The study analyzed CT imaging characteristics of bone erosion in rheumatoid arthritis (RA) and bioinformatically investigated inflammation-related genes regulated by rG4 structures. Clinical data, CT images, and peripheral blood RNA-seq data from 49 RA patients at Yancheng Third People's Hospital and from healthy controls were collected. Differentially expressed RA-related genes (irDEGs) were screened using DESeq2 software, 67 irDEGs were identified, of which 42 contained potential rG4 structures. A U-Net CNN deep learning model was developed for automatic segmentation and quantification of bone erosion severity in CT images with high accuracy, cube similarity coefficient (DSC), sensitivity, and specificity. The quantitative model score was significantly correlated with the clinical disease activity score (DAS28). CT characteristics of bone erosion in RA patients were closely associated with the expression of rG4-regulated irDEGs. The model provided an accurate method for the clinical assessment of bone erosion and candidate targets for studying the molecular mechanisms of bone destruction.