Researchers at the University of California, Los Angeles have developed a compact, cost-effective diagnostic platform that combines lens-free holography and deep learning to automate HER2 scoring of breast cancer tissue samples. Their findings were officially published. BME Frontiers (BMEF)offers an innovative alternative to expensive and bulky traditional clinical digital pathology instruments.
Accurate HER2 assessment is the basis for breast cancer diagnosis, prognosis, and targeted treatment planning. Traditional whole-slide imaging scanners rely on advanced optical components and precision mechanical systems, making them unaffordable for many distributed clinics. The newly designed lens-free holographic device does not require an objective lens or mechanical focusing. It captures holographic diffraction signals from stained tissue sections over a 1,250 mm² field of view under RGB laser illumination, outperforming many commercially available pathology scanners with an effective imaging throughput of 84 mm² per minute.
To ensure reliable results, the team employed a five-model neural network ensemble strategy and Bayesian Monte Carlo dropout for real-time uncertainty quantification. When evaluated on a blinded dataset of 412 independent tissue samples, the system reached 84.9% accuracy for four-level HER2 classification and 94.8% accuracy for binary scoring. We successfully excluded 30.4% of the misclassified samples, and the loss of correct predictions was only 7.2%, effectively reducing the diagnostic risk. A simplified single blue light mode also maintains decent performance and further reduces hardware costs.
The entire imaging hardware costs less than $980, providing a great balance of affordability and functionality. Overall classification performance is comparable to high-end brightfield microscopes used in standard digital pathology workflows.
This all-new imaging AI framework provides a practical solution for high-throughput, on-site HER2 testing. This will expand access to standardized breast cancer pathology services and facilitate the uptake of low-cost computational pathology techniques, especially in regions where advanced medical equipment is lacking.
sauce:
Reference magazines:
Shen, C.-Y., others. (2026). Automated HER2 scoring with uncertainty quantification using Lensfree holography and deep learning. BME frontier. DOI: 10.34133/bmef.0278. https://spj.science.org/doi/10.34133/bmef.0278
