Optic Fiber Automatic Identification And Analysis Of Key Technologies

Presently, in the neurological research related to human and animal, histological examination of nerve fibers is often complemented by morphometric analysis in both clinical and research settings. Nerve morphometric analysis could provide researcher some valuable parameters.The accurate isolation of the ROI (Region of Interesting) is the key to the nerve morphometric studies. In the early investigation of nerve morphometry, the ROI is achieved manually i. e. the experienced neuroscientists draw the outline of the ROI in the microscopy image. Gradually, the disadvantage of manual morphometry is noticeable. Manual nerve morphometry is extremely tedious, labour intensive and time-consuming, and segmentation of ROI depend on the subjective decision and experience of operator. To overcome this there have been efforts to devise representative sampling schemes to reduce the amount of data analysis required. However, the accuracy of results is dubious.As a result, automatic identification of nerve fibers has draws much attention from image processing and neuroscience communities. Since the eighties of the last century, algorithms and commercial softwares related to nerve fibers detection were reported and promoted to improve the efficiency of fiber detection.Nevertheless, precise segmentation is the basis of the fiber analysis. In the process of fiber detection, the failure always results from the diversity of nerve fiber, for instance, lack of prominent features and irregular in shape. In the nerve with dense population, for example the optic nerve, the detection becomes more difficult.At present investigation, an original scheme for automatic segmentation of optic nerve fiber is proposed. Most importantly, the anatomical structure of fiber is taken into account. First, the center line of myelin sheath is picked up; therefore, the location of fiber is determined. Next, the contour is moved in, moved out, and finally, the fine contour is achieved by evolution of contour.Motivated by the above original scheme, we introduced watersnake to segment nerve fibers. Initially, watershed extracts continuous and topological reserving sketch of the nerve fibers and the false fibers are get rid of. In the following step, the initial contours are steered to be good fit to the fine location of axon and myelin sheath. The final contours are smooth and accurate.The proposed method has been tested using nerve image by random sampling; the results from automation were compared with those of manual method with the help of PHOTOSHOP software. The overall detection rate and false alarm rate were found to be 95. 7% and 3. 3%, respectively. The axon diameter frequency distribution was comparable to that done manually, and the peak has no offset.The current results indicate that the detection rate is slightly superior to the research of Romero (94.8%) , but the false alarm rate is inferior to that (1.0%) . It was found that the false fibers primarily originate from the close region among the crowded nerve fibers. Probably the sciatic nerve fibers adopted in Romero's research are sparse, which result in the false alarm rate is lower than ours.In our investigation, we adopted electron microscopy photographs with high resolution. In the following research, we would continue our investigation and made the technique of watersnake to be generally applied, so that it could be adaptable to the images with low resolution (light microscopy).