for in-depth and comprehensive analysis. (3) Multiomics research faces the additional challenge of dealing with high dimensionality and heterogeneity in the data. Standardizing and fusing the multiomics data is necessary. The fusion of multiomics can be effectively achieved through machine learning techniques. (4) The processes of radiomics and pathomics are more time-consuming regarding image segmentation. Models that have a lower level of reproducibility and study results that exhibit weaker generalization. In the future, it is imperative to develop automated image segmentation methods, standardize and normalize the research process, and incorporate external validation to ensure the robustness of the findings