Biomedical Engineering Reference
In-Depth Information
Figure 2.
Flowchart of a typical automatic tongue diagnosis system.
See attached CD for
color version.
tongue diagnosis is dependent on the subjective experience and knowledge of the
doctor, which has inevitably impeded the development of TCM.
To circumvent this problem, some research groups [68-75], especially in
China, have designed computerized tongue diagnosis systems based on image
analysis techniques that attempt to alleviate the diagnostic discrepancy among
doctors and provided systematic and quantitative diagnosis. Figure 2 presents a
flowchart of a typical automatic tongue diagnosis system based on image analysis.
First, the tongue of a patient or a subject to be examined is captured using a color
CCD (Charge Coupled Device) camera under standard imaging geometry condi-
tions. Second, some image analysis techniques are applied to the tongue image:
including color calibration, tongue body segmentation and extraction, and color
recognition. Third, quantitative representation of the tongue follows to charac-
terize the pathological features defined in TCM, such as the colors of the tongue
substance and coating, the thickness and wetness of coating, and the shape of
tongue. Finally, the quantitative features obtained are mapped to diseases via
data-mining techniques or learning theories, like the Bayesian network adopted in
[72]. Here our study is focused on tongue body extraction and segmentation.
