Digital mammography images
Our team was one of the best performes groups in the Breast Cancer Contest organized by DREAM.
Each screening mammography exam usually comprises two images per breast: a view from above called a craniocaudal (CC) view and an oblique or angled view called mediolateral oblique (MLO) view. These two views show the medial part as well the external lateral portion of the breast as much as possible. More than 99% of all the exams provided by Group Health have both CC and MLO views for each breast imaged. Figure 2 shows an example of CC and MLO views for the left and right breast of an healthy subject.
2 Diferent views from real data
The data is shown as an example of left right Medio Lateral Oblique (MLO) and Craneo Caudal (CC), show in the orientation link.
The main problem derived with this data is that the data is taken from the wild, this means that no preprocesing or analisys is performed, just, a minor clasification and shor metadata. This issue affect to the following:
- The samples are unbalanced, a large number of healty is provide compared with organs with lessions
- Most of the probles arises with data augmentation
- There is not imput from the radilogist to show what are the hot-regions, (suspicius)
- Sage Bionetworks
- IBM Research
- Kaiser Permanente Washington Health Research Institute
- Icahn School of Medicine at Mount Sinai
- Seattle Cancer Care Alliance
- U.S. Food & Drug Administration
- Radish Medical Solutions
- National Cancer Institute
In the press..
- The IFIC, among the elite of the global research to improve the detection of breast cancer
- Published in the UPV
- New AI-based method to diagnose breast cancer
- Artificial intelligence to improve the precision of mammograms
- Artificial Intelligence combined with evaluations by radiologists improves the precision of mammograms