Control Vision (United States)

Future space missions require more autonomous capabilities. Modern autonomy-enabling software, such as deep learning algorithmic approaches, offers the potential to enhance current and future missions but only if these algorithms can be matched to the constraints and requirements of flight hardware, including limited size, weight, power, memory, and radiation tolerance. An important option to accelerate the deployment of deep learning onboard is the hybridization across software and firmware in embedded flight computers. In this paper we present one of the first cases of hybridization of a convolutional neural network in space, validated onboard ESA’s OPS-SAT with the SmartCam convolutional neural network. We use a custom compiler and runtime designed to deploy neural networks across different hardware targets for space missions and implement a generic matrix multiplier to execute multiply-accumulate cycles from the SmartCam architecture on the onboard field programmable gate array (FPGA). We demonstrate identical output classification performance between the original and hybridized implementations of SmartCam with one hundred images taken in orbit.

Histiocytic sarcoma (HS) represents a rare type of malignant disease, characterized by neoplastic proliferation of interstitial dendritic cells (iDCs). In rabbits, there are only two previous reports of HS with pulmonary and cutaneous primary localizations. A 2-year-old, male, European rabbit (Oryctolagus cuniculus) with a clinical history of chronic respiratory distress and progressive weight loss was necropsied. Postmortem examination revealed a gray-white, well-demarcated ovoid mass that deformed the ventral region of the larynx. Similar nodules were identified in both lungs. Histologically, the laryngeal mass and pulmonary nodules were composed of dense sheets of neoplastic histiocytes with marked atypia and multinucleated neoplastic cells. Immunohistochemically, the neoplastic cells were intensely positive for Iba1, CD1a, and vimentin, occasionally for E-cadherin, and negative for multi-cytokeratin. A diagnosis of laryngeal HS with multiple pulmonary metastases was made. To the authors' knowledge, laryngeal histiocytic tumor has not been previously reported in rabbits.

To provide safety and effectiveness outcomes 12 months postsurgically for sequential canaloplasty and trabeculotomy with the OMNI system combined with cataract surgery in mild-to-moderate open-angle glaucoma (OAG). Eleven ophthalmology practices and surgery centers located in 8 U.S. states. Retrospective, multicenter, single arm. Twelve surgeons contributed 81 patients meeting eligibility criteria: OAG, 12-month follow-up, and medicated intraocular pressure (IOP) 36 mm Hg or less on 4 or less medications preoperatively. Analysis was stratified by baseline IOP: more than 18 mm Hg (Group 1) and 18 mm Hg (Group 2) or less. Success was defined as proportion with 20% reduction or greater in IOP or IOP between 6 and 18 mm Hg (inclusive) and on the same or fewer medications without secondary surgical intervention (SSI). Other endpoints included mean IOP and medications at 12 months. Safety included corrected distance visual acuity and adverse events (AEs). Of the 81 patients included, primary success was met by 79% in Group 1 and 81% in Group 2. The mean IOP was reduced in Group 1 (21.9 to 15.1 mm Hg, P < .0001) and remained controlled in Group 2 (14.1 to 13.4 mm Hg, P = .3177). Medications went from 2.0 ± 1.3 to 1.1 ± 1.1 in Group 1 and from 1.6 ± 1.3 to 0.9 ± 1.2 in Group 2. AEs were typical for cataract or angle surgery: mild inflammation (11%), IOP spikes (5%), and hyphema (4%). Four patients (5%) required an SSI. The OMNI system provided effective IOP reduction, sustained IOP control, and meaningful medication reduction for up to 12 months postoperative.

3D Augmented Reality (AR) Heads‐up Displays (HUDs) have the potential of overlaying virtual objects at the correct locations with accurate motion parallax. Accurate overlays require tracking the pupils of the driver's eyes. We developed an ML‐based pupil tracking system based on a convolutional neural network (CNN) to find the precise location of the pupils.

.Significance: Hyperspectral imaging (HSI) can support intraoperative perfusion assessment, the identification of tissue structures, and the detection of cancerous lesions. The practical use of HSI for minimal-invasive surgery is currently limited, for example, due to long acquisition times, missing video, or large set-ups.Aim: An HSI laparoscope is described and evaluated to address the requirements for clinical use and high-resolution spectral imaging.Approach: Reflectance measurements with reference objects and resected human tissue from 500 to 1000 nm are performed to show the consistency with an approved medical HSI device for open surgery. Varying object distances are investigated, and the signal-to-noise ratio (SNR) is determined for different light sources.Results: The handheld design enables real-time processing and visualization of HSI data during acquisition within 4.6 s. A color video is provided simultaneously and can be augmented with spectral information from push-broom imaging. The reflectance data from the HSI system for open surgery at 50 cm and the HSI laparoscope are consistent for object distances up to 10 cm. A standard rigid laparoscope in combination with a customized LED light source resulted in a mean SNR of 30 to 43 dB (500 to 950 nm).Conclusions: Compact and rapid HSI with a high spatial- and spectral-resolution is feasible in clinical practice. Our work may support future studies on minimally invasive HSI to reduce intra- and postoperative complications.

Wheat is susceptible to contamination by deoxynivalenol (DON) which is regarded as a class III carcinogen. In this paper, a rapid and nondestructive method for DON content determination and contamination degree discrimination in wheat was developed by using a multispectral imaging (405-970 nm) system. Genetic algorithm (GA) and principal component analysis (PCA), as preprocessing methods, were used to obtain the best spectral characteristics. The determination model was established by combining preprocessing methods and chemometric methods including partial least squares (PLS), support vector machines (SVM) and back propagation neural network (BPNN). The best quantitative determination result was obtained based on GA-SVM with a correlation coefficient of prediction (Rp), root mean square error of prediction (RMSEP) and residual predictive deviation (RPD) of 0.9988, 365.3 μg kg-1 and 8.6, respectively. Furthermore, the accuracy of contamination degree classification was up to 94.29% in the prediction set by using the PCA-PLS model. The results showed that the combination of multispectral imaging technology and chemometrics was an effective and nondestructive method for the determination of DON in wheat.

Modern large displacement optical flow algorithms usually use an initialization by either sparse descriptor matching techniques or dense approximate nearest neighbor fields. While the latter have the advantage of being dense, they have the major disadvantage of being very outlier-prone as they are not designed to find the optical flow, but the visually most similar correspondence. In this article we present a dense correspondence field approach that is much less outlier-prone and thus much better suited for optical flow estimation than approximate nearest neighbor fields. Our approach does not require explicit regularization, smoothing (like median filtering) or a new data term. Instead we solely rely on patch matching techniques and a novel multi-scale matching strategy. We also present enhancements for outlier filtering. We show that our approach is better suited for large displacement optical flow estimation than modern descriptor matching techniques. We do so by initializing EpicFlow with our approach instead of their originally used state-of-the-art descriptor matching technique. We significantly outperform the original EpicFlow on MPI-Sintel, KITTI 2012, KITTI 2015 and Middlebury. In this extended article of our former conference publication we further improve our approach in matching accuracy as well as runtime and present more experiments and insights.

ImportanceDemonstrating that success of Descemet stripping automated endothelial keratoplasty is similar across donor cornea preservation times (PTs) could increase the donor pool.ObjectiveTo determine whether the 3-year rate of graft success using corneal donor tissue preserved 8 to 14 days is noninferior to that of donor tissue preserved 7 days or less.Design, Setting, and ParticipantsA multicenter, double-masked, randomized noninferiority clinical trial was conducted from April 16, 2012, to June 5, 2017, at 40 clinical sites (70 surgeons) in the United States, with donor corneas provided by 23 US eye banks. A total of 1090 individuals (1330 study eyes) underwent Descemet stripping automated endothelial keratoplasty (1255 eyes [94.4%] for Fuchs endothelial corneal dystrophy).InterventionsDescemet stripping automated endothelial keratoplasty with random assignment of a donor cornea with a PT of 7 days or less (0-7d PT) or 8 to 14 days (8-14d PT).Main Outcomes and MeasuresGraft success at 3 years.ResultsOf the 1090 participants (1330 study eyes; 60.2% women and 39.8% men; median age at enrollment, 70 years [range, 42-90 years]), the 3-year cumulative probability of graft success was 95.3% (95% CI, 93.6%-96.9%) in the 0-7d PT group and 92.1% (95% CI, 89.9%-94.2%) in the 8-14d PT group (difference, 3.2%). The upper limit of the 1-sided 95% CI on the difference was 5.4%, exceeding the prespecified noninferiority limit of 4%. The difference was mostly owing to more primary donor failures in the 8-14d PT group, with the conditional probability of failure after the first month being 2.4% in the 0-7d PT group and 3.1% in the 8-14d PT group. In preplanned secondary analyses, longer PT was associated with a lower rate of graft success (unadjusted hazard ratio for graft failure per additional day of PT, 1.10; 95% CI, 1.03-1.18; P = .008 [PT analyzed as days]), with success rates of 96.5% (95% CI, 92.3%-98.4%) for PT of 4 days or less, 94.9% (95% CI, 92.5%-96.6%) for PT of 5 to 7 days, 93.8% (95% CI, 91.0%-95.8%) for PT of 8 to 11 days, and 89.3% (95% CI, 84.4%-92.7%) for PT of 12 to 14 days (P = .01 [PT analyzed as categorical variable]).Conclusions and RelevanceThe 3-year success rate in eyes undergoing Descemet stripping automated endothelial keratoplasty was high irrespective of PT. However, the study was unable to conclude that the success rate with donor corneas preserved 8 to 14 days was similar to that of corneas preserved 7 days or less with respect to the prespecified noninferiority limit. Although longer PT was associated with a lower success rate, the difference in rates was small when PT was less than 12 days.Trial Registrationclinicaltrials.gov Identifier: NCT01537393

Wide color gamut displays on the market were studied. TVs displaying over 90% NTSC gamut area and monitors displaying over 100% Adobe coverage were selected. CdSe QD based solutions achieve the highest color and energy efficiency, which is linked to fundamental attributes of QDs: narrow emission, efficiency, and operational stability.