Jens Gregor

Dylan B. Lewis, Jens Gregor, Hector Santos-Villalobos

Modern vision pipelines increasingly rely on pretrained image encoders whose representations are reused across tasks and models, yet these representations are often overcomplete and model-specific. We propose a simple, training-free method to improve the efficiency of image representations via a post-hoc canonical correlation analysis (CCA) operator. By leveraging the shared structure between representations produced by two pre-trained image encoders, our method finds linear projections that serve as a principled form of representation selection and dimensionality reduction, retaining shared semantic content while discarding redundant dimensions. Unlike standard dimensionality reduction techniques such as PCA, which operate on a single embedding space, our approach leverages cross-model agreement to guide representation distillation and refinement. The technique allows representations to be reduced by more than 75% in dimensionality with improved downstream performance, or enhanced at fixed dimensionality via post-hoc representation transfer from larger or fine-tuned models. Empirical results on ImageNet-1k, CIFAR-100, MNIST, and additional benchmarks show consistent improvements over both baseline and PCA-projected representations, with accuracy gains of up to 12.6%.

William Whiteley, Vladimir Panin, Chuanyu Zhou et al.

Direct reconstruction of positron emission tomography (PET) data using deep neural networks is a growing field of research. Initial results are promising, but often the networks are complex, memory utilization inefficient, produce relatively small 2D image slices (e.g., 128x128), and low count rate reconstructions are of varying quality. This paper proposes FastPET, a novel direct reconstruction convolutional neural network that is architecturally simple, memory space efficient, works for non-trivial 3D image volumes and is capable of processing a wide spectrum of PET data including low-dose and multi-tracer applications. FastPET uniquely operates on a histo-image (i.e., image-space) representation of the raw data enabling it to reconstruct 3D image volumes 67x faster than Ordered subsets Expectation Maximization (OSEM). We detail the FastPET method trained on whole-body and low-dose whole-body data sets and explore qualitative and quantitative aspects of reconstructed images from clinical and phantom studies. Additionally, we explore the application of FastPET on a neurology data set containing multiple different tracers. The results show that not only are the reconstructions very fast, but the images are high quality and lower noise than iterative reconstructions.

William Whiteley, Wing K. Luk, Jens Gregor

Purpose: Neural network image reconstruction directly from measurement data is a relatively new field of research, that until now has been limited to producing small single-slice images (e.g., 1x128x128). This paper proposes a novel and more efficient network design for Positron Emission Tomography called DirectPET which is capable of reconstructing multi-slice image volumes (i.e., 16x400x400) from sinograms. Approach: Large-scale direct neural network reconstruction is accomplished by addressing the associated memory space challenge through the introduction of a specially designed Radon inversion layer. Using patient data, we compare the proposed method to the benchmark Ordered Subsets Expectation Maximization (OSEM) algorithm using signal-to-noise ratio, bias, mean absolute error and structural similarity measures. In addition, line profiles and full-width half-maximum measurements are provided for a sample of lesions. Results: DirectPET is shown capable of producing images that are quantitatively and qualitatively similar to the OSEM target images in a fraction of the time. We also report on an experiment where DirectPET is trained to map low count raw data to normal count target images demonstrating the method's ability to maintain image quality under a low dose scenario. Conclusion: The ability of DirectPET to quickly reconstruct high-quality, multi-slice image volumes suggests potential clinical viability of the method. However, design parameters and performance boundaries need to be fully established before adoption can be considered.