Image processing methods to study the living hearing organ

Sammanfattning: In this thesis we are developing image processing methods allowing one to probe the structure and function of the mammalian hearing organ, an intricate cellular assembly with a highly specialized three-dimensional organization. Confocal as well as electron microscopy was used for image acquisition. Three-dimensional confocal imaging is hampered by noise and blur, and this is even more significant when working with deep biological imaging, such as our temporal bone preparation of the guinea pig cochlea. Blur is not only caused by the optical properties of the microscope, but by the scattering of light and distortions caused by the sample itself. Deconvolution is widely used in image processing to reverse these effects. If the system s point-spread function (PSF) is known, a deconvolution can be performed leading to increased image resolution. The idea of getting all the information about the PSF and deconvolution from the image itself is explored in this thesis. A set of tools were developed that screens the images for structures that can be used as PSF estimates. Using the extracted structures, a PSF model for deconvolution was designed, resulting in a significant improvement in deconvolutions of images acquired from the inner ear. To study functionally relevant movements inside the hearing organ, a wavelet-based threedimensional brightness-adjusted optical flow algorithm was designed. Numerical experiments show that the algorithm allows motion detection with less than 10% magnitude error over the range from 0.5 5 pixels under conditions of noise more severe than that typically found in real experiments. The angular error was less than 10° for the same range of motions. Accurate quantification of motion and deformation patterns in the hearing organ is therefore possible. The algorithm was applied to three-dimensional image sequences showing the sound-evoked motion of cochlearsensory cells. Motion directed perpendicular and parallel to the surface of the hearing organ were robust and displayed a linear relationship to each other, suggesting that they are governed by passive cellular properties such as stiffness, mass and damping. In contrast, components directed away from the center of the spiral-shaped cochlea showed much higher variability, and thesecomponents may be more tightly linked to the active, force-generating property of the outer hair cells. Outer hair cells are unique because of their capacity for high-frequency force generation, which requires specialized connections between molecular motors and the cell skeleton. Using electron tomography and three-dimensional reconstruction, we examined one such connecting structure, the pillars , which links the plasma membrane to the cell skeleton. Individual pillars show structural variability, which may have functional relevance. However, tissue processing and imperfect imaging conditions may cause a part of the variability. Information gained from these studies will be useful to develop a more detailed understanding of the three dimensional micromechanics of the cochlea, and to test existing models describing the mechanical vibrations of this organ.