Search

The advancement in bio-engineering technology has enabled tissues to be artificially cultivated from human cells, providing the opportunity to model disease and discover potential treatments 1 . Blood vessel is an important category of human tissues that can be artificially engineered to facilitate the development of treatment plans for vascular diseases. The growth of tissue engineered blood vessels (TEBVs) is a costly procedure, and effective quality control during the growing process could help reduce waste and optimize the cultivation process. Imaging technologies, such as optical coherence tomography5,6 (OCT), have been applied to obtain cross-sectional images of TEBVs, which could be used as a nondestructive method to assess blood vessel during cultivation. Ultrasound (US) imaging has been widely accepted in clinical practice due to its real-time imaging capacity and zero radiation emission; and compared to optics-based imaging modality it is more accessible financially. We implemented an US computer tomography (USCT) based monitoring system on assisting quality control in TEBV growth. In this prototype, a single element transducer is placed in a circular stand that rotates around the TEBV bioreactor to collect A-lines from different angles. Mechatronics systems are used to actuate the transducer for circular motion. A circular back-projection method is used in image reconstruction. Experiments were carried out with point phantom and the bioreactor to validate the imaging functionality of the prototype. Reconstructed images provide validation to the feasibility of using USCT to monitor the growth of TEBV growth.

The conventional transrectal ultrasound (TRUS) probe for transperineal needle insertion is typically a sagittal viewing linear array or combined with another transverse curvilinear array. Physicians are required to manually move and rotate the probe back and forth to determine the needle trajectory or to render 3D volume. For the hands-free 3D imaging guidance of transperineal needle insertion, this paper proposes a reflector-based 3D TRUS imaging probe. The proposed probe is enabled by motorizing rotation of a forward-viewing array transducer and bending the ultrasound beam to cover whole prostate utilizing an acoustic mirror as reflector. The proposed device is capable of providing an arbitrary angled B-mode slice with respect to prostate based on the motorized probe rotation angle adjustment, and enables to render 3D images with volume-based reconstruction. In the experimental validation with the cyst phantom, the contrast-noise ratios in the 0 and 90-degree of rotation angle were 4.36 and 4.05 dB in the image acquisition with reflector, and 4.26 and 5.07 dB in that without reflector, respectively. Also, the 3D appearance of cyst can be rendered with maximum intensity projection. These results demonstrated the limited image quality deterioration by adding reflector, and indicate the reflector-based TRUS probe has a potential to improve the guidance of transperineal needle insertion.

Making sophisticated, robust, and safe sequential decisions is at the heart of intelligent systems. This is especially critical for planning in complex multi-agent environments, where agents need to anticipate other agents' intentions and possible future actions. Traditional methods formulate the problem as a Markov Decision Process, but the solutions often rely on various assumptions and become brittle when presented with corner cases. In contrast, deep reinforcement learning (Deep RL) has been very effective at finding policies by simultaneously exploring, interacting, and learning from environments. Leveraging the powerful Deep RL paradigm, we demonstrate that an iterative procedure of self-play can create progressively more diverse environments, leading to the learning of sophisticated and robust multi-agent policies. We demonstrate this in a challenging multi-agent simulation of merging traffic, where agents must interact and negotiate with others in order to successfully merge on or off the road. While the environment starts off simple, we increase its complexity by iteratively adding an increasingly diverse set of agents to the agent zoo as training progresses. Qualitatively, we find that through self-play, our policies automatically learn interesting behaviors such as defensive driving, overtaking, yielding, and the use of signal lights to communicate intentions to other agents. In addition, quantitatively, we show a dramatic improvement in the success rate of merging maneuvers from 63% to over 98%.

Methods for ''cleaning up'' (or recognizing) states of a neural network are crucial for the functioning of many neural cognitive models. This process takes a noisy approximation of a state and recovers the original information. As a particular example, we consider the cleanup required for the use of Vector Symbolic Architectures, which provide a method for manipulating symbols using a fixed-length vector representation. To recognize the result of these manipulations, a mechanism for cleaning up the resulting noisy representation is needed, as this noise increases with the number of symbols being combined. While these symbolic manipulations have previously been modelled with biologically plausible neurons, this paper presents the first spiking neuron model of the cleanup process. We demonstrate that it approaches ideal performance and that the neural requirements scale linearly with the number of distinct symbols in the system. While this result is relevant for any biological model requiring cleanup, it is crucial for VSAs, as it completes the set of mechanisms needed to provide a full neural implementation of symbolic reasoning.

While Boltzmann Machines have been successful at unsupervised learning and density modeling of images and speech data, they can be very sensitive to noise in the data. In this paper, we introduce a novel model, the Robust Boltzmann Machine (RoBM), which allows Boltzmann Machines to be robust to corruptions. In the domain of visual recognition, the RoBM is able to accurately deal with occlusions and noise by using multiplicative gating to induce a scale mixture of Gaussians over pixels. Image denoising and in-painting correspond to posterior inference in the RoBM. Our model is trained in an unsupervised fashion with unlabeled noisy data and can learn the spatial structure of the occluders. Compared to standard algorithms, the RoBM is significantly better at recognition and denoising on several face databases.