• automated breast volume scanner
• breast cancer
• deep learning
• digital breast tomosynthesis
• machine learning
• personalized medicine

Breast cancer is the most common type of cancer among women, according to published data. Recent research demonstrates how machine learning and deep learning may use personal health data to forecast the risk of breast cancer in five years.
This PhD thesis will look into modern, technological ways for early detection of breast cancer in order to improve diagnostic accuracy. The study takes a multimodal approach, investigating breast cancer diagnosis from several perspectives. Key research themes include (i) how effective techniques from other oncology domains can be used in breast cancer early detection fields and (ii) strategies for gathering patient lifestyle data.
Regarding point (i), the research presented in this thesis resulted in the gathering of Digital Breast Tomosynthesis (DBT) and Automated Breast Volume Scanner (ABVS) images from over 80 Italian patients, which proved to be among the most promising medical images for distinguishing between benign and malignant lesions.
An analysis of the most relevant features was performed on these images with the goal of comparing the features considered most important by the model in identifying benign and malignant to the features of the images used by professionals to make a diagnosis.
Given the limited availability of images, in order to use more advanced approaches for benign/malignant categorization, public datasets had to be adapted to the sample that previously existed. An Attention UNET was thus constructed for the automated segmentation of DBT from the bounding box. This reduces image segmentation time because specialists simply need to identify the bounding box rather than perform the segmentation manually.
In contrast, for item (ii) on gathering information on Italian women’s lifestyles, a web application was created. Over the course of about a month, this app asks Italian women of all ages questions about their food habits, health condition, physical activity, and well-being. Information about Italian women’s lifestyles is gathered in order to analyze what the literature deems to be breast cancer risk factors. The results of the first responses demonstrate how such a device might help Italian women improve their knowledge and daily habits.