AI-Powered Label-Free Optical Imaging Accurately Identifies Thyroid Cancer During Surgery

Thyroid cancer is the most common endocrine cancer, and its rising detection rates have increased the number of patients undergoing surgery. During tumor removal, surgeons often face uncertainty in distinguishing cancerous tissue from healthy structures, increasing the risk of incomplete excision or unnecessary surgery. Current diagnostic tools, such as fine-needle aspiration and post-operative pathology, are accurate but slow and provide no real-time guidance in the operating room. New research now demonstrates a label-free optical imaging approach combined with artificial intelligence (AI) that can rapidly identify and localize thyroid cancer tissue.

Researchers at Duke University (Durham, NC, USA) and the University of California, Los Angeles (UCLA, Los Angeles, CA, USA) employed Dynamic Optical Contrast Imaging, or DOCI, a technique that illuminates tissue and measures its natural autofluorescence rather than relying on dyes or contrast agents. Each DOCI scan captures data from 23 optical channels across a wide field of view, generating a detailed spectral fingerprint that reflects underlying tissue biology.

Freshly excised thyroid specimens were imaged using DOCI and analyzed with a two-stage machine learning framework. In the first stage, an interpretable classification model distilled complex optical data into a small set of features to categorize tissue as healthy, follicular thyroid cancer, or papillary thyroid cancer. In the second stage, deep-learning models based on a U-Net architecture were applied to generate spatial tumor probability maps, identifying the precise location of cancerous regions within each specimen.

The findings, published in Biophotonics Discovery, show that the AI system accurately classified thyroid tissue types and achieved perfect accuracy on an independent test set. Importantly, it also correctly identified samples from aggressive anaplastic thyroid cancer as malignant, despite not being explicitly trained on that subtype. The deep-learning segmentation models produced highly accurate tumor maps, particularly for papillary thyroid cancer, while maintaining very low false-positive rates in cancer-free tissue.

Although the current study analyzed tissue after surgical removal, the results point toward future intraoperative use. By providing rapid, label-free visualization of cancer margins, DOCI combined with AI could help surgeons remove tumors more precisely, reduce repeat surgeries, and spare healthy tissue. With further development, the approach may offer real-time guidance in the operating room, improving outcomes for patients with thyroid cancer.

Related Links:
Duke University
UCLA