1. 北京大学第三附属医院海淀院区（北京市海淀医院） a. 放射科|b. 胸外科，北京 100080|
2. 北京推想科技有限公司 先进研究院，北京 100025
Research on the Correlation between Pathological Findings of Multiple Primary
Lung Cancer and AI-Assisted CT Diagnosis
LI Dasheng1a，WANG Dawei2，HUANG Yuqing1b，LIU Xiaoxu1a，HUO Zhiyi1a
1. a. Department of Radiology| b. Department of Thoracic Surgery, Beijing Haidian Section of Peking University Third Hospital (Beijing
Haidian Hospital), Beijing 100080, China| 2. Institute of Advanced Research, Infervision Technology Co., Ltd, Beijing 100025, China
Abstract：Objective To explore the clinical application value of artificial intelligence (AI) in the diagnosis of multiple primary
lung cancer (MPLC) by analyzing the consistency between the diagnoses based on AI assistance and the clinical pathological results.
Methods Twenty-six patients with MPLC who were admitted to Beijing Haidian Hospital from 2017 to 2019 were enrolled in this
study. A total of 57 cancer lesions were found in these enrolled patients and divided into 2 groups according to their pathological
grades, including group 1 (0 and T1a1 stages) and group 2 (T1a2, T1a3, T1b, T3a, and T3b stages). AI-assisted diagnostic system was
utilized to measure the size and density of nodules quantitatively. The χ2-test was utilized to study the correlation between pathological
groups (binary classification) and AI system predicted nodule types. In addition, linear correlation and regression analysis were
performed between pathological grades (0, T1a1, T1a2, T1a3, T1b, T3a, and T3b stages) and AI system-output measurements.
Results A significant correlation between AI predicted nodule types and pathological groups (binary classification) was observed
(P<0.05). In addition, pathological grades (0, T1a1, T1a2, T1a3, T1b, T3a, and T3b stages) was shown to be significantly correlated
with the nodule volume, the longest and shortest diameter of nodules measured by the AI system (P<0.001)| the longest diameter
measurement value of nodules increased with the advancement of pathological grades| the nodule volume and the longest diameter
measured by AI system increased along with the growth of tumor area as well. The volume of nodules measured by AI system also
increased with the advancement of pathological stages (model 3) even after the adjustment of tumor area| meanwhile, the longest
dimension of nodules and the longest diameter measured by the AI system increased along with the increase of tumor area even after
the adjustment of pathological grades| in contrast, the measured volume of nodule by AI system decreased along with the increase
of tumor area. Conclusion The AI diagnostic system displayed a decent diagnostic performance for multiple primary lung cancer
(MPLC) in different pathological grades evidence by the strong consistency to clinical pathological diagnosis results. In clinical
imaging diagnoses, suspicious malignant nodes should be focused and followed up intensely by referring to AI system predicted
results. In addition to AI measured volume of nodules, other signs and features could suggest the possibilities of MPLC incidence and
improve the detection rate of MPLC in clinical practice.
李大胜1a，王大为2，黄宇清1b，刘晓旭1a，霍志毅1a. 多原发肺癌病理结果与AI辅助CT诊断的相关性研究[J]. 中国医疗设备, 2021, 36(2): 77-80.
LI Dasheng1a，WANG Dawei2，HUANG Yuqing1b，LIU Xiaoxu1a，HUO Zhiyi1a. Research on the Correlation between Pathological Findings of Multiple Primary
Lung Cancer and AI-Assisted CT Diagnosis. China Medical Devices, 2021, 36(2): 77-80.
