School of Mechanical Engineering Seminar
Monday, May 9, 2016 at 15:00
Wolfson Building of Mechanical Engineering, Room 206

 Rotating microfluidics for Mixing and Process Intensification

Wallace WF Leung

Department of Mechanical Engineering, The Hong Kong Polytechnic University,

Hung Hom, Kowloon, Hong Kong

 A novel method using transient inertial effect is used to mix a small milli- to micro-liter sample of different species in a rotating chamber. The primary flow generated from inertia induces a complex three-dimensional Coriolis flow that speeds up mixing by orders of magnitude in an otherwise viscous-force dominant micro-chamber. Experiments and numerical simulations are used to investigate the mixing due to the complex secondary flows generated. Such mixing and process intensification can be applied to cell culture, chemical reaction studies, and various other applications.

In a continuous-flow microchannel, a steady angular rotation generates throughflow down the channel by centrifugal acceleration, and at the same time the Coriolis acceleration generates crossflow that mixes fluids across the channel. The mixing can be carried out in either subcritical or supercritical mode depending on the channel width-to-height ratio. Experimental and numerical simulations have been used to confirm the results. A scale-up law on mixing is also presented based on dimensionless groups determined from Buckingham-π theorem.

.

Assessment of breast density and bilateral asymmetry for risk stratification and early detection of breast cancer

Dr. Dror Lederman

Breast cancer is the second most common type of non-skin cancer and the fifth most common cause of cancer death. It has long been shown that early detection of breast cancer may increase the treatment options, survivability, and chance for full recovery. Therefore, methods for earlier detection of breast cancer and risk stratification have long been of great interest. Mammography has been widely used for this purpose, both for screening and diagnosis. However, mammography interpretation suffers from relatively low detection sensitivity and specificity, especially in younger women (i.e., less than 50 years old), due to the low prevalence of breast cancer and denser breast tissue. In order to cope with this problem, we have been developing a novel methodology for mammography interpretation, which assesses the level of mammographic asymmetry between bilateral breasts as a major risk and diagnostic factor. The bilateral mammographic asymmetry information is combined with unilateral mammographic tissue information, to yield a fused risk-probabilistic model. In this talk, I will introduce our ongoing research work in this field, and in particular the proposed framework and set of algorithms that we have been developing for this purpose. I will also present our work on tissue density estimation. Some experimental results will be presented alongside with a discussion on the challenges and trends in this field of research.

Dror Lederman received the B.Sc., M.Sc. and Ph.D. degrees in Electrical and Computers Engineering in 1998, 2003 and 2009, respectively, and the B.EMS. (Bachelor in Emergency Medicine) degree in 2005, all from Ben-Gurion University of the Negev, Beer Sheba, Israel. Between 2009 and 2011 he was a research fellow at the Imaging Research Division, Department of Radiology, University of Pittsburgh, Pittsburgh, PA, USA, where he was primarily involved in development of computer-aided diagnosis for breast cancer and lung computed tomography. Dr. Lederman is now a research scientist at Intel Corporation and a senior lecturer at the Holon Institute of Technology.

ההרצאה תתקיים ביום ראשון 01.05.16, בשעה 14:00

 בחדר 315, הבניין הרב תחומי, אוניברסיטת תל אביב

Abstract :

Matrix factorization is a popular method for modeling user and item behaviors in recommender systems. Matrix factorization is used to capture the item and user biases when building a recommendation model, thus representing the User-Item-Rating space in a compact form. One of the most known methods used for matrix factorization is SVD (Singular Value Decomposition). With SVD, the usage of matrix factorization results in a representation of the original user-item-rating matrix as a combination of latent (implicit) dimensions, which are perceived as the user interests and the item attributes. Understanding these dimensions and their translation to true user preferences is a difficult task. Our research focuses on non-latent (explicit) factorization of the user-item-rating matrix. We show that, given existing item attributes, we may build an explicit model describing the user preferences. Using this model we demonstrate how to predict ratings using explicit attributes from the data. To select the best attribute for our model, we utilize the realization based entropy approach and define two new measures: (1) the space entropy and (2) the consent entropy. Using a combination of these entropies, we may quantify the contribution of each non-latent attribute to our prediction model. Using the Movielens datasets, we compare our results to the known "SVD" model, and show how our explicit model yields similar results.