Background & Objective: Breast cancer is the most widespread invasive cancer in women and the second leading cause of mortality. A good number of work had done to harness the potential available in the electronic healthcare data but many unstructured healthcare data remained unexplored because of its complex nature. Mammography reports are the dictations of radiologist and prime source of information about the lesion. Analyzing and using this information-rich source can significantly improve the quality of living and saves the life. Method: In this paper, we developed an automated system that can automatically extract the information from the mammography reports and effectively used to classify them based on BI-RADS standards. The performance of the proposed method is evaluated using classification accuracy, precision, recall, f1-score and confusion matrix. Results: The results show the highest classification accuracy (97.58%) is achieved for this study. Results suggest that it is a feasible system for quicker and more precise diagnosis of breast cancer. Conclusion: Unstructured free text mammography dictations are the rich source of domain-specific information about breast lesion. Using this information to learn about interesting patterns and helps in identifying lesion malignancy accurately. The proposed model reads the free text mammography observations and detects the malignancy with high accuracy and recall.
Volume 11 | 06-Special Issue
Pages: 1319-1331