When a patient is diagnosed with cancer, fluorescence-guided surgery is often their best treatment option. This highly successful procedure uses an inert near-infrared dye to discern tumors from healthy cells. However, every potential tumor must be checked by an oncologist. This step adds upwards of 30 minutes and costs thousands of dollars per potential tumor. 

The authors developed OptiDX, a smart camera powered by AI to evaluate tumor fluorescence in real time. This tool provides immediate and accurate tumor identification without the frozen section staining process normally required. 

Fluorescence-guided methods are powerful tools to identify tumors as small as 2mm in length. The diversity of shapes, sizes, and intensities among differing health populations makes direct fluorescence evaluation difficult. Thus, the authors employed deep learning and computer vision toolkits to predict cancer probability from fluorescence, demographic, and co-morbidity data. 

• Identifies tumors immediately, saving over 30 minutes and thousands of dollars per tumor per operation.
• Requires energies under 0.1 eV, significantly lower than the operating room lights.
• Reliable and consistent algorithm eliminates observer bias.
• Area under receiver-operating curve is 0.897, close to a perfect score of 1.0, indicating few false positives and false negatives.
• Intuitive and easy to use