Breast Cancer is one of the highest cause for death in women, it's early identification increases the chances of survival and can help save many lives. We Trained a model to detect the level of calcification on a scale of 1 to 4 in a breast cancer image submitted by the user and classify it as benign or malignant. The model has an accuracy score of 90% on validation dataset.
##Images from Webapp
To use the webapp visit https://GPW5JFXKJ4OYCXAQ.anvil.app/TX2LVB3MJEOYZY3JUPJ3QEA5
Layer (type) Output Shape Param #
conv2d (Conv2D) (None, 224, 224, 32) 320
conv2d_1 (Conv2D) (None, 224, 224, 64) 18496
batch_normalization (BatchN (None, 224, 224, 64) 256
ormalization)
max_pooling2d (MaxPooling2D) (None, 223, 223, 64) 0
dropout (Dropout) (None, 223, 223, 64) 0
conv2d_2 (Conv2D) (None, 223, 223, 64) 36928
batch_normalization_1 (Batc (None, 223, 223, 64) 256
hNormalization)
max_pooling2d_1 (MaxPooling (None, 222, 222, 64) 0
2D)
dropout_1 (Dropout) (None, 222, 222, 64) 0
dropout_2 (Dropout) (None, 222, 222, 64) 0
flatten (Flatten) (None, 3154176) 0
dense (Dense) (None, 64) 201867328
batch_normalization_2 (Batc (None, 64) 256
hNormalization)
dropout_3 (Dropout) (None, 64) 0
dense_1 (Dense) (None, 128) 8320
batch_normalization_3 (Batc (None, 128) 512
hNormalization)
dropout_4 (Dropout) (None, 128) 0
dense_2 (Dense) (None, 8) 1032
Total params: 201,933,704 Trainable params: 201,933,064 Non-trainable params:
The web app accepts mammograms either coloured or gray scale standardizes it, gives a density rating of 1 to 4 and type of benign or malignant.
Prospective functionality may include a failed for the users to submit their images to our database which will then be used to update our model on new data ans possibly integrating our model with the use of transfer learning.