| ObjectiveBased on real-time dynamic image artificial intelligence system to carry out research.To analyze and compare the diagnostic efficacy of real-time dynamic artificial intelligence system and sonographers with different seniority in benign and malignant breast diseases by comprehensive three-dimensional observation of breast lesions.The value of the artificial intelligence system in assisting doctors with different seniority in the diagnosis of benign and malignant breast diseases was comprehensively analyzed.The ultimate goal is to improve the correct rate of ultrasound diagnosis of breast diseases.Materials and MethodsA total of 194 patients who visited the First Hospital of Dalian Medical University for ultrasound-guided aspiration biopsy or conventional surgery for breast nodules with clear pathological diagnosis between September 2020-July 2022 were screened.All patients were female,the total number of breast lesions was 210,the age span of patients was 18-88 years,the mean age was(49.03±12.72)years.A total of 66 cases with malignant lesions,the age span of the patients was 36-88 years,with a mean age of(57.55±17.50)years.Benign lesions were 144 patients with an age span of 18-69 years and a mean age of(45.12±10.90)years.The ultrasound diagnostic instrument used in this study was Philips EPIQ-7,and the ultrasound probe was a superficial linear array probe with model L12-5.The AI system is a product of Beijing Yi Zhun Intelligent Technology Co.,LTD.The AI system is connected to and matched with the ultrasound instrument.During the ultrasound examination,the patient was instructed to lie in the supine position,with arms as far as possible,and to expose the complete bilateral breasts and armpits as much as possible.After that,a junior physician(working time < 2 years),a middle senior physician(working time 2-10 years),and a senior physician(working time > 10years)radially scanned each quadrant gland and bilateral axillary lymph nodes with the nipple as the center to find out the corresponding breast lesions and confirm the BI-RADS classification results.Then,the AI system entered the lesion analysis function,and the lesion was scanned by another physician by longitudinal and transverse sections,respectively.Each scan started from one end of the tumor to the other end.Finally,the BI-RADS classification results of real-time dynamic scanning lesions under the AI system were obtained.The results of the BI-RADS classification and the ultrasound description of the breast nodes determined by the senior,middle-aged and junior physicians were recorded separately.Ultrasound description specifically including the specific location,size,aspect ratio,borders,margins,morphology,internal echogenicity,posterior acoustic shadow,calcification and internal and peripheral blood flow of the breast nodule.In addition,the BI-RADS classification results obtained by the AI system scanning were recorded,and all ultrasound descriptions under the AI system will be automatically generated and saved after the completion of breast nodule scanning.According to the latest edition of BI-RADS Ultrasound dictionary formulated by the American College of Radiology in 2013,the malignancy rate of class 4A nodules was 2%-10%,indicating low suspicious malignancy;The malignancy rate of nodules in class 4B was 10%-50%,indicating moderate suspicious malignancy.The nodules of category 4A and below(including category 3 and 4A)were regarded as benign nodules,and nodules of category 4B and above(including category 4B,4C and 5)were regarded as malignant nodules.Result1.The sensitivity,specificity,accuracy,Youden index,negative predictive value,and positive predictive value of AI in the diagnosis of benign and malignant breast diseases were 89.40%,86.81%,87.62%,76.21%,91.24%,and 75.64%,and the area under the ROC curve(AUC)was 0.881.2.The diagnostic values of junior doctors in the diagnosis of benign and malignant breast diseases were 90.77%,71.72%,78.10%,62.49%,95.54% and 59.00%,respectively,and the AUC was 0.816.The diagnostic values of middle-aged doctors were 93.94%,80.56%,84.76%,74.50%,96.67%,68.89%,and the AUC value was0.852.The diagnostic values of senior physicians were 96.97%,88.19%,90.95%,85.16%,98.45%,79.01%,and the AUC value was 0.926.The specificity,accuracy,Youden index and positive predictive value of the AI system in the diagnosis of benign and malignant breast diseases were higher than those of low and middle seniority doctors,but lower than those of senior doctors.3.The AUC of the real-time dynamic image-based AI system in diagnosing benign and malignant breast diseases was higher than that of junior and middle senior doctors,and lower than that of senior doctors,and the differences were statistically significant(P< 0.01).ConclusionThe diagnostic efficiency of the AI system based on real-time dynamic images for benign and malignant breast diseases is significantly better than that of low and middle seniority doctors,and slightly lower than that of senior doctors.It can effectively help low and middle seniority ultrasound doctors to improve the accuracy of breast cancer diagnosis,and it also has a certain auxiliary effect on the diagnosis of senior doctors.Therefore,the diagnostic quality of conventional breast ultrasound can be improved. |
PDF Full Text Request