For esophageal cancer, as with most cancers, early diagnosis and treatment is the most effective way to handle the disease. For this type of cancer, physicians look for lesions that could be malignant by using an endoscope to examine the lining of the esophagus and take random biopsies of suspicious sites. However, this method has some drawbacks: Doctors sometimes miss lesions or biopsy the wrong area, the procedure takes time, and the biopsies are uncomfortable for the patient. Scientists from the University of California at Davis’ Center for Biophotonics and Medical Center in Sacramento and at Lawrence Livermore National Laboratory are working on an instrument that could one day be used in conjunction with conventional endoscopes. It would use multiple optical modalities to classify lesions according to their risk of malignancy, directing the need for additional sampling or other intervention, said Stavros G. Demos, a researcher at both the university and the national laboratory. This type of instrument could increase the efficiency and comprehensiveness of malignancy screening, he added. The instrument would use nine modalities based on near-infrared imaging techniques using either emitted or scattered light. “This approach offers key benefits as a guidance tool during exploratory or therapeutic procedures,” Demos said. Near-IR light is absorbed less by blood, which reduces interference when used in living tissue and which allows it to penetrate further into tissue than shorter wavelengths. This allows deeper tissue to be interrogated. He said that because near-IR images can be separated from the visible image, both can be acquired and displayed simultaneously. This allows the technology to mesh more easily with current techniques because doctors are trained to use a conventional color video image for endoscopy. But before starting on the instrument itself, the researchers had to determine whether combining the modalities would provide the sensitivity and specificity needed for this type of screening. To do so, they used three lasers for illumination: a HeNe laser at 632.8 nm and two diode lasers from Edmund Scientific, at 532 and 405 nm. They also used a broadband white-light source equipped with three bandpass filters centered at 700, 850 and 1000 nm, with bandwidths of 40 nm. The research was published in the March 20 issue of Optics Express. By combining nine optical techniques, researchers could more accurately predict whether an esophageal lesion was high- or low-risk than by using an individual technique alone. They used laser sources at 408, 532 and 633 nm to excite autofluorescence in images A, B and C, respectively. To collect parallel and cross-polarized scattering, they used light from a filtered broadband source to illuminate the sample at 700 nm (D and E, respectively), 850 nm (F and G) and 1000 nm (H and I). The researchers also created degree-of-polarization images for the respective wavelengths (J,K and L). Reprinted from Optics Express. To collect the images, the researchers used a CCD camera from Princeton Instruments. A longpass filter with a cut-on wavelength of 700 nm was placed in the acquisition path to filter out any short wavelength emission. They also used a removable polarizer in the acquisition path to choose parallel or orthogonal polarization. For each sample, they collected three tissue autofluorescence images, each with a different laser illumination. They used the white light and filter wheel to make tissue scattering measurements — three parallel polarization images and three orthogonal polarization images. They combined the two sets to create three-degrees-of-polarization images. According to Demos, using three lasers helped them find the optimal excitation wavelength for tissue classification with high sensitivity. The 700-, 850- and 1000-nm filters roughly cover the spectral region that is suitable for imaging biological tissues and enabled them to test for optimal contrast. The parallel and orthogonal polarization could help to increase sensitivity and specificity. The researchers tested the system on freshly acquired biopsies from human patients. The equipment was situated just outside the examination room so that the autofluorescence images could be acquired immediately, and the results were compared against histological readings by trained pathologists. The system acquired all nine images in less than two minutes; however, increasing the laser power might reduce the acquisition time to a few seconds. The investigators imaged 40 sample sets consisting of 18 normal biopsies and eight noncancerous lesions (or 26 low-risk samples) and 14 cancerous lesions deemed at high risk. At its best, the system correctly classified 12 of the 14 high-risk samples and 25 of the 26 low-risk samples. Analysis indicated that the combination of modes provided information not available from a single mode. They theorize that the three classification errors could have resulted from the relatively wide field of view, which imaged the entire microscope slide at once. In addition, each set of biopsies was reduced to a mean image intensity for each imaging modality. A cancerous lesion might not be large enough to affect the mean image intensity. Analyzing the images with higher spatial resolution might overcome this problem. Demos said that the team will work on developing a prototype that can be used in endoscopy to further test the technology and determine its accuracy for tissue classification and margin delineation in vivo.