Use of A'Scape for the Detection of Lung Tumors

摘要

The Automatic Statistical characterization and Partitioning of Environment (A'SCAPE) has previously been successfully used to characterize infrared (IR) and radar land scenes as a target predetection state (DoD application) and, in another application (law enforcement), to detect weapons concealed underneath clothing in scenes collected by different types of sensors including IR and Millimeter-wave sensors. In this dual-use technology project, the prime emphasis was placed on structuring A'SCAPE to (1) detect tumors in lung tissue, and (2) classify a particular tumor as being either benign or malignant wing computer tomography (CT) data. It is shown that A'SCAPB (1) can successfully highlight suspected tumor tissue within the lung image, and (2) has the potential to detect and classify tumors. The latter point can be proven however only if more data of both benign and malignant tumors is made available. During this effort, a library of only 5 usable data sets were available. Two of these sets are wed to develop rules and the remaining three are used to test the rules. However, the results demonstrate promise for future application of this approach.