Fig. 1: Overview of the entire research process. | npj Precision Oncology

Fig. 1: Overview of the entire research process.

From: Integration of radiomics, deep learning, transcriptomics, and metabolomics reveals prognostic risk stratification and underlying biological mechanisms in colorectal cancer

Fig. 1: Overview of the entire research process.The alternative text for this image may have been generated using AI.

The study framework was organized into five main phases. A Data collection: Data were obtained from multiple sources to support different analytical stages. B Risk stratification: Radiomics and DL features from CT images were evaluated using 117 machine learning algorithm combinations to identify the optimal model for patient risk stratification. In parallel, three models (clinical, DLRM-RS, and combined models) were constructed and evaluated for performance comparison. C Metabolomic analysis: NMR-based profiling identified differential metabolic signatures between risk groups. D Transcriptomic analysis: Transcriptomic differences and immune infiltration patterns between risk groups were explored using TCIA data. E Pathway validation: Two commonly enriched pathways (butanoate metabolism and nitrogen metabolism) were identified in both transcriptomic and metabolomic analyses. Their prognostic significance was confirmed using GSVA scoring and Kaplan-Meier analysis in the TCGA cohort. CRC colorectal cancer, DLRM deep learning-radiomics model, DL deep learning, TCIA The Cancer Imaging Archive, TCGA The Cancer Genome Atlas, NMR Nuclear Magnetic Resonance, GSVA Gene Set Variation Analysis.

Back to article page