Structural plasticity mediated by addition and elimination of dendritic spines is thought to underlie the formation of long-term memory. However, the spatial relationship of those structural activities during learning and memory remains unclear. Using in vivo two-photon microscopy, I track spine dynamics in mouse retrosplenial cortex (RSC) during contextual and spatial learning. I report that learning leads to addition of new spines that are spatially clustered, and the amount of clustering is predicted by spine turnover prior to learning. Both spine measures are correlated with learning and memory performance. Accordingly, a genetic manipulation by heterozygous mutation of Ccr5 that enhances pre-learning spine turnover also enhances learning-related spine clustering, as well as future learning and memory performance. New spines related to one memory do not form cluster with spines related to a distinct memory. In contrast, spines gained from one repetitive task tend to cluster with each other. Remarkably, clustered new spines are usually added on dendritic segments with rapid spine turnover, revealing the presence of hotspots on dendritic tree where elevated rates of spine turnover facilitate clustered spine addition associated with memory. One implication of these findings is that increased spine turnover may allow neurons to more efficiently sample the synaptic space during learning in order to optimize information acquisition. Once acquired, spine clustering may stabilize this information, thus strengthening memory circuits.

In surgical oncology, intraoperative tissue identification is vital for the ideal surgical outcome: resection of the entire tumor tissue and maximum preservation of the adjacent normal tissue. Meanwhile, the need for negative margins should be carefully balanced against excessive resection of healthy tissue which can result in significant patient morbidity. Currently, intraoperative tissue identification relies on a combination of visual cues, palpation, and surgeon experience as well as frozen section. In this dissertation, we demonstrate that Dynamic Optical Contrast Imaging (DOCI) can help surgeons accurately identify tissue intraoperatively. DOCI is a label-free and non-invasive optical modality that provides relative fluorescence lifetime images with extremely short acquisition and computational times. We report recent progress on system implementation towards intraoperative use, as well as systematically characterized performance of the system. We report the use of DOCI system in several head and neck surgical applications. Data from animal tissue shows that DOCI accurately differentiates tissue type and clearly delineate margins from mouse cancer models. Results from ex vivo experiments demonstrate that DOCI enables sensitive and specific mapping of hypercellular parathyroid gland. It is demonstrated that DOCI allows for real-time visualization and accurate delineation of OCSCC and adjacent healthy tissue, and that DOCI can be used in to clearly delineate margins between oropharyngeal cancer and surrounding healthy tissue.