PURPOSE: The value of computer-assisted detection (CAD) used in magnetic resonance (MR) mammography in a clinical setting is currently a subject of controversy. This study evaluated the extent to which color-coded CAD systems aid radiologists with different levels of experience in their reading of MR mammographies. MATERIALS AND METHODS: In this prospective study, 48 patients with a total of 88 lesions (43 malignant, 45 benign) were included. All examinations were performed on a 1.5 Tesla MR scanner with intravenous application of 0.1 mmol gadopentetate dimeglumine/kg body weight. Three readers independently analyzed the images without knowledge of the clinical data; radiologists 1 and 2 were much more experienced in the interpretation of MR mammographies than radiologist 3. Initially, the observers visually categorized the lesions as benign or malignant following classification of BI-RADS (Breast Imaging Reporting and Data System). The readers also scored their own confidence level using a dichotome score (1: unsure vs. 2: sure) according to the BI-RADS classification. The images were then analyzed in a blinded manner with two technically different CAD systems: the full-time point (FTP) method (Cadsciences; White Plains, NY, USA) and the Dynacad version 1.1 (Invivo; Pewaukee, WI, USA). RESULTS: After CAD, all three readers classified more malignant lesions as BI-RADS 4 or 5 (suspicious or highly suggestive of malignancy). However, this increase in sensitivity revealed only statistical significance for observer 3 (p < 0.05). After CAD, the two experienced readers categorized about the same quantity of benign lesions correctly as BI-RADS 2. Observer 3 classified less benign lesions as BI-RADS 2 after both CAD analyses; i. e. the specificity decreased. The subjective confidence of all observers increased after analysis with Dynacad. Observer 3 also reported to be more confident after the FTP method. CONCLUSION: Computer-assisted detection can be a useful additional diagnostic tool for the radiologist in the interpretation of MR mammographies, but does not have the potential to replace the professional experience of a radiologist.
PURPOSE: The value of computer-assisted detection (CAD) used in magnetic resonance (MR) mammography in a clinical setting is currently a subject of controversy. This study evaluated the extent to which color-coded CAD systems aid radiologists with different levels of experience in their reading of MR mammographies. MATERIALS AND METHODS: In this prospective study, 48 patients with a total of 88 lesions (43 malignant, 45 benign) were included. All examinations were performed on a 1.5 Tesla MR scanner with intravenous application of 0.1 mmol gadopentetate dimeglumine/kg body weight. Three readers independently analyzed the images without knowledge of the clinical data; radiologists 1 and 2 were much more experienced in the interpretation of MR mammographies than radiologist 3. Initially, the observers visually categorized the lesions as benign or malignant following classification of BI-RADS (Breast Imaging Reporting and Data System). The readers also scored their own confidence level using a dichotome score (1: unsure vs. 2: sure) according to the BI-RADS classification. The images were then analyzed in a blinded manner with two technically different CAD systems: the full-time point (FTP) method (Cadsciences; White Plains, NY, USA) and the Dynacad version 1.1 (Invivo; Pewaukee, WI, USA). RESULTS: After CAD, all three readers classified more malignant lesions as BI-RADS 4 or 5 (suspicious or highly suggestive of malignancy). However, this increase in sensitivity revealed only statistical significance for observer 3 (p < 0.05). After CAD, the two experienced readers categorized about the same quantity of benign lesions correctly as BI-RADS 2. Observer 3 classified less benign lesions as BI-RADS 2 after both CAD analyses; i. e. the specificity decreased. The subjective confidence of all observers increased after analysis with Dynacad. Observer 3 also reported to be more confident after the FTP method. CONCLUSION: Computer-assisted detection can be a useful additional diagnostic tool for the radiologist in the interpretation of MR mammographies, but does not have the potential to replace the professional experience of a radiologist.
Authors: Matthias Dietzel; Pascal A T Baltzer; Tibor Vag; Aimee Herzog; Mieczyslaw Gajda; Oumar Camara; Werner A Kaiser Journal: Korean J Radiol Date: 2010-02-22 Impact factor: 3.500
Authors: Monique D Dorrius; Marijke C Jansen-van der Weide; Peter M A van Ooijen; Ruud M Pijnappel; Matthijs Oudkerk Journal: Eur Radiol Date: 2011-03-15 Impact factor: 5.315