next up previous contents
Next: CC Views with MLO Up: Málaga Database Previous: MLO Views   Contents

CC Views

We evaluated our algorithm using also the set of $ 35$ CC mammograms obtained from Málaga database. In this experiment, we used as a training set a database of RoIs extracted from the CC views. Thus, the same leave-one-out methodology used before is applied in order to not bias the results.

The mean $ A_z$ obtained for all mammograms was $ 90.5\pm 9.5$ , $ 89.1\pm 5.9$ , and $ 91.9\pm 4.7$ for algoritms d1, d2, and Eig, respectively. Moreover, Table [*] shows the performance of the algorithms depending on the mass size. Note that this time the proposed algorithm obtains similar performance independently on this factor, while the other algorithms vary its performance depending on the mass size.

Using this database we can also compare the performance of our proposal when dealing with MLO views and CC views. Note that the performance for CC images seems more size independent than in MLO images. Further, in this case the mean is slightly increased. However, we consider that there are not enough images to obtain reliable conclusions.


Table 4.3: Influence of the lesion size (in $ cm^2$ ) for algorithms d1, d2, and the proposed approach, using the CC images of Málaga database. The results show mean and standard deviation of the $ A_z$ values.
  Lesion Size (in $ cm^2$ )
 
  $ <$ 0.70 0.70-1.20 1.20-2.00 $ >$ 2.00
 
 -||-- d1 $ 94.2\pm 9.3$ $ 93.1\pm 7.8$ $ 87.7\pm 6.8$ $ 87.8\pm 10.5$
 -||-- d2 $ 87.4\pm 4.3$ $ 89.7\pm 6.0$ $ 88.2\pm 6.7$
 -||-- Eig $ 91.7\pm 5.3$ $ 93.2\pm 5.8$ $ 92.9\pm 4.0$
 -||--



next up previous contents
Next: CC Views with MLO Up: Málaga Database Previous: MLO Views   Contents
Arnau Oliver 2008-06-17