Search

A major challenge in applying deep learning to medical imaging is the paucity of annotated data. This study demonstrates that synthetic colonoscopy images generated by Generative Adversarial Network (GAN) inversion can be used as training data to improve the lesion classification performance of deep learning models. This approach inverts pairs of images with the same label to a semantically rich & disentangled latent space and manipulates latent representations to produce new synthetic images with the same label. We perform image modality translation (style transfer) between white light and narrowband imaging (NBI). We also generate realistic-looking synthetic lesion images by interpolating between original training images to increase the variety of lesion shapes in the training dataset. We show that these approaches outperform comparative colonoscopy data augmentation techniques without the need to re-train multiple generative models. This approach also leverages information from datasets that may not have been designed for the specific colonoscopy downstream task. E.g. using a bowel prep grading dataset for a polyp classification task. Our experiments show this approach can perform multiple colonoscopy data augmentations, which improve the downstream polyp classification performance over baseline and comparison methods by up to 6%., Comment: 9 pages, 3 figures

There has been an increasing interest in residual feed intake (RFI) as a measure of net feed efficiency in dairy cattle. RFI phenotypes are obtained as residuals from linear regression encompassing relevant factors (i.e., energy sinks) to account for body tissue mobilization. However, fitting energy sink phenotypes as regression variables in standard linear regression was criticized because phenotypes are subject to measurement errors. Multiple-trait models have been proposed which derive RFI by follow-up partial regression. By re-arranging the single-trait linear regression, we showed a causal RFI interpretation underlying the linear regression for RFI. It postulates recursive effects in energy allocation from energy sinks on dry matter intake, but the feedback or simultaneous effects are assumed to be nonexistent. A Bayesian recursive structural equation model was proposed for directly predicting RFI and energy sinks and estimating relevant genetic parameters simultaneously. A simplified Markov chain Monte Carlo algorithm that implemented the Bayesian recursive model was described. The recursive model is asymptotically equivalent to one-step linear regression for RFI, yet extends the analytical capacity to multiple-trait analysis. It is equivalent to Bayesian-implemented, multiple-trait model reparameterized based on Cholesky decomposition of phenotypic (co)variance matrix for evaluating RFI, but varied in assumptions about relationships between energy sinks., Comment: 12 pages, 2 Tables, and 1 Figure

Cows are typically milked two or more times on a test day, but not all these milkings are sampled and weighed. Statistical methods have been proposed to estimate daily yields in dairy cows, centering on various yield correction factors in two broad categories. The initial approach estimated a test-day yield with doubled morning (AM) or evening (PM) yield in the AM-PM milking plans, assuming equal AM and PM milking intervals. However, AM and PM milking intervals can vary, and milk secretion rates may be different between day and night. Additive correction factors (ACF) are evaluated by the average differences between AM and PM milk yield for various milking interval classes (MIC). We show that an ACF model is equivalent to a regression model of daily yield on categorical regressor variables, and a continuous variable for AM or PM yield with a fixed regression coefficient. Similarly, a linear regression model can be implemented as an ACF model with the regression coefficient for AM or PM yield estimated from the data. Multiplicative correction factors (MCF) are ratio of daily yield to yield from single milkings, but their statistical interpretations vary. Overall, MCF were more accurate for estimating daily milk yield than ACF. MCF have biological and statistical challenges. An exponential regression model was proposed as an alternative model for estimating daily milk yield, which improved the accuracy in the present study. Characterization of ACF and MCF showed how ACF and MCF improved the accuracy compared to doubling AM or PM yield as the daily milk yield. The methods were explicitly described to estimate daily milk yield in AM and PM milking plans. Still, the principles are generally applicable to cows milked more than two times a day, and they apply similarly to the estimation of daily fat and protein yields with some necessary modifications., Comment: 14 pages, 2 tables, and 1 figure

Urinary tract infections (UTIs) are a common condition that can lead to serious complications including kidney injury, altered mental status, sepsis, and death. Laboratory tests such as urinalysis and urine culture are the mainstays of UTI diagnosis, whereby a urine specimen is collected and processed to reveal its cellular and chemical composition. This process requires precise specimen collection, handling infectious human waste, controlled urine storage, and timely transportation to modern laboratory equipment for analysis. Holographic lens free imaging (LFI) can measure large volumes of urine via a simple and compact optical setup, potentially enabling automatic urine analysis at the patient bedside. We introduce an LFI system capable of resolving important urine clinical biomarkers such as red blood cells, white blood cells, crystals, casts, and E. Coli in urine phantoms. This approach is sensitive to the particulate concentrations relevant for detecting several clinical urine abnormalities such as hematuria, pyuria, and bacteriuria. We show bacteria concentrations across eight orders of magnitude can be estimated by analyzing LFI measurements. LFI measurements of blood cell concentrations are relatively insensitive to changes in bacteria concentrations of over seven orders of magnitude. Lastly, LFI reveals clear differences between UTI-positive and UTI-negative urine from human patients. Together, these results show promise for LFI as a tool for urine screening, potentially offering early, point-of-care detection of UTI and other pathological processes.

CANVAS caused by RFC1 biallelic expansions is a major cause of inherited sensory neuronopathy. Detection of RFC1 expansion is challenging and CANVAS can be associated with atypical features. We clinically and genetically characterized 50 patients, selected based on the presence of sensory neuronopathy confirmed by EMG. We screened RFC1 expansion by PCR, repeat-primed PCR, and Southern blotting of long-range PCR products, a newly developed method. Neuropathological characterization was performed on the brain and spinal cord of one patient. Most patients (88%) carried a biallelic (AAGGG)n expansion in RFC1. In addition to the core CANVAS phenotype (sensory neuronopathy, cerebellar syndrome, and vestibular impairment), we observed chronic cough (97%), oculomotor signs (85%), motor neuron involvement (55%), dysautonomia (50%), and parkinsonism (10%). Motor neuron involvement was found for 24 of 38 patients (63.1%). First motor neuron signs, such as brisk reflexes, extensor plantar responses, and/or spasticity, were present in 29% of patients, second motor neuron signs, such as fasciculations, wasting, weakness, or a neurogenic pattern on EMG in 18%, and both in 16%. Mixed motor and sensory neuronopathy was observed in 19% of patients. Among six non-RFC1 patients, one carried a heterozygous AAGGG expansion and a pathogenic variant in GRM1. Neuropathological examination of one RFC1 patient with an enriched phenotype, including parkinsonism, dysautonomia, and cognitive decline, showed posterior column and lumbar posterior root atrophy. Degeneration of the vestibulospinal and spinocerebellar tracts was mild. We observed marked astrocytic gliosis and axonal swelling of the synapse between first and second motor neurons in the anterior horn at the lumbar level. The cerebellum showed mild depletion of Purkinje cells, with empty baskets, torpedoes, and astrogliosis characterized by a disorganization of the Bergmann's radial glia. We found neuronal loss in the vagal nucleus. The pars compacta of the substantia nigra was depleted, with widespread Lewy bodies in the locus coeruleus, substantia nigra, hippocampus, entorhinal cortex, and amygdala. We propose new guidelines for the screening of RFC1 expansion, considering different expansion motifs. Here, we developed a new method to more easily detect pathogenic RFC1 expansions. We report frequent motor neuron involvement and different neuronopathy subtypes. Parkinsonism was more prevalent in this cohort than in the general population, 10% versus the expected 1% (p < .001). We describe, for the first time, the spinal cord pathology in CANVAS, showing the alteration of posterior columns and roots, astrocytic gliosis and axonal swelling, suggesting motor neuron synaptic dysfunction.

A modern Fortran implementation of three Dirac operators (Wilson, Brillouin, Susskind) in lattice QCD is presented, based on OpenMP shared-memory parallelization and SIMD pragmas. The main idea is to apply a Dirac operator to $N_v$ vectors simultaneously, to ease the memory bandwidth bottleneck. All index computations are left to the compiler and maximum weight is given to portability and flexibility. The lattice volume, $N_x N_y N_z N_t$, the number of colors, $N_c$, and the number of right-hand sides, $N_v$, are parameters defined at compile time. Several memory layout options are compared. The code performs well on modern many-core architectures (480 Gflop/s, 880 Gflop/s, and 780 Gflop/s with $N_v=12$ for the three operators in single precision on a 72-core KNL processor, a $2\times24$-core Skylake node yields similar results). Explicit run-time tests with CG/BiCGstab inverters confirm that the memory layout is relevant for the KNL, but less so for the Skylake architecture. The ancillary code distribution contains all routines, including the single, double, and mixed precision Krylov space solvers, to render it self-contained and ready-to-use., Comment: v1: 58 pages, 24 tables, 13 figures

The domain of online algorithms with predictions has been extensively studied for different applications such as scheduling, caching (paging), clustering, ski rental, etc. Recently, Bamas et al., aiming for an unified method, have provided a primal-dual framework for linear covering problems. They extended the online primal-dual method by incorporating predictions in order to achieve a performance beyond the worst-case case analysis. In this paper, we consider this research line and present a framework to design algorithms with predictions for non-linear packing problems. We illustrate the applicability of our framework in submodular maximization and in particular ad-auction maximization in which the optimal bound is given and supporting experiments are provided.

The one-loop determination of the coefficient $c_\text{SW}$ of the Wilson quark action has been useful to push the leading cut-off effects for on-shell quantities to $\mathcal{O}(\alpha^2 a)$ and, in conjunction with non-perturbative determinations of $c_\text{SW}$, to $\mathcal{O}(a^2)$, as long as no link-smearing is employed. These days it is common practice to include some overall link-smearing into the definition of the fermion action. Unfortunately, in this situation only the tree-level value $c_\text{SW}^{(0)}=1$ is known, and cut-off effects start at $\mathcal{O}(\alpha a)$. We present some general techniques for calculating one loop quantities in lattice perturbation theory which continue to be useful for smeared-link fermion actions. Specifically, we discuss the application to the 1-loop improvement coefficient $c_\text{SW}^{(1)}$ for overall stout-smeared Wilson fermions., Comment: 7 pages, 1 figure, talk given at the 38th International Symposium on Lattice Field Theory (LATTICE2021), 26th-30th July 2021, Zoom/Gather@Massachusetts Institute of Technology Version 3: Corrected some equations in section 3.2

We use the two-dimensional Schwinger model to investigate how lattice fermions perceive the global topological charge $q\in\mathbf{Z}$ of a given gauge background $U$. After a warm-up part devoted to staggered, Adams, Wilson and naive fermions, we consider Karsten-Wilczek and Borici-Creutz fermions, which are in the class of minimally doubled lattice fermion actions. We focus on the eigenvalue spectrum and the chiralities of the pertinent eigenmodes. Without modification both minimally doubled actions are found to be insensitive to topology, but in either case it is possible to define a suitable species-splitting term to make the resulting operator topology aware., Comment: 8 pages, 7 figures, talk given at the 38th International Symposium on Lattice Field Theory (LATTICE2021), 26th-30th July 2021, Zoom/Gather@Massachusetts Institute of Technology

Illustrations are an essential transmission instrument. For an historian, the first step in studying their evolution in a corpus of similar manuscripts is to identify which ones correspond to each other. This image collation task is daunting for manuscripts separated by many lost copies, spreading over centuries, which might have been completely re-organized and greatly modified to adapt to novel knowledge or belief and include hundreds of illustrations. Our contributions in this paper are threefold. First, we introduce the task of illustration collation and a large annotated public dataset to evaluate solutions, including 6 manuscripts of 2 different texts with more than 2 000 illustrations and 1 200 annotated correspondences. Second, we analyze state of the art similarity measures for this task and show that they succeed in simple cases but struggle for large manuscripts when the illustrations have undergone very significant changes and are discriminated only by fine details. Finally, we show clear evidence that significant performance boosts can be expected by exploiting cycle-consistent correspondences. Our code and data are available on http://imagine.enpc.fr/~shenx/ImageCollation., Comment: accepted to ICDAR 2021

This article reviews deep learning applications in biomedical optics with a particular emphasis on image formation. The review is organized by imaging domains within biomedical optics and includes microscopy, fluorescence lifetime imaging, in vivo microscopy, widefield endoscopy, optical coherence tomography, photoacoustic imaging, diffuse tomography, and functional optical brain imaging. For each of these domains, we summarize how deep learning has been applied and highlight methods by which deep learning can enable new capabilities for optics in medicine. Challenges and opportunities to improve translation and adoption of deep learning in biomedical optics are also summarized.

Oblique plane microscopy (OPM) enables high speed, volumetric fluorescence imaging through a single-objective geometry. While these advantages have positioned OPM as a valuable tool to probe biological questions in animal models, its potential for in vivo human imaging is largely unexplored due to its typical use with exogenous fluorescent dyes. Here we introduce a scattering-contrast oblique plane microscope (sOPM) and demonstrate label-free imaging of blood cells flowing through human capillaries in vivo. The sOPM illuminates a capillary bed in the ventral tongue with an oblique light sheet, and images side- and back- scattered signal from blood cells. By synchronizing sOPM with a conventional capillaroscope, we acquire paired widefield and axial images of blood cells flowing through a capillary loop. The widefield capillaroscope image provides absorption contrast and confirms the presence of red blood cells (RBCs), while the sOPM image may aid in determining whether optical absorption gaps (OAGs) between RBCs have cellular or acellular composition. Further, we demonstrate consequential differences between fluorescence and scattering versions of OPM by imaging the same polystyrene beads sequentially with each technique. Lastly, we substantiate in vivo observations by imaging isolated red blood cells, white blood cells, and platelets in vitro using 3D agar phantoms. These results demonstrate a promising new avenue towards in vivo blood analysis., Comment: 12 pages, 6 figures

We would like to reply to Neuray et al. who report a series of five new patients from four unrelated families with bi-allelic mutations of GRN. Their work nicely completes the few existing reports of similar cases, and refers to our recent publication describing six homozygous GRN pathogenic variant carriers with divergent phenotypes and ages at onset (Huin et al., 2020). In summary, the Letter from Neuray et al., reports valuable findings that lead to better define CLN11 due to bi-allelic GRN pathogenic variants. Despite the small sample number that does not allow statistical analysis, the authors underlined the occurrence of cognitive deterioration and epilepsy. Further study of the CLN11 families with functional brain imaging and neuropsychological examinations may be highly informative for the understanding and the clinical characterization of this rare disease., Comment: Brain - A Journal of Neurology , Oxford University Press (OUP), 2020

Oblique back-illumination capillaroscopy has recently been introduced as a method for high-quality, non-invasive blood cell imaging in human capillaries. To make this technique practical for clinical blood cell counting, solutions for automatic processing of acquired videos are needed. Here, we take the first step towards this goal, by introducing a deep learning multi-cell tracking model, named CycleTrack, which achieves accurate blood cell counting from capillaroscopic videos. CycleTrack combines two simple online tracking models, SORT and CenterTrack, and is tailored to features of capillary blood cell flow. Blood cells are tracked by displacement vectors in two opposing temporal directions (forward- and backward-tracking) between consecutive frames. This approach yields accurate tracking despite rapidly moving and deforming blood cells. The proposed model outperforms other baseline trackers, achieving 65.57% Multiple Object Tracking Accuracy and 73.95% ID F1 score on test videos. Compared to manual blood cell counting, CycleTrack achieves 96.58 $\pm$ 2.43% cell counting accuracy among 8 test videos with 1000 frames each compared to 93.45% and 77.02% accuracy for independent CenterTrack and SORT almost without additional time expense. It takes 800s to track and count approximately 8000 blood cells from 9,600 frames captured in a typical one-minute video. Moreover, the blood cell velocity measured by CycleTrack demonstrates a consistent, pulsatile pattern within the physiological range of heart rate. Lastly, we discuss future improvements for the CycleTrack framework, which would enable clinical translation of the oblique back-illumination microscope towards a real-time and non-invasive point-of-care blood cell counting and analyzing technology., Comment: 10 pages, 6 figures