“Kernel Locally Sensitive Hashing for the Content Based Image Retrieval”
Masoud Badiei Khuzani, PhD
ABSTRACT: Due to the development of the Internet at large scale and the availability of various image capturing devices such as digital cameras, smart mobile phones, image scanners, digital image databases are expanding very rapidly. With the popularity of the computer based smart system, content based image retrieval (CBIR) has grown in different areas to research. Efficient image browsing, image retrieval and searching tools are needed to users in various domains including histopathology image datasets. To achieve image retrieval, many retrieval systems have been developed. Two main frameworks are the content based retrieval and text based retrieval. Text based frameworks were introduced in 1970s. In this approach, text descriptors are used to annotate the image. However, annotating a large data-base is laborious. In CBIR, contents of image are used to annotate to perform retrieval in an unsupervised manner. In this talk, we propose a novel framework for the CBIR using kernel locally sensitive hash functions. In particular, we propose a novel framework for constructing a set of new hash functions based on radial kernels to speed up the image retrieval process. As a preliminary result, we validate our proposed retrieval system on the MNIST data-set.
“Prediction of Clinical Outcomes in Diffuse Large B-Cell Lymphoma (DLBCL) Utilizing Radiomic Features Derived from Pretreatment Positron Emission Tomography (PET) Scan”
Eduardo Somoza, MD
ABSTRACT: Diffuse Large B-Cell lymphoma (DLBCL) is the most common type of lymphoma, accounting for a third of cases worldwide. Despite established prognostic scores and advancements in treatment, the five-year percent survival rate for this patient population nears sixty percent, possibly related to the known heterogeneity of DLBCL. PET imaging features may characterize this diverseness and help predict clinical outcomes even before treatment initiation. The approach we have been employing to address this need is the creation of a prognostic model from pretreatment clinical and imaging data of DLBCL patients seen at Stanford Hospital and Clinics (SHC). In this presentation, we will provide an update on the radiomics component of our model. Preliminary results, efforts towards standardization, and future directions will be covered. Ultimately, we hope our efforts will lead to the development of a prognostic model that can be utilized to guide treatment selection in high risk DLBCL patients in an attempt to circumvent relapse or refractory disease.