This study investigates whether reconstructed spectral reflectance from macropathology multi-spectral images (macroMSI) can assist binary classification of tissue malignancy to identify excised tissue margin during skin cancer diagnosis. We captured high resolution 7-channel macroMSI of 10 samples before and after formalin fixing and a pathologist labeled 115 regions of interest. We reconstructed spectral reflectance by adaptive Wiener Estimation. Subsets of reconstructed spectra were input to k-Nearest Neighbors (kNN) and Support Vector Machine (SVM) classifiers and evaluated by average area under curve of stratified 5-fold cross validation. Results revealed that unfixed spectra were a superior feature set as classifier input. SVM outperformed kNN classifier.
各種検索
サポート
ヘルプ