Health Research and Policy

Abstract

DATE: May 30, 2013
TIME: 1:15 - 3:00 pm
LOCATION: Medical School Office Building, Rm x303
TITLE: Mass Spectrometry Imaging Data Treated by the Lasso Method for Cancer Margin Assessment in Gastric Cancer Surgery
SPEAKERS: Livia S. Eberlin, Post Doctoral Scholar, Zare Lab, Dept of Chemistry
Robert Tibshirani, Professor of Health Research and Policy (Biostatistics) and of Statistics

Desorption electrospray ionization (DESI) is an ambient mass spectrometry imaging technique that has been increasingly applied for molecular imaging in many fields of science. DESI-MS allows samples to be chemically imaged two-dimensionally in the ambient environment and without the need of extensive sample preparation, giving a detailed picture of the distribution of a multitude of molecules within the sample surface. In cancer research, DESI-MS has been shown to provide diagnostic information based on the metabolic profiles obtained from human tissue sections in which the molecular expression of lipids and other small metabolites are altered due to malignancy. Nevertheless, the large amount of molecular features obtained in a single mass spectrum makes data interpretation quite difficult.

In this work, we are applying the lasso method (multiple-logistic with L1 penalty) to treat the DESI-MS molecular imaging data obtained from a total of 46 tissue samples, including human normal and cancerous gastric tissue. In results so far, we are able to find the important spectral features that contribute to the discrimination between gastric cancer, normal stromal and normal epithelial gastric tissue. When cross-validation and an independent set of samples were used to test the method, high overall accuracy values of 99% and 97% were obtained, respectively. To further test our method, we are now applying the classifier to evaluate the presence of cancer in margins samples obtained from gastric cancer surgery, for which current standard histological methods lack the specificity and accuracy needed for clear diagnosis. Our results indicate that DESI/lasso may potentially be a valuable tool in assessing cancer margins in gastric cancer surgery.

Suggested Readings:
J. Friedman, T. Hastie, R. Tibshirani. Regularization Paths for Generalized Linear Models via Coordinate Descent. April, 2009.

Back to Workshop in Biostatistics

Stanford Medicine Resources:

Footer Links: