| Περίληψη: | <p>Optical microscopy has been an indispensable tool for many biomedical applications. Compared to rapid developments in research labs striving for higher spatio-temporal resolution and
information extraction effcacy, advances towards the clinic, especially those targeted at surgical applications, have been slow in comparison. Complex practicalities make the translation
of cutting-edge technology much more challenging. Consequently, there is a constant need to address the limitations of conventional medical imaging devices using optical techniques and to develop new instrumentation that is task-oriented and capable of providing stabilised real-time high-resolution image guidance for improved surgical outcome. In particular, imaging in the human brain entails the most stringent requirements, for which label-free techniques would be most beneficial. Apart from the microscope design, optical aberrations also affect the image quality. Until now, the concept of adaptive optics (AO) has not been introduced to surgical microscopy, as it most noticeably benefits systems with higher theoretical resolution. For a high-resolution surgical microscope, AO would have many roles to play, making its integration a potentially rewarding and meaningful contribution. </p>
<p>This thesis concerns to date the focused development of a compact and contactless label-free neurosurgical microscope and the investigation of using AO methods to address practical needs in surgical scenarios. Specifically, a long-working-distance reflectance confocal microscope was designed and built with special considerations for compactness and portable transit. It provides cellular-level resolution over wide imaging fields of view (FOVs) and visualises individual cells even in normal brain tissue exhibiting minimal label-free image contrast. The elimination of direct tissue contact and fluorescence labelling also greatly reduces surgical risk and complexity. Closed-loop sensor-based AO and remote focusing was later implemented and characterised following the development of an open-source Python software, SenAOReFoc, which is now publicly available on GitHub https://github.com/jiahecui/SenAOReFoc for the wider community. It was designed with a user-friendly graphic user interface and modular functional units for easy integration into existing AO microscopes. The use of remote focusing for fast axial scanning avoids slow mechanical movements of the translation stage and simultaneously corrects for system aberrations at each refocusing depth. Both volumetric and depth-wise random access imaging functions have been enabled. To fulfil requirements for real-time display in clinical
settings, some post-processing and image enhancement techniques are then demonstrated to remove image artefacts and improve image contrast with minimal computational effort. An extended range autofocusing technique that combines remote focusing with sequence-dependent learning using a bi-directional long short term memory network was developed to address possible respiratory and pulsing movements during surgery. A much larger system-aberration-free autofocusing range as compared to conventional methods was achieved in tissue both exposed to air and immersed in liquid. Network generality has been validated for different imaging FOVs
and for specimens not seen during the network training process. Continuous tracking of axial sample motion has also been demonstrated under varying experimental conditions. </p>
<p>In the remainder of this thesis, a generalised AO method was proposed for high-numerical aperture aberration-free refocusing in refractive-index-mismatched media involving both stage
translation and remote focusing. Two new sets of orthogonal modes were established using QR decomposition for either independent or balanced remote focusing and refractive index mismatch correction. Finally, first investigations of multi-conjugate AO in microscopy were carried out to correct for spatially-variant aberrations and increase the imaging FOV. For this purpose and to incorporate different imaging and AO modalities in the same system, a multi-purpose AO microscope was designed and developed beforehand.</p>
|
|---|