Directional Kernel Density Estimation for Classification of Breast Tissue Spectra

Abstract

Abstract-In Breast Conserving Therapy, surgeons measure the thickness of healthy tissue surrounding an excised tumor (surgical margin) via post-operative histological or visual assessment tests that, for lack of enough standardization and reliability, have recurrence rates in the order of 33%. Spectroscopic interrogation of these margins is possible during surgery, but algorithms are needed for parametric or dimension reduction processing. One methodology for tumor discrimination based on dimensionality reduction and nonparametric estimation -in particular, Directional Kernel Density Estimation -is proposed and tested on spectral image data from breast samples. Once a hyperspectral image of the tumor has been captured, a surgeon assists by establishing Regions of Interest where tissues are qualitatively differentiable. After proper normalization, Directional KDE is used to estimate the likelihood of every pixel in the image belonging to each specified tissue class. This information is enough to yield, in almost real time and with 98% accuracy, results that coincide with those provided by histological H&E validation performed after the surgery.

Optical biopsy of head and neck cancer using hyperspectral imaging and convolutional neural networks

“…Preliminary research demonstrates that HSI has potential for providing diagnostic information for a myriad of diseases, including anemia, hypoxia, cancer detection, skin lesion and ulcer identification, urinary stone analysis, enhanced endoscopy, and many potential others in development. [11][12][13][14][15][16][17][18][19][20][21][22] Supervised machine learning and artificial intelligence algorithms have demonstrated the ability to classify images after being allowed to learn features from training or example images. One such method, convolutional neural networks (CNNs), has demonstrated astounding performance at image classification tasks due to their capacity for robust handling of training sample variance and ability to extract features from large training data sizes.…”

Calibration and analysis of a multimodal micro-CT and structured light imaging system for the evaluation of excised breast tissue

“…This technique could potentially be useful in determining the degree of stromal alignment in breast specimens, which could be related to disease progression (Provenzano et al 2006, Conklin et al 2011). Furthermore, spatial and spectral analysis of raw reflectance images of excised breast tissue have offered great diagnostic potential in a dark-field confocal microscopy configuration, where scatter contrast was enhanced similarly to high-spatial frequency SLI (Laughney et al 2012, Pardo et al 2017). With future development of a classification model validated with histopathology, the technology described here can be evaluated in the clinical theater.…”

Automated skin lesion segmentation with kernel density estimation

“…Hypothesis H0 where the pixel belongs to the lesion region, and Hypothesis H1 where the pixel belongs to the non-lesion region. Maximum likelihood classification is used to decide if a pixel belongs to lesion or non-lesion region based on the value obtained for H0 and H1 [7].…”