Article Impact Level: HIGH Data Quality: STRONG Summary of Nature Biomedical Engineering. https://doi.org/10.1038/s41551-025-01435-3 Dr. Xin Tong et al.
Points
- A new study evaluated panoramic photoacoustic computed tomography, a non-invasive imaging technique designed to overcome the limitations of conventional breast cancer screening methods, such as mammography and ultrasound.
- The novel PACT workflow was clinically tested on 39 patients, analyzing a total of 78 breasts to assess its effectiveness in a real-world diagnostic setting.
- A machine learning model analyzing the imaging data successfully distinguished between suspicious and normal tissue with a maximum area under the receiver operating characteristic curve of 0.89.
- This rapid, radiation-free system utilizes laser and ultrasonic sensors to provide high-resolution images of tumor-related vasculature by detecting hemoglobin signatures, eliminating the need for painful breast compression.
- Researchers identified PACT as a sensitive and promising diagnostic tool, with future work focused on expanding patient datasets to refine the machine learning classification models further.
Summary
A new study evaluated a non-invasive breast imaging workflow using panoramic photoacoustic computed tomography (PACT) combined with machine learning. The research aimed to address the limitations of conventional imaging, including the use of ionizing radiation in mammography and the operator dependency of ultrasound. The technique was tested in a clinical setting on 39 patients, encompassing 78 breasts. The PACT system provides high-resolution visualization of tumor-related vasculature, a key indicator of malignancy, by detecting hemoglobin signatures. This approach is effective irrespective of breast density and does not require contrast agents or painful breast compression.
The PACT system utilizes a near-infrared laser and an array of 512 ultrasonic sensors to produce detailed images in under 15 seconds. It can resolve structures as small as 0.25 mm at a tissue depth of 4 cm. Researchers developed learning-based classifiers to analyze the imaging data. A model designed to distinguish between normal and suspicious tissue achieved a maximum area under the receiver operating characteristic curve (AUC) of 0.89, a performance level comparable to conventional imaging standards. A subsequent classifier was developed to differentiate between malignant and benign lesions using 13 distinct PACT features, including those related to angiogenesis and tumor hypoxia.
These findings establish PACT as a sensitive and promising imaging tool for breast lesion evaluation. The integration of machine learning enhances the system’s diagnostic capability, allowing for the objective and accurate classification of tissue. By offering a rapid, radiation-free, and painless alternative, the PACT system presents a significant potential advancement in breast cancer diagnostics. Future work will focus on expanding the patient dataset to refine the classification models further and move the technology toward broader clinical applications.
Link to the article: https://www.nature.com/articles/s41551-025-01435-3
References Tong, X., Liu, C. Z., Luo, Y., Lin, L., Dzubnar, J., Invernizzi, M., Delos Santos, S., Zhang, Y., Cao, R., Hu, P., Zheng, J., Torres, J., Kasabyan, A., Lai, L. L., Yee, L. D., & Wang, L. V. (2025). Panoramic photoacoustic computed tomography with learning-based classification enhances breast lesion characterization. Nature Biomedical Engineering. https://doi.org/10.1038/s41551-025-01435-3
