TLDR: A Caltech-led research team has developed an innovative AI-enhanced photoacoustic computed tomography (PACT) technique for breast cancer detection. This method offers a safer, more comfortable, and potentially more precise alternative to traditional mammography by using machine learning to differentiate between malignant and benign tissue. Tested in a clinical study with 39 patients, PACT matched the diagnostic accuracy of mammography and MRI, without exposing patients to radiation or causing physical discomfort, and shows particular promise for screening dense breast tissue.
Pasadena, CA – A groundbreaking advancement in breast cancer detection has been unveiled by a Caltech-led research team, introducing an AI-enhanced photoacoustic computed tomography (PACT) technique that promises to revolutionize screening methods. This innovative approach offers a safer, more comfortable, and potentially more precise alternative to conventional mammography, addressing many of its long-standing limitations.
For decades, X-ray mammography has been the cornerstone of early breast cancer detection. However, it comes with significant drawbacks, including exposure to ionizing radiation, the discomfort of breast compression, and reduced efficacy in patients with dense breast tissue, often leading to false positives. Other existing methods like ultrasound, while safe, can be operator-dependent and lack specificity, while MRI is costly, time-intensive, and unsuitable for patients with certain implants or claustrophobia. “We were strongly motivated to work on this problem because none of the current techniques are perfect,” stated Lihong Wang, the Bren Professor of Medical Engineering and Electrical Engineering at Caltech, emphasizing, “The future of medicine has to be better than that.”
The PACT system, refined by Wang and his colleagues over two decades, combines near-infrared lasers, ultrasonic sensors, and sophisticated artificial intelligence. Instead of directly imaging tissue, PACT captures functional changes by detecting oxygenated hemoglobin, which is typically elevated in tumor areas. Short pulses of laser light are absorbed by molecules like hemoglobin, causing them to rapidly expand and contract, generating ultrasonic vibrations. These vibrations are then picked up by an array of 512 tiny ultrasonic sensors, allowing for the reconstruction of high-resolution images of the breast’s internal structures. This process provides a clear view of structures as small as a quarter of a millimeter at a depth of 4 centimeters.
A key differentiator of PACT is its patient-friendly experience. Unlike mammography, which requires painful compression, PACT scans are performed with the breast submerged in warm water. Each scan takes a mere 15 seconds during a breath-hold, significantly improving patient comfort and reducing anxiety. This non-invasive method eliminates radiation exposure entirely.
Machine learning plays a pivotal role in PACT’s diagnostic capabilities. Caltech researchers have trained neural networks on thousands of breast tumor images, enabling the algorithm to recognize subtle cancer-related microvascular patterns, such as uneven blood vessel growth and oxygen deprivation, which are often too faint for the human eye to detect. The system has been trained on images of malignant and benign growths, as well as suspicious and healthy tissue, enhancing its ability to notice minute variations and differentiate tissue types with high accuracy.
In a clinical study conducted in collaboration with researchers at the City of Hope Comprehensive Cancer Center in Duarte, California, PACT was tested on 39 patients, involving 78 breasts. The results, published in Nature Biomedical Engineering, demonstrated that PACT achieved diagnostic accuracy comparable to both mammography and MRI in distinguishing between normal and suspicious breast tissue, with a maximum area under the receiver operating characteristic curve of 0.89. The technology shows particular promise for screening dense breast tissue, where X-rays often struggle, contributing to up to 15% false positives.
“We basically use molecules to figure out the body’s physiology,” Dr. Wang explained. “That’s the beauty of photoacoustic tomography: By detecting molecules, we can figure out exactly how the body is functioning. When there’s a functional difference, that means we can potentially detect disease better.”
Also Read:
- Mayo Clinic Unveils AI System for Early Detection of Surgical Site Infections via Patient Photos
- T-SYNTH: Advancing Breast Imaging AI with Synthetic Data
Looking ahead, the research team aims to further enhance PACT’s capabilities by incorporating additional laser wavelengths and new features, such as imaging lipid content and collagen, to add new layers of specificity. They also plan to validate the system with larger datasets and continue improving the classification model, paving the way for PACT to play a vital role in improving early breast cancer detection and patient outcomes globally.
