By Dave DeFusco
Katz School researchers have designed a mobile app using AI technology that analyzes mammogram images and provides real-time diagnostic predictions, all from the convenience of a smartphone. They will present their project findings, outlined in the white paper, “Pink Guardian: A Gateway to Early Breast Cancer Detection,” at the 16th International Conference on Human System Interaction in Paris in July.
At the core of the app, called Pink Guardian, is the BC-InceptionV3 architecture, a sophisticated model developed within TensorFlow Lite, an open-source library from Google that allows developers to run machine learning models on mobile devices. This powerful model excels at identifying potential signs of breast cancer from mammogram scans, offering users diagnostic predictions along with confidence scores. These scores help users understand the certainty of the app’s analysis.
“Pink Guardian represents an important advancement in mobile health technology, offering a reliable and accessible tool for early breast cancer detection,” said Kanchan Subhashchandra Maurya, lead author of the paper and a student in the M.S. in Artificial Intelligence. “By combining cutting-edge AI with a user-friendly interface, the app empowers users to take proactive steps in monitoring their breast health. This innovation promises to significantly impact healthcare practices and patient outcomes worldwide.”
Development of the app involved gathering and refining a diverse dataset of mammogram images, known as the CBIS-DDSM, sourced from Kaggle, a web-based platform and online community for data scientists and machine learning practitioners. This dataset includes over 10,000 mammogram images, providing a broad range of cases for training and testing. The app utilizes advanced machine learning techniques, such as deep learning and convolutional neural networks, to accurately classify images as either benign or malignant.
The app’s key features are a user-friendly interface, allowing users to upload mammogram images from their device’s gallery or directly from the camera. The app then processes the image and provides a diagnosis along with a confidence score, such as “malignant (75%)”; enables quick and accurate predictions, making it a valuable tool for preliminary breast cancer screening; and optimizes for mobile devices, ensuring smooth operation across various smartphones.
The app cautions that while Pink Guardian offers preliminary assessments, users should consult healthcare professionals for further evaluation and confirmation. The researchers also aim to implement security measures to ensure the privacy and protection of user data.
“The app marks a significant step forward in the integration of AI and healthcare,” said Youshan Zhang, Maurya’s mentor, co-author of the paper and assistant professor of artificial intelligence and computer science. “It not only enhances the accuracy of breast cancer detection but also democratizes access to advanced diagnostic tools. We’re committed to continuous improvement, with plans to expand its capabilities to other medical imaging modalities, such as ultrasound and CT scans.”
