Breast cancer patient survival analysis using the Cox proportional hazard model with the clinical and computed tomography datasets

Breast cancer is one of the most common fatal cancers worldwide, with 12.5% of the new cases diagnosed in 2020. The Computed Tomography, commonly referred to as CT scan, is used to determine the breast cancer treatment and prognosis of the cancer as it is challenging to detect. Breast cancer is common in women, but sometimes can be seen in men as well. Although there are models to predict Lung Cancer using the Convolutional Neural Network or ConvNet (CNN) from the CT scans and clinical data, no such implementation has been applied to the CT scans of the breast cancer using the images as well as clinical data. The overall relative 5 year survival rate for breast cancer is about 90%. This means 90 out of 100 women are alive 5 years after they’ve been diagnosed with breast cancer. Currently, there is a need to accurately predict the survival and malignancy score for this type of cancer, to diagnose at the very early stage. This work develops a model to predict the survival outcome of the patient with the breast cancer and gives a survival score. Our model uses the Concordance Index (c-index) as the evaluation metrics and it achieved a c-index score of 0.635 on the test data being 25% of the whole data on the German Breast Cancer Study Group 2 dataset.