Early-stage tumor detection using photoacoustic microscopy: a pattern recognition approach

Abstract

We report photoacoustic microscopy (PAM) of arteriovenous (AV) shunts in early stage tumors in vivo, and develop a pattern recognition framework for computerized tumor detection. Here, using a high-resolution photoacoustic microscope, we implement a new blood oxygenation (sO_2)-based disease marker induced by the AV shunt effect in tumor angiogenesis. We discovered a striking biological phenomenon: There can be two dramatically different sO_2 values in bloodstreams flowing side-by-side in a single vessel. By tracing abnormal sO_2 values in the blood vessels, we can identify a tumor region at an early stage. To further automate tumor detection based on our findings, we adopt widely used pattern recognition methods and develop an efficient computerized classification framework. The test result shows over 80% averaged detection accuracy with false positive contributing 18.52% of error test samples on a 50 PAM image dataset.

Files

Additional details