Feasibility of multi-spectral imaging system to provide enhanced demarcation for skin tumor resection

Invading tumors like basal cell carcinoma have usually no distinct demarcation for the human eye. Therefore, during resection, an additional rim around the tumor is removed. However, extending sprouts can be missed since most lesions are not uniform. To improve the visualization of the tumor demarcation, we developed a multi-spectral imaging system especially adapted for dermatological applications based on tunable liquid crystal spectral tunable filter technology and LED illumination. Enhanced visualization of skin tumor demarcation was achieved using three strategies. The first strategy is based on creating false color images by combining narrow band spectral filtered images by placing them into the red, green and blue image components of a color image at three specific wavelengths. These specific wavelengths were determined using a trial on error tool to achieve the highest contrast between malignant and healthy tissue. The second strategy is to make ratio images of narrow band spectral filtered images at specific wavelengths. A trail on error tool was created which enables the user to try multiple wavelengths to obtain optimal contrast. This method could be applied in real-time. For the third strategy, on pixel spectral segmentation is applied by selecting the pixel spectra in the center of a tumor, surrounding tissue and healthy tissue far away from the tumor. The correlation between these specific spectra and all image pixels is calculated using a fast algorithm. The degree is correlation is graded by color coding and presented in a false color images showing a detailed demarcation of suspicious regions in the tissue. Although this strategy is expected to provide a higher specificity, it takes more time to calculate than the first strategy.