Feasibility of non-imaging, random-sampling second harmonic generation measurements to distinguish colon cancer

In the United States, colorectal cancer is the third leading cause of cancer death. Colonoscopy with polyp removal may suffer from incomplete resection. Collagen is altered in dysplastic tissue and can be studied with second harmonic generation (SHG) imaging. SHG imaging endoscopes require miniaturized scanning components, which greatly adds to endoscope complexity.

We investigate whether non-imaging, randomly sampled SHG line or point intensity measurements are sufficient to distinguish normal tissue from tumor and tumor-adjacent tissue.

Unstained tumor, normal, and tumor-adjacent thin sections from 10 colorectal cancer subjects were imaged using a multiphoton microscope with constant power. SHG signal from collagen was isolated by grayscale thresholding, and the grayscale mean of the image was calculated. Supra-threshold pixels and lines of pixels in the image were randomly selected to simulate point sampling and line scanning.

The mean SHG signal from normal samples was significantly greater than adjacent samples ( $p<0.05$ ) and tumor samples ( $p<0.01$ ). For both sampling types, the $p$ -value becomes reliable after randomly sampling only 1000 times.

Reliable cancer detection information may be obtained through non-imaging SHG intensity measurements. A simple endoscope with this capability could help identify suspicious masses or optimum surgical margins.