Transfer learning, or inductive transfer, refers to the transfer of knowledge from a source task to a target task. In the context of convolutional neural networks (CNNs), transfer learning can be implemented by transplanting the learned feature layers from one CNN (derived from the source task) to initialize another (for the target task). Previous research has shown that the choice of the source CNN impacts the performance of the target task. In the current literature, there is no principled way for selecting a source CNN for a given target task despite the increasing availability of pre-trained source CNNs. In this paper we investigate the possibility of automatically ranking source CNNs prior to utilizing them for a target task. In particular, we present an information theoretic framework to understand the source-target relationship and use this as a basis to derive an approach to automatically rank source CNNs in an efficient, zero-shot manner. The practical utility of the approach is thoroughly evaluated using the PlacesMIT dataset, MNIST dataset and a real-world MRI database. Experimental results demonstrate the efficacy of the proposed ranking method for transfer learning.

No abstract

Purpose MRI-based cell tracking has emerged as a useful tool for identifying the location of transplanted cells, and even their migration. Magnetically labeled cells appear as dark contrast in T2*- weighted MRI, with sensitivities of individual cells. One key hurdle to the widespread use of MRI-based cell tracking is the inability to determine the number of transplanted cells based on this contrast feature. In the case of single cell detection, manual enumeration of spots in 3D MRI in principle is possible; however, it is a tedious and time-consuming task that is prone to subjectivity and inaccuracy on a large scale. This research presents the first comprehensive study on how a computer based intelligent, automatic and accurate cell quantification approach can be designed for spot detection in MRI scans. Methods Magnetically labeled mesenchymal stem cells (MSCs) were transplanted into rats using an intracardiac injection, accomplishing single cell seeding in the brain. T2*- weighted MRI of these rat brains were performed where labeled MSCs appeared as spots. Using machine learning and computer vision paradigms, approaches were designed to systematically explore the possibility of automatic detection of these spots in MRI. Experiments were validated against known in vitro scenarios. Results Using the proposed deep convolutional neural network (CNN) architecture, an in vivo accuracy up to 97.3% and in vitro accuracy of up to 99.8% was achieved for automated spot detection in MRI data. Conclusion The proposed approach for automatic quantification of MRI-based cell tracking will facilitate the use of MRI in large scale cell therapy studies.

No abstract

No abstract