Thyroid cancer, the most common type of endocrine cancer, continues to see rising detection rates. Surgeons often face difficulties during tumor removal in distinguishing between cancerous and healthy tissue in real time due to the delicate nature of nearby structures. Traditional diagnostic methods like fine-needle aspiration (FNA) and pathology, while accurate, are slow and do not provide immediate feedback during surgery. This can result in unnecessary surgeries for benign nodules or additional procedures if malignant tissue is missed.
A recent multi-institutional study published in Biophotonics Discovery introduced Dynamic Optical Contrast Imaging (DOCI) as a new approach for examining tissue. Unlike traditional methods that require dyes or contrast agents, DOCI uses illumination to measure natural autofluorescence from molecules within cells. This produces distinct optical signatures for healthy and cancerous tissues, which can be analyzed across 23 different optical channels to create detailed maps of tissue biology.
Researchers at Duke University and the University of California, Los Angeles combined DOCI with machine learning techniques. Tyler Vasse at Duke developed a two-stage AI analysis framework under the guidance of Tuan Vo-Dinh. Clinical deployment was led by Yazeed Alhiyari with Maie St. John's team at UCLA.
In the first stage of their research, an interpretable machine-learning model categorized specimens into healthy thyroid tissue, follicular thyroid cancer, or papillary thyroid cancer—the latter two being the most common differentiated forms of the disease. The system distilled data from 23 optical channels into a smaller set of features and achieved perfect accuracy on an independent test set. It also correctly identified samples from aggressive anaplastic thyroid cancers as malignant.
The second stage focused on pinpointing tumor locations using deep-learning models based on U-Net architecture. These models produced probability maps that accurately highlighted regions with cancerous cells, especially in cases of papillary thyroid cancer, while minimizing false positives in healthy tissue.
"By merging the speed of optical imaging with the power of AI, DOCI has the potential to reduce uncertainty in the operating room, prevent unnecessary surgeries, spare healthy tissue, and improve outcomes for patients with thyroid cancer," according to researchers involved in the study.
While current results were obtained from tissues immediately after removal rather than during live surgery, this technology suggests a future where surgeons could receive real-time guidance without needing labels or dyes during operations.
