Scanning Probe Microscopy With Landmark-Referenced Control For Direct Biological Investigations

V.M. Ayres, F.M. Salam, N. Xi (Dept. of Electrical & Computer Engineering, Michigan State University), D. Wang (Dept. of Human Medicine, Michigan State University)

Scanning Probe Microscopy provides high resolution imaging of specimens, including biological specimens. Scanning Probe Microscope-based nanomanipulation is a newly emerging area that offers an orders-of-magnitude improvement over current manipulation capabilities. Together, the two offer the possibility of site-specific direct investigations of biological events. We present our research toward the development of a landmark recognition scheme for use within an adaptive nonlinear neural network controller, for high-end control of the X-Y motion of an SPM tip. To achieve efficient and reliable manipulation in a micro/nano environment, it is essential to possess capabilities of sensing, processing and actuation in dynamic interactions. In our approach, we acquire the signals which would normally be projected as an SPM image and extract essential information for subsequent use as the sensing component within a feedback control loop formulation which is used to accurately steer the probe's tip along a prescribed trajectory. The feedback loop formulation involves two components: (1) landmark recognition and (2) dynamic pattern matching through adaptive learning. Using this approach, we are developing a new scanning probe microscope capability in which the tip is able to return to and stay centered on a specific site by recognizing the way that the site feels to the probe tip, rather than the way an image of the site appears to the eye of a human operator.

Principal Component Analysis (PCA) is used for landmark recognition of specific biological features. Principal Component Analysis is a pattern recognition technique that selects/extracts key features from a data set. The feature selection process transforms the data space into the feature space by reducing the dimensionality of the data set. The reduced data set is comprised of the most effective features that contain the intrinsic information of the data. We will present our recent research using Principal Component Analysis to sort leukocytes (white blood cells) and erythrocytes (red blood cells) into distinct classes, and also to identify several different erythrocyte specimens as part of the same class. We find that that information from an initial 512x512 (xyz) SPM data set can be effectively represented by about eight essential features. We will further present our research on the introduction of the pattern classification information into the adaptive learning chip for dynamic tip control based on feature recognition.