Deep neural network models can learn clinically relevant features from millions of histopathology images. However generating high-quality annotations to train such models for each hospital, each cancer type, and each diagnostic task is prohibitively laborious. On the other hand, terabytes of training data -- while lacking reliable annotations -- are readily available in the public domain in some cases. In this work, we explore how these large datasets can be consciously utilized to pre-train deep networks to encode informative representations. We then fine-tune our pre-trained models on a fraction of annotated training data to perform specific downstream tasks. We show that our approach can reach the state-of-the-art (SOTA) for patch-level classification with only 1-10% randomly selected annotations compared to other SOTA approaches. Moreover, we propose an uncertainty-aware loss function, to quantify the model confidence during inference. Quantified uncertainty helps experts select the best instances to label for further training. Our uncertainty-aware labeling reaches the SOTA with significantly fewer annotations compared to random labeling. Last, we demonstrate how our pre-trained encoders can surpass current SOTA for whole-slide image classification with weak supervision. Our work lays the foundation for data and task-agnostic pre-trained deep networks with quantified uncertainty.
Compressive sensing (CS) reconstructs images from sub-Nyquist measurements by solving a sparsity-regularized inverse problem. Traditional CS solvers use iterative optimizers with hand crafted sparsifiers, while early data-driven methods directly learn an inverse mapping from the low-dimensional measurement space to the original image space. The latter outperforms the former, but is restrictive to a pre-defined measurement domain. More recent, deep unrolling methods combine traditional proximal gradient methods and data-driven approaches to iteratively refine an image approximation. To achieve higher accuracy, it has also been suggested to learn both the sampling matrix, and the choice of measurement vectors adaptively. Contrary to the current trend, in this work we hypothesize that a general inverse mapping from a random set of compressed measurements to the image domain exists for a given measurement basis, and can be learned. Such a model is single-shot, non-restrictive and does not parametrize the sampling process. To this end, we propose MOSAIC, a novel compressive sensing framework to reconstruct images given any random selection of measurements, sampled using a fixed basis. Motivated by the uneven distribution of information across measurements, MOSAIC incorporates an embedding technique to efficiently apply attention mechanisms on an encoded sequence of measurements, while dispensing the need to use unrolled deep networks. A range of experiments validate our proposed architecture as a promising alternative for existing CS reconstruction methods, by achieving the state-of-the-art for metrics of reconstruction accuracy on standard datasets.
Sleep plays a crucial role in the well-being of human lives. Traditional sleep studies using Polysomnography are associated with discomfort and often lower sleep quality caused by the acquisition setup. Previous works have focused on developing less obtrusive methods to conduct high-quality sleep studies, and ear-EEG is among popular alternatives. However, the performance of sleep staging based on ear-EEG is still inferior to scalp-EEG based sleep staging. In order to address the performance gap between scalp-EEG and ear-EEG based sleep staging, we propose a cross-modal knowledge distillation strategy, which is a domain adaptation approach. Our experiments and analysis validate the effectiveness of the proposed approach with existing architectures, where it enhances the accuracy of the ear-EEG based sleep staging by 3.46% and Cohen's kappa coefficient by a margin of 0.038.
Transformers combined with convolutional encoders have been recently used for hand gesture recognition (HGR) using micro-Doppler signatures. We propose a vision-transformer-based architecture for HGR with multi-antenna continuous-wave Doppler radar receivers. The proposed architecture consists of three modules: a convolutional encoderdecoder, an attention module with three transformer layers, and a multi-layer perceptron. The novel convolutional decoder helps to feed patches with larger sizes to the attention module for improved feature extraction. Experimental results obtained with a dataset corresponding to a two-antenna continuous-wave Doppler radar receiver operating at 24 GHz (published by Skaria et al.) confirm that the proposed architecture achieves an accuracy of 98.3% which substantially surpasses the state-of-the-art on the used dataset.
* Accepted to be published in IEEE Sensors Letters, 4 pages
Accurate sleep stage classification is significant for sleep health assessment. In recent years, several deep learning and machine learning based sleep staging algorithms have been developed and they have achieved performance on par with human annotation. Despite improved performance, a limitation of most deep-learning based algorithms is their Black-box behavior, which which have limited their use in clinical settings. Here, we propose Cross-Modal Transformers, which is a transformer-based method for sleep stage classification. Our models achieve both competitive performance with the state-of-the-art approaches and eliminates the Black-box behavior of deep-learning models by utilizing the interpretability aspect of the attention modules. The proposed cross-modal transformers consist of a novel cross-modal transformer encoder architecture along with a multi-scale 1-dimensional convolutional neural network for automatic representation learning. Our sleep stage classifier based on this design was able to achieve sleep stage classification performance on par with or better than the state-of-the-art approaches, along with interpretability, a fourfold reduction in the number of parameters and a reduced training time compared to the current state-of-the-art. Our code is available at https://github.com/Jathurshan0330/Cross-Modal-Transformer.
Computational optical imaging (COI) systems leverage optical coding elements (CE) in their setups to encode a high-dimensional scene in a single or multiple snapshots and decode it by using computational algorithms. The performance of COI systems highly depends on the design of its main components: the CE pattern and the computational method used to perform a given task. Conventional approaches rely on random patterns or analytical designs to set the distribution of the CE. However, the available data and algorithm capabilities of deep neural networks (DNNs) have opened a new horizon in CE data-driven designs that jointly consider the optical encoder and computational decoder. Specifically, by modeling the COI measurements through a fully differentiable image formation model that considers the physics-based propagation of light and its interaction with the CEs, the parameters that define the CE and the computational decoder can be optimized in an end-to-end (E2E) manner. Moreover, by optimizing just CEs in the same framework, inference tasks can be performed from pure optics. This work surveys the recent advances on CE data-driven design and provides guidelines on how to parametrize different optical elements to include them in the E2E framework. Since the E2E framework can handle different inference applications by changing the loss function and the DNN, we present low-level tasks such as spectral imaging reconstruction or high-level tasks such as pose estimation with privacy preserving enhanced by using optimal task-based optical architectures. Finally, we illustrate classification and 3D object recognition applications performed at the speed of the light using all-optics DNN.
With applications ranging from metabolomics to histopathology, quantitative phase microscopy (QPM) is a powerful label-free imaging modality. Despite significant advances in fast multiplexed imaging sensors and deep-learning-based inverse solvers, the throughput of QPM is currently limited by the speed of electronic hardware. Complementarily, to improve throughput further, here we propose to acquire images in a compressed form such that more information can be transferred beyond the existing electronic hardware bottleneck. To this end, we present a learnable optical compression-decompression framework that learns content-specific features. The proposed differentiable optical-electronic quantitative phase microscopy ($\partial \mu$) first uses learnable optical feature extractors as image compressors. The intensity representation produced by these networks is then captured by the imaging sensor. Finally, a reconstruction network running on electronic hardware decompresses the QPM images. The proposed system achieves compression of $\times$ 64 while maintaining the SSIM of $\sim 0.90$ and PSNR of $\sim 30$ dB. The promising results demonstrated by our experiments open up a new pathway for achieving end-to-end optimized (i.e., optics and electronic) compact QPM systems that provide unprecedented throughput improvements.
Ever since the first microscope by Zacharias Janssen in the late 16th century, scientists have been inventing new types of microscopes for various tasks. Inventing a novel architecture demands years, if not decades, worth of scientific experience and creativity. In this work, we introduce Differentiable Microscopy ($\partial\mu$), a deep learning-based design paradigm, to aid scientists design new interpretable microscope architectures. Differentiable microscopy first models a common physics-based optical system however with trainable optical elements at key locations on the optical path. Using pre-acquired data, we then train the model end-to-end for a task of interest. The learnt design proposal can then be simplified by interpreting the learnt optical elements. As a first demonstration, based on the optical 4-$f$ system, we present an all-optical quantitative phase microscope (QPM) design that requires no computational post-reconstruction. A follow-up literature survey suggested that the learnt architecture is similar to the generalized phase concept developed two decades ago. We then incorporate the generalized phase contrast concept to simplify the learning procedure. Furthermore, this physical optical setup is miniaturized using a diffractive deep neural network (D2NN). We outperform the existing benchmark for all-optical phase-to-intensity conversion on multiple datasets, and ours is the first demonstration of its kind on D2NNs. The proposed differentiable microscopy framework supplements the creative process of designing new optical systems and would perhaps lead to unconventional but better optical designs.
Raw point cloud processing using capsule networks is widely adopted in classification, reconstruction, and segmentation due to its ability to preserve spatial agreement of the input data. However, most of the existing capsule based network approaches are computationally heavy and fail at representing the entire point cloud as a single capsule. We address these limitations in existing capsule network based approaches by proposing PointCaps, a novel convolutional capsule architecture with parameter sharing. Along with PointCaps, we propose a novel Euclidean distance routing algorithm and a class-independent latent representation. The latent representation captures physically interpretable geometric parameters of the point cloud, with dynamic Euclidean routing, PointCaps well-represents the spatial (point-to-part) relationships of points. PointCaps has a significantly lower number of parameters and requires a significantly lower number of FLOPs while achieving better reconstruction with comparable classification and segmentation accuracy for raw point clouds compared to state-of-the-art capsule networks.
We propose a weighted least-square (WLS) method to design autoregressive moving average (ARMA) graph filters. We first express the WLS design problem as a numerically-stable optimization problem using Chebyshev polynomial bases. We then formulate the optimization problem with a nonconvex objective function and linear constraints for stability. We employ a relaxation technique and convert the nonconvex optimization problem into an iterative second-order cone programming problem. Experimental results confirm that ARMA graph filters designed using the proposed WLS method have significantly improved frequency responses compared to those designed using previously proposed WLS design methods.