Weimin Xu,1 Xinyu Wang,2 Lei Yang,2 Muzi Meng,3,4 Chenyu Sun,1 Wanwan Li,1 Jia Li,1 Lu Zheng,1 Tong Tang,1 WenJun Jia,1 Xiao Chen1 

1Department of Thyroid and Breast Surgery, The Affiliated Second Hospital of Anhui Medical University, Hefei, People’s Republic of China; 2School of Second Clinical Medicine, AnHui Medical University, Hefei, People’s Republic of China; 3School of Medicine, American University of the Caribbean, Sint Maarten, Kingdom of the Netherlands; 4General Surgery, BronxCare Health System, New York, NY, USA

Correspondence: WenJun Jia; Xiao Chen, Email jiawenjun@smail.nju.edu.cn; chenxiao950616@Gmai.com

Background: The Chinese Society of Clinical Oncology Artificial Intelligence System (CSCO AI) serves as a clinical decision support system developed utilizing Chinese breast cancer data. Our study delved into the congruence between breast cancer treatment recommendations provided by CSCO AI and their practical application in clinical settings.
Methods: A retrospective analysis encompassed 537 breast cancer patients treated at the Second Affiliated Hospital of Anhui Medical University between January 2017 and December 2022. Proficient senior oncology researchers manually input patient data into the CSCO AI system. “Consistent” and “Inconsistent” treatment categories were defined by aligning our treatment protocols with the classification system in the CSCO AI recommendations. Cases that initially showed inconsistency underwent a second evaluation by the Multi-Disciplinary Treatment (MDT) team at the hospital. Concordance was achieved when MDTs’ treatment suggestions were in the ‘Consistent’ categories.
Results: An impressive 80.4% concurrence was observed between actual treatment protocols and CSCO AI recommendations across all breast cancer patients. Notably, the alignment was markedly higher for stage I (85.02%) and stage III (88.46%) patients in contrast to stage II patients (76.06%, P=0.023). Moreover, there was a significant concordance between invasive ductal carcinoma and lobular carcinoma (88.46%). Interestingly, triple-negative breast cancer (TNBC) exhibited a high concordance rate (87.50%) compared to other molecular subtypes. When contrasting MDT-recommended treatments with CSCO AI decisions, an overall 92.4% agreement was established. Furthermore, a logistic multivariate analysis highlighted the statistical significance of age, menstrual status, tumor type, molecular subtype, tumor size, and TNM stage in influencing consistency.
Conclusion: In the realm of breast cancer treatment, the alignment between recommendations offered by CSCO AI and those from MDT is predominant. CSCO AI can be a useful tool for breast cancer treatment decisions.

Keywords: Chinese Society of clinical oncology artificial intelligence system, CSCO AI, breast cancer, multi-disciplinary treatment, MDT, consistency