Abstract： Objective To evaluate the accuracy of cone?beam breast CT (CBBCT) on tumor sizing in patients with invasive breast carcinoma and analyze the influence factors. Methods The preoperative CBBCT images of 82 female patients (85 breast lesions) with invasive breast carcinoma confirmed by pathology from November, 2011 to December, 2017 in Tianjin Medical University Cancer Hospital were included in this retrospective study. All the patients underwent the pathology and immunohistochemical test after operation. Tumor size estimation were performed on preoperative CBBCT images. Referring to tumor size measurement on pathology, all the lesions were divided into two groups. Concordance was defined as the discrepancy in diameter less than 0.5 cm, and the discordance was more than 0.5 cm. Pathology examination was performed after resection, and estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2(HER?2) and Ki?67 result were recorded. All the lesions were classified into molecular subtype, including 14 Luminal A, 50 Luminal B, 11 HER?2?enriched and 10 triple?negative. Intraclass correlation coefficient (ICC) and Pearson correlation coefficient were used to analyze the reliability of CBBCT on tumor sizing. CBBCT?pathology discordance was analyzed based on the clinical, histopathology and CBBCT features by using t test, Chi?square and Fisher exact test. ROC curve was used to analyze the cut?off value between tumor size and CBBCT?pathology discordance. Results The agreement between CBBCT (2.155 ± 0.799) cm and pathology (1.986 ± 0.933) cm measurement was on moderate degree based on the ICC value (ICC=0.781, P<0.01) and had positive correlation (r=0.803, P<0.01). CBBCT?pathology concordance was found in 71 lesions, and discordance in 14 lesions. The factors of family history, symptom, pathology type, molecular subtypes, histological grade, surrounding fat invasion, lymphatic invasion, axillary lymph node metastasis, HER?2 positive and Ki?67 high expression had no significant difference between two groups. ER or PR positive had significant difference, suggesting that the accuracy of evaluation on ER or PR negative lesions could be reduced. The cut?off value of ROC curve between tumor size and CBBCT?pathology discordance was 2.08 cm, and the area under curve was 0.70. In 85 lesions. 66 of 75 mass lesions and 5 of 10 non?mass lesions were consistent. The lesion type had significant difference between two groups (χ2=6.705, P=0.010), which suggested the CBBCT evaluation on non?mass could have discrepancy with pathology. Conclusion CBBCT has high accuracy on tumor size evaluation on invasive carcinoma. ER or PR negative, large lesions or non?tumor lesions should be alert to the impact of CBBCT?pathology discordance before surgery which may cause the alteration of surgical protocols.