Cancer Detection Using Machine Learning...

Breast cancer is commonly found among women and hence its early detection is a major step in saving a life.Mammography makes use of low energy x-ray to examine the human breast for diagnosis and screening. According to the National Cancer Institute, breast cancer is the second most common cancer for women in the US, after skin cancer. Mammograms are not meant for curing breast cancer but they are used for detecting breast cancer as early as possible. The most important step here is the classification of the mammogram image as normal-abnormal. The texture of a breast in a mammogram image plays an important role in classification purposes. This article deals with the detection of breast cancer using a variation of Histogram of Gradient (HOG) and Gabor filter combination called Histogram of Oriented Texture (HOT) for feature extraction. The density of the mammogram image is important for classification.

Histogram Equalization

The intensity values can be adjusted using histogram equalization.By using Histogram equalization, the transformation of the intensity values matches the histogram of the output image with a specified histogram.

Canny edge detection

The Canny edge detector is an edge detector, it uses a multi-stage algorithm to detect the edges in images. It was developed by John F.Canny in 1986. The output of the Canny edge detection for normal images are given below.

Feature Extraction

This technique is used for reducing the dimensions of the image and it will be useful when image size is large, and reduced feature representation is required to quickly complete a task such as an image matching and retrieval. The features can be extracted by using the following techniques,

1.Histogram of Gradients.

2.Gabor filter.

1.Histogram of Gradients

The HOGfeature extraction method is applicable for denser images.It does so by dividing the image into small 8X8 pixels cells and blocks 4X4 cells.Each cell has a fixed number of gradients orientation bins. The gradient of an image is represented in horizontal and vertical directions

2.Gabor filter

Gabor filters are the acclimatization sensitive filters used for edge and texture analysis.It is a linear filter that analyzes specific frequency content in an image in specific directions and a localized region around point or region of analysis.

Classification

In this article, the K-NN classifier is used to classify the mammogram images into normal or abnormal. The KNN classifier has been used in statistical estimation and pattern recognition.

Capturing texture information is important for mammogram classification (both normal-abnormal and benign-malignant). HOG and Gabor filter captures this texture information better than the other descriptor and hence it performs better compared to others.

CONCLUSION

In the proposed work the images are collected internally and by using MIAS data collection. After that, the feature is extracted from a preprocessed image using feature extraction techniques such as HoG and Gabor filter. Using the extracted features, the KNN classifier classifies the input image into normal or abnormal. The K value of the classifier is changed and the corresponding accuracy is measured using certain parameters. This system attains 100 % of accuracy for k=1 and 2. Also, it achieves 95 % accuracy at k=4. The proposed system requires a minimum time to compile the steps.