Study On Prediction Of Neoadjuvant Chemotherapy Response For Breast Cancer Based On DCE-MRI

Breast cancer is a common malignancy of women. The morbidity of breast cancer in the countryis still an increasing trend in recent years, and is still the highest morbidity of malignancy in female.Neoadjuvant chemotherapy (NACT) for breast cancer has become a criteria model for breast cancertreatment at present. NACT can not only improve local tumor treatment, also can be controlledexisting micrometastases. NACT can reduce the size of cancer, breast-conserving aspects show a lotof advantages by a large scale of randomized clinical trials, and clinical studies have demonstratedthat after NACT treatment the survival of patients with pathologic complete remission significantlyimproved.The dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) have a high sensitivityto detect small, multi-center and multi-focal lesions, and the three-dimensional imaging allows moreaccurate to locate lesions position. Besides, DCE-MRI can identify breast cystic and real lesion part,imaging of both breast at the same time with no radiation and so on. Due to the DCE-MRI has somany advantages, it become the popular breast cancer examination of current clinical evaluation ofthe NACT.Research work of this paper mainly focus on the ability of overall dynamic features, lesionbackground dynamic features and the lesion dynamic features of breast DCE-MRI to predict theclinical response of the NACT. The content includes:(1) According to the characteristics of breast DCE-MR images, our segmentation method isproposed based on the combination of breast-horizontal and breast-sagittal. Then according to theDCE-MR images characterized by a sequence, taking the advantages of the three-dimensionaladjacent correlation between images to correct the wrong sub-site. By using of this method to segmentbreast DCE-MR images can be maximize retain the breast glandular tissue that provide a moreaccurate for subsequent studies.(2) In this study, we proposed a method to calculate the three-dimensional dynamic enhancedfeatures of overall area, lesion background area and the lesion area of the breast. All dynamic featuresare calculated by statistical methods, including an enhanced rate of the mean, variance and maximumat each time point and the bilateral asymmetry dynamic features based on left and right breast. Thesefeatures can be a good characterization of the relevant information of overall area, lesion backgroundarea and lesion area.(3) In order to investigate the value of the breast overall dynamic features, lesion backgrounddynamic features and lesion dynamic features in predicting the clinical response of the NACT, our study take a deep investigate of the ability of these three feature types in predicting the clinicalresponse of the NACT. Each feature have done divisibility assessment. And we used geneticalgorithm to select the optimal subset from all features. The encouraging results shown that the breastoverall dynamic features, the lesion background dynamic features and the lesion dynamic featurescan tell which case contain related information that could predict the clinical response of the NACTby objective. That means breast DCE-MR images does contained related information that couldpredict clinical response of the NACT.(4) According to the compilation of DCE-MR imaging with other clinical examinations in theclinical response of the NACT, the results showed that in all examination methods of clinicalevaluation of the NACT in breast DCE-MR imaging is the better valuable method to examine theclinical response of the NACT than other clinical examination.