Application of artificial intelligence ultrasound in breast cancer screening in Shigatse, Tibet
10.3760/cma.j.cn131148-20220319-00186
- VernacularTitle:应用人工智能辅助乳腺癌超声筛查在西藏日喀则地区的初步研究
- Author:
Yongzhi HOU
1
;
You TIAN
;
Zhen BAI
;
Ma MI
;
Yulin WU
;
Xun ZHANG
;
Cai CHANG
;
Shichong ZHOU
Author Information
1. 日喀则市人民医院超声医学科，日喀则 857000
- Keywords:
Ultrasonography;
Artificial intelligence;
Breast cancer screening
- From:
Chinese Journal of Ultrasonography
2022;31(11):927-932
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To explore the feasibility and clinical value of artificial intelligence-assisted breast ultrasound in screening breast cancer in Tibet.Methods:Two hundred and eighty-six women who participated in breast cancer screening in Shigatse People′s Hospital from August to September in 2021 were selected. The study included four groups. Group 1, ultrasound screening by senior breast ultrasound doctors from Shanghai; Group 2: local ultrasound doctors used intelligent-assisted ultrasound equipment for screening; Group 3: local ultrasound technicians used intelligent-assisted ultrasound equipment for screening; Group 4: ultrasound screening by local ultrasound doctors. The pathological results of screening positive cases and six-month ultrasound follow-up results of negative cases were set as the gold standard.Results:Twenty-seven lesions of 21 persons were screened positive. Pathology showed that 1 case of invasive ductal carcinoma, 1 case of severe atypical hyperplasia, 6 cases of fibroadenoma, 5 cases of breast disease, 14 cases of breast hyperplasia. Two hundred and sixty-five persons were screened negative, and the results of the six-month ultrasound follow-up were still negative. The accuracy, sensitivity, and specificity of group 2 were 0.966, 1, and 0.964 respectively; The accuracy, sensitivity, and specificity of group 3 were 0.935, 0.769, and 0.943 respectively; The accuracy, sensitivity, and specificity of group 4 were 0.860, 0.308 and 0.885 respectively. The accuracy and area under the curve of groups 2 and 3 were significantly different from that of group 4 (all P<0.001), and there was no significant difference from that of group 1 ( P=0.063, P=0.055). Conclusions:Artificial intelligence-assisted breast ultrasound screening technology can effectively improve the screening efficiency of non-breast ultrasound specialists and technicians. It is very suitable to solve the problems faced by grass-roots screening in Tibet and has great social significance and clinical value.
