Abstract 4555: Remote sensing image dtection; a new tool for evaluation the tumor thickness in tongue SCC

Abstract

Abstract A preoperative predictor of the risk of subclinical nodal metastasis would be useful. Studies have shown a strong correlation between histological tumor depth and the risk of nodal metastasis. To assess if preoperatively measured tumor depth predicts an increased risk of subclinical metastatic neck disease and thus the need for elective neck dissection. To further pin-point the boundary between normal to infective tissues, biopsy is generally performed. However, the similarity and the general nature of tissue substances between normal and cancer cells are sometimes hard to distinguish by human eyes. This is also true for experienced radiologist. Therefore, computer aided cancer cell classification is highly beneficial for such diagnoses. In this presentation, we implemented methods that are found highly successfully in satellite remote sensing image detection and classification. These methods can be grouped into two major categories, supervised and unsupervised classification. The supervised methods, Least Squares Orthogonal Subspace Projection (LSOSP), Non-negative Constraint Least Squares (NCLS) and Fully Constraint Least Squares (FCLS), require a detail library of every possible tissue signatures consist in the MR image. But, the classification results can be generally more accurate. On the other hand, the unsupervised methods, Automatic Target Detection (ATGP), Unsupervised Non-negative Constraint Least Squares (UNCLS), and Unsupervised Fully Constraint Least Squares (UFCLS), can automatically generate tissue classification results. However, parameters for these techniques such as the number of tissues need to be predefined. Also, the cancer cell classification still needs human intervention to justify the location of infectious region. Based on our experiments, both category of methods are show promising results. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 101st Annual Meeting of the American Association for Cancer Research; 2010 Apr 17-21; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2010;70(8 Suppl):Abstract nr 4555.