Researchers have developed a method that combines uncertainty-aware artificial intelligence with lensfree holography to automate the assessment of HER2 status in breast cancer tissue samples. This approach aims to overcome two major obstacles to widespread AI use in pathology: the high cost of imaging equipment and the lack of reliable confidence measures in AI predictions.
The new technique, described in a report on Medical Xpress, uses a lensfree holographic microscope that can capture large tissue areas without expensive lenses. An AI algorithm then analyzes the images to determine HER2 expression levels while also estimating how certain it is about each prediction. This built-in uncertainty quantification could help pathologists trust automated results and reduce false diagnoses.
Key takeaways
- Lensfree holography eliminates the need for costly microscope lenses, making imaging more accessible.
- Uncertainty-aware AI provides confidence scores for each HER2 assessment, helping to flag uncertain cases for manual review.
- The combined system could enable automated HER2 testing in settings where traditional digital pathology infrastructure is not available.
- Accurate and reliable HER2 status is critical for determining whether breast cancer patients will benefit from targeted therapies like trastuzumab.
Understanding the challenge with AI in pathology
Artificial intelligence has shown promise for analyzing medical images, including tissue samples used to diagnose cancer. In breast cancer, determining the HER2 status of tumor cells is essential for treatment planning. HER2-positive cancers often respond to drugs that target this protein, while HER2-negative patients are spared unnecessary side effects.
Traditional HER2 testing relies on immunohistochemistry and sometimes fluorescence in situ hybridization, both of which require specialized equipment and trained pathologists. AI-based digital pathology could automate parts of this workflow, but two issues have limited its adoption. First, high-resolution digital scanners used to digitize slides are expensive, creating a financial barrier for many hospitals and laboratories. Second, most AI models provide a single answer without indicating how confident they are, which makes it risky to rely on them in clinical decisions.
Lensfree holography as a cost-effective imaging solution
Lensfree holography is an imaging technique that records holograms of tissue samples without using bulky objective lenses. Instead, it relies on the interference patterns created when light passes through the sample. The researchers used this method to capture wide-field images of breast cancer tissue sections, achieving sufficient resolution for HER2 analysis while keeping hardware costs low.
According to the report on Medical Xpress, the lensfree setup can digitize entire tissue sections in a single shot, reducing the time and complexity associated with traditional slide scanning. This makes the technology particularly attractive for low-resource settings or facilities that cannot afford standard digital pathology systems.
Uncertainty-aware AI for reliable predictions
Standard deep learning models output a fixed classification for each image, such as “HER2 positive” or “HER2 negative.” The uncertainty-aware AI used in this study goes a step further. For each prediction, the algorithm calculates a confidence score that reflects how sure it is about the result. When confidence is low, the system flags the case for human review rather than automatically outputting potentially unreliable results.
By incorporating uncertainty quantification, the method reduces the risk of false positives or false negatives that could lead to inappropriate treatment. The researchers tested the approach against conventional AI models and found that uncertainty-aware predictions were more reliable and better at identifying challenging cases where manual verification was needed.
Future implications for breast cancer diagnostics
The combination of cheap imaging hardware and trustworthy AI could accelerate the adoption of automated pathology in clinical practice. If further validated, the lensfree holography system might one day be used for on-site HER2 testing in smaller clinics, potentially speeding up diagnosis and treatment decisions.
The work also highlights the growing importance of uncertainty quantification in medical AI. As regulatory agencies consider how to approve AI-based diagnostic tools, built-in confidence metrics may become a requirement for safe deployment. The researchers aim to expand their approach to other cancer biomarkers and tissue types.
Frequently Asked Questions
What is HER2 and why is it important in breast cancer?
HER2 is a protein found on the surface of some breast cancer cells. When overexpressed, it indicates an aggressive form of cancer that can be treated with targeted therapies like trastuzumab (Herceptin). Accurate assessment of HER2 status is crucial for selecting the right treatment plan.
How does lensfree holography differ from traditional microscopy?
Lensfree holography records holograms of tissue using light interference, eliminating the need for complex optical lenses. This makes the imaging system simpler and cheaper, while still capturing high-resolution images of large tissue areas. It is especially useful for digital pathology applications where cost is a barrier.
How does uncertainty-aware AI improve diagnostic accuracy?
Uncertainty-aware AI models not only make a prediction but also estimate how confident they are in that prediction. This allows the system to flag uncertain results for pathologist review rather than providing potentially incorrect outputs. It reduces the risk of misdiagnosis and builds trust in automated systems.
===FEATURED_IMAGE_CAPTION===This is an original report by Vital Signs Today, informed by reporting from Medical Xpress. Read the original source.
This article is for information only and is not medical advice. See our Medical Disclaimer.
