 Sök
och bläddra bland Karolinska Institutets avhandlingar |
|
|||||||||||||||||||||||||||||||||||
Doktorsavhandling vid Karolinska Institutet |
|
Szabo, Botond KDynamic magnetic resonance imaging of the breast : Imaging features for diagnosis and prognosis of breast cancerFredagen den 19 november 2004, kl. 9.00. Föreläsningssal M63, Karolinska Universitetssjukhuset, Huddinge. |
|
| ISBN: 91-7140-089-3 | Diss: 04:485 |
Abstract:
Study I studied the kinetic and morphological MR features and their diagnostic power were studied in a total of 92 women with 109 histopathologically verified breast lesions. Logistic regression analysis was performed in order to select the most important diagnostic criteria. Time-to-peak enhancement and margin characteristics appear to be the most relevant criteria, and a scoring system based on these parameters showed a comparable diagnostic performance to the routine interpretation approach.
Study II was based on the same material as presented in Study I, but focused on the applicability of artificial neural networks (ANNs). Neurostatistical analysis confirmed the findings of Study 1, in addition, wash-out parameters also turned out to be relevant criteria. Morphologic enhancement features are of importance in the evaluation of breast masses, but their value is inferior to kinetic information. ANN is a valuable tool in analyzing complex data sets, and we found it valuable in supporting the diagnosis in dynamic breast MR imaging.
Study III correlated MR features to prognostic factors of breast cancer in 61 patients. Univariate and multivariate analyses were performed. Presence of rim-enhancement, early maximal enhancement and wash-out phenomenon were independently associated with established predictors of poor prognosis (histologic grade, Ki-67, ER status). Our results suggest that these MR signs may be useful to non-invasively identify highly aggressive breast carcinomas.
Study IV investigated the value of preoperative CE-MRI in predicting disease-free and overall survival in 50 consecutive patients with breast cancer. The median follow-up for surviving patients was 95 months. Among the examined prognostic factors, signal enhancement ratio (SER), and tumor size were independently related to disease-free survival at multivariate analysis. These results suggest that CE-MRI can predict disease-free survival and may be useful as a prognostic tool.
In Study V, an ANN based segmentation method was developed for dynamic CE-MRI of the breast and compared to quantitative and empiric parameter mapping techniques. ANN was successfully applied to the classification of breast MR images identifying structures with benign or malignant enhancement kinetics. Correlation coefficient to a reference wash-out curve, pharmacokinetic parameter kep, and time-to-peak were independently associated to ANN output classes.
Conclusion: Through its ability to monitor both morphological and dynamic features of breast tumors, contrast media enhanced MRI is suitable to diagnose and differentiate breast lesions. Computerized technology including the use of neural networks may facilitate the interpretation of the data collected from the MRI diagnosis, and can be a useful complementary tool to the radiologist. CE-MRI is not only an excellent tool for diagnosis, but can also predict disease-free survival and may be a useful tool in exploring the prognosis for patients with breast cancer.
Keywords: Breast neoplasm, magnetic resonance imaging, prognostic factors, survival analysis, image segmentation, artificial neural networks
List of papers
 |
Dynamic MR imaging of the breast. Analysis of kinetic and morphologic diagnostic criteria. Szabo BK, Aspelin P, Wiberg MK, Bone B Acta Radiol, 2003; 44(4): 379-86 |
|
Application of artificial neural networks to the analysis of dynamic MR imaging features of the breast. Szabo BK, Wiberg MK, Bone B, Aspelin P Eur Radiol, 2004; 14(7): 1217-25. Epub 2004 Mar 18 |
|
Invasive breast cancer: correlation of dynamic MR features with prognostic factors. Szabo BK, Aspelin P, Kristoffersen Wiberg M, Tot T, Bone B Eur Radiol, 2003; 13(11): 2425-35. Epub 2003 Jul 26 |
|
Can contrast-enhanced MR imaging predict survival in breast cancer? Bone B, Szabo BK, Perbeck LG, Veress B, Aspelin P Acta Radiol, 2003; 44(4): 373-8 |
|
Neural network approach to the segmentation and classification of dynamic MR images of the breast: comparison with empiric and quantitative kinetic parameters. Szabo BK, Aspelin P, Kristoffersen Wiberg M Academic Radiology, Accepted |
