| Endoscopy is a medical procedure for identifying and visualizing anomalies in body cavities,which is widely used in minimal invasive surgery and clinical diagnosis.However,most clinical endoscopes can only provide structural information of tissues.Diagnostic results are obtained based on the experiences of clinicians to a large extent.During diagnosis and minimally invasive surgery,functional information is critical for comprehensive assessment of the physiology and pathology in tissues.Tissue blood supply,for example,is an important indicator.At present,there are multiple modalities of techniques for measuring tissue blood flow,including optical coherence tomography(OCT),photoacoustic imaging(PAI),ultrasound-doppler flowmetry and laser doppler imaging(LDI),and etc.These techniques,however,have the limitation in acquisition time.Laser speckle contrast imaging(LSCI)has been a promising technique for measuring tissue blood flow,providing full-field,high-resolution and two-dimensional blood flow map.Because of the similar imaging components with endoscopes,combining endoscopic imaging with LSCI has been a topic of high interest.The endoscopic laser speckle contrast imaging(ELSCI)would be very helpful for imaging the tissue blood flow in body cavities.Currently there are a few commercial LSCI devices.However,most of them are bulky and inflexible,limiting their combination with endoscopes.Therefore,a portable LSCI device is needed.In this thesis,we designed a multimodal endoscopic system with LSCI.A portable laser speckle imager was firstly developed for easier integration with endoscopes.We performed finger occlusion tests and photothrombotic stroke experiments for the performance evaluation.The results showed that the portable laser speckle imager had comparable performance with the conventional LSCI systems.Then the portable imager was combined with a commercial laparoscope to build up the multimodal imaging system.Finger occlusion tests and real-time cerebral blood flow(CBF)monitoring were performed for the performance evaluation,and the results showed that the multimodal endoscopic system could provide high-quality multimodal images.In addition,we proposed a method for laser speckle contrast image segmentation based on deep learning and unsupervised domain adaptation without using the target modality ground truth,which could be potentially applied in computer-aided diagnosis for the endoscopic system. |
