Breast cancer is the most common type of diseases among women, affecting 2.1 million women every year. According to the World Health Organization (WHO), 627,000 deaths have been registered in women due to breast cancer. This is 15% of deaths in women caused by cancer. Early detection and accurate diagnosis are the best methods to treat breast cancer; in such cases, the disease can completely be cured. Hence, it is essential to classify the presence of breast cancer accurately.
What it does
The work is trying to develop a powerful end-to-end classification model for screening breast cancer.
How I built it
The dataset is from CBIS-DDSM (Curated Breast Imaging Subset of Digital Database for Screening Mammography) database. I have tested that data before with TF 1.7, which has a steep learning curve and is not user-friendly. Just several hours ago, I saw this hackathon and decided to test it with Tensorflow 2.0.
Challenges I ran into
Got several errors in 2.0 with the code I used in Tensorflow 1.7. It was a little bit painful when there were only a few hours left, and I needed to figure out the problem quickly by checking out documentation related to those changes.
Accomplishments that I'm proud of
I 've done. Even though the model could be better, but now I am satisfied with the result as well as Tensorflow 2.0. The knowledge could be useful for further development.
What I learned
To finish a project in last minutes does not always bring negative experience. I forced myself to go through Tensorflow2.0 documents quickly, and I may not be so productive if I was not under such a pressure :) But Tensorflow2.0 is also well documented. That helps a lot.
What's next for Breast cancer prediction using Tensorflow2.0
Well, refine the model and improve its performance.