Using information metrics and molecular communication to detect cellular tissue deformation

Calcium-signaling-based molecular communication has been proposed as one form of communication for short range transmission between nanomachines. This form of communication is naturally found within cellular tissues, where Ca$²⁺ ions propagate and diffuse between cells. However, the naturally flexible structure of cells usually leads to the cells dynamically changing shape under strain. Since the interconnected cells form the tissue, a change in shape of one cell will change the shape of the neighboring cells and the tissue as a whole. This will in turn dramatically impair the communication channel between the nanomachines. We propose a process for nanomachines utilizing Ca$²⁺ based molecular communication to infer and detect the state of the tissue, which we term the Molecular Nanonetwork Inference Process. The process employs a threshold based classifier that identifies its threshold boundaries based on a training process. The inference/detection mechanism allows the destination nanomachine to determine: i) the type of tissue deformation; ii) the amount of tissue deformation; iii) the amount of Ca$²⁺ concentration emitted from the source nanomachine; and iv) its distance from the destination nanomachines. We evaluate the use of three information metrics: mutual information, mutual information with generalized entropy and information distance. Our analysis, which is conducted on two different topologies, finds that mutual information with generalized entropy provides the most accurate inferencing/detection process, enabling the classifier to obtain 80% of accuracy on average.