Optimizing Acquisition and Reconstruction of Under-sampled MRI for Signal Detection

点击翻译按钮获取中文摘要

PROJECT SUMMARY Magnetic resonance imaging (MRI) is a versatile imaging modality that suffers from slow acquisition times, which is a challenge, for both time sensitive applications and for patient throughput. Accelerating MRI would benefit patients both by reducing the time they need to be in the scanner and in reducing the cost of healthcare. This project is part of a larger scientific effort to accelerate MRI while maintaining the diagnostic quality. Acceleration, even by a factor of two, would result in a major advance for public health. Two of the current approaches to accelerate MRI rely on collecting less data (under-sampling) and deep learning reconstruction. These approaches can lead to images with diagnostic quality using significant under-sampling but may suffer from artifacts which are hard to characterize and may sometimes resemble anatomy. Specifically, this project will optimize the performance of accelerated MRI, including undersampling patterns and deep learning reconstructions, on detecting and localizing subtle lesions. To carry out this optimization, we will first develop the methods required for detection of lesions by machine and human observer models. The human observer models will be validated by psychophysical studies where humans perform the detection task. In the first aim of this project, we will apply and develop detection tasks and model observers. We will consider under-sampled acquisition strategies in MRI including one and two-dimensional subsampling methods using deep learning reconstructions which enforce data consistency. We will develop detection tasks for signals in anatomical backgrounds were the signal location is known and when the observer needs to search for the signal. The human and machine performance in these tasks will be modeled. In the second aim, we will optimize data acquisition and neural network reconstruction using signal detection with observer models and psychophysical experiments. We will also introduce a detectability-based loss function to neural network reconstructions. There is recent interest in exploring the benefits of low/mid field MRI which has a trade-off with higher noise. In the third aim, we will evaluate the effect of field strength on signal detection. We will use data from high field acquisitions from a publicly available database to model images from lower magnetic fields. Using the detection of subtle lesions, we will evaluate detection performance with varying field strength. This research project will help to strengthen the research environment and broaden participation at Manhattan College by involving students in biomedical research incorporating applied mathematics, statistics and data science.