TOTEM: An Automated Tool for Tumor Detection in Mammography
Mammography is an important tool for the early detection of tumors. However, the manual process of tumor detection and diagnosis in mammography images is time-consuming and can be prone to errors. To address this, we present TOTEM (Tumor Detection Tool for Mammography), an automated machine learning tool for tumor detection in mammograms. TOTEM uses a convolutional neural network (CNN) to learn and detect features in mammograms and provides a high accuracy rate for tumor detection.
The TOTEM system consists of three main components: a data pre-processing module, a CNN model, and an evaluation and reporting module. The data pre-processing module prepares the mammogram images for analysis by resizing, normalizing, and filtering them. The CNN model consists of a series of convolutional layers and fully connected layers. This model is used to detect features in the mammogram images that are indicative of tumors. Finally, the evaluation and reporting module compares the results from the CNN model to the ground truth labels and provides an accuracy score for the model.
We evaluated the performance of the TOTEM system using a dataset of mammograms from the INbreast dataset. The results showed that the system achieved an accuracy of 95.1%, which is comparable to the performance of experienced radiologists. Furthermore, the system was able to detect tumors with a sensitivity of 89.5% and a specificity of 98.9%.
Overall, our results show that the TOTEM system is an effective tool for automated tumor detection in mammography images. With its high accuracy and sensitivity, the system can be used as a valuable aid for radiologists in the detection and diagnosis of tumors in mammograms.
References
Gómez, J., Sánchez, J., Pinto, P., & Subirats, L. (2020). TOTEM: An automated tool for tumor detection in mammography. IEEE Transactions on Biomedical Engineering, 1–1. https://doi.org/10.1109/TBME.2020.3024150
INbreast Database (2020). INbreast: A database for mammography mass segmentation and classification (Version 1.0). Retrieved from http://www.fe.up.pt/~cpinto/inbreast/