Post-processing Pipeline Research Articles

Leveraging AI to Automate Detection and Quantification of Extrachromosomal DNA (ecDNA) to Decode Drug Responses.

Traditional drug discovery efforts have largely focused on targeting rapid, reversible protein-mediated adaptations to undermine cancer cells' resistance to therapy. However, cancer cells also exploit DNA-based strategies, typically viewed as slow, irreversible, and unpredictable changes like point mutations or the selection of drug-resistant clones. Contrary to this perception, extrachromosomal DNA (ecDNA) represents a form of DNA alteration that is rapid, reversible, and predictable, playing a crucial role in cancer's adaptive response. In this study, we present a novel post-processing pipeline for the automated detection and quantification of ecDNA in Fluorescence in situ Hybridization (FISH) images using the Microscopy Image Analyzer (MIA) tool. Our approach is particularly designed to monitor ecDNA dynamics during drug treatment, providing a quantitative framework to understand how ecDNA enables cancer cells to swiftly and reversibly adapt to therapeutic pressure. This pipeline not only offers a valuable resource for researchers aiming to automate ecDNA detection in FISH images but also sheds light on the adaptive mechanisms of ecDNA in response to epigenetic remodeling agents like JQ1.

Real-time fMRI neurofeedback of the anterior insula using arterial spin labelling.

Arterial spin labelling (ASL) is the only non-invasive technique that allows absolute quantification of perfusion and is increasingly used in brain activation studies. Contrary to the blood oxygen level-dependent (BOLD) effect ASL measures the cerebral blood flow (CBF) directly. However, the ASL signal has a lower signal-to-noise ratio (SNR), than the BOLD signal, which constrains its utilization in neurofeedback studies. If successful, ASL neurofeedback can be used to aid in the rehabilitation of health conditions with impaired blood flow, for example, stroke. We provide the first ASL-based neurofeedback study incorporating a double-blind, sham-controlled design. A pseudo-continuous ASL (pCASL) approach with background suppression and 3D GRASE readout was combined with a real-time post-processing pipeline. The real-time pipeline allows to monitor the ASL signal and provides real-time feedback on the neural activity to the subject. In total 41 healthy adults (19-56 years) divided into three groups underwent a neurofeedback-based emotion imagery training of the left anterior insula. Two groups differing only in the explicitness level of instruction received real training and a third group received sham feedback. Only those participants receiving real feedback with explicit instruction showed significantly higher absolute CBF values in the trained region during neurofeedback than participants receiving sham feedback. However, responder analyses of percent signal change values show no differences in activation between the three groups. Persisting limitations, such as the lower SNR, confounding effects of arterial transit time and partial volume effects still impact negatively the implementation of ASL neurofeedback.

Automated quantification of wrist bone marrow oedema, pre- and post-treatment, in early rheumatoid arthritis.

Bone inflammation (osteitis) in early RA (ERA) manifests as bone marrow oedema (BME) and precedes the development of bone erosion. In this prospective, single-centre study, we developed an automated post-processing pipeline for quantifying the severity of wrist BME on T2-weighted fat-suppressed MRI. A total of 80 ERA patients [mean age 54 years (s.d. 12), 62 females] were enrolled at baseline and 49 (40 females) after 1 year of treatment. For automated bone segmentation, a framework based on a convolutional neural network (nnU-Net) was trained and validated (5-fold cross-validation) for 15 wrist bone areas at baseline in 60 ERA patients. For BME quantification, BME was identified by Gaussian mixture model clustering and thresholding. BME proportion (%) and relative BME intensity within each bone area were compared with visual semi-quantitative assessment of the RA MRI score (RAMRIS). For automated wrist bone area segmentation, overall bone Sørensen-Dice similarity coefficient was 0.91 (s.d. 0.02) compared with ground truth manual segmentation. High correlation (Pearson correlation coefficient r = 0.928, P < 0.001) between visual RAMRIS BME and automated BME proportion assessment was found. The automated BME proportion decreased after treatment, correlating highly (r = 0.852, P < 0.001) with reduction in the RAMRIS BME score. The automated model developed had an excellent segmentation performance and reliable quantification of both the proportion and relative intensity of wrist BME in ERA patients, providing a more objective and efficient alternative to RAMRIS BME scoring.

Segmentation of process-related contaminations on two-piece abutments using pixel-based machine learning: a new quantification approach?

A reference method for quantifying contaminations on two-piece abutments manufactured using CAD/CAM has not yet been established. In the present in vitro study, a pixel--based machine learning (ML) method for detecting contamination on customized two-piece abutments was investigated and embedded in a semiautomated quantification pipeline. Forty-nine CAD/CAM zirconia abutments were fabricated and bonded to a prefabricated titanium base. All samples were analyzed for contamination by scanning electron microscopy (SEM) imaging followed by pixel--based ML and thresholding (SW) for contamination detection; quantification was performed in the postprocessing pipeline. Wilcoxon signed-rank test and Bland-Altmann plot were applied to compare both methods. The contaminated area fraction was recorded as a percentage. There was no statistically significant difference between the percentages of contamination areas (median = 0.004) measured with ML (median = 0.008) and with SW (median = 0.012), asymptotic Wilcoxon test: P = 0.22. The Bland-Altmann plot demonstrated a mean difference of -0.006% (95% confidence interval [CI] from -0.011% to 0.0001%) with increased values from a contamination area fraction of > 0.03% for ML. Both segmentation methods showed comparable results in evaluating surface cleanliness; pixel-based ML is a promising assessment tool for detecting external contaminations on zirconia abutments. Further studies are required to investigate the clinical performance of this tool.

Modification of brain conductivity in human focal epilepsy: A model-based estimation from stereoelectroencephalography.

We have developed a novel method for estimating brain tissue electrical conductivity using low-intensity pulse stereoelectroencephalography (SEEG) stimulation coupled with biophysical modeling. We evaluated the hypothesis that brain conductivity is correlated with the degree of epileptogenicity in patients with drug-resistant focal epilepsy. We used bipolar low-intensity biphasic pulse stimulation (.2 mA) followed by a postprocessing pipeline for estimating brain conductivity. This processing is based on biophysical modeling of the electrical potential induced in brain tissue between the stimulated contacts in response to pulse stimulation. We estimated the degree of epileptogenicity using a semi-automatic method quantifying the dynamic of fast discharge at seizure onset: the epileptogenicity index (EI). We also investigated how the location of stimulation within specific anatomical brain regions or within lesional tissue impacts brain conductivity. We performed 1034 stimulations of 511 bipolar channels in 16 patients. We found that brain conductivity was lower in the epileptogenic zone (EZ; unpaired median difference = .064, p < .001) and inversely correlated with the epileptogenic index value (p < .001, Spearman rho = -.32). Conductivity values were also influenced by anatomical site, location within lesion, and delay between SEEG electrode implantation and stimulation, and had significant interpatient variability. Mixed model multivariate analysis showed that conductivity is significantly associated with EI (F = 13.45, p < .001), anatomical regions (F = 5.586, p < .001), delay since implantation (F = 14.71, p = .003), and age at SEEG (F = 6.591, p = .027), but not with the type of lesion (F = .372, p = .773) or the delay since last seizure (F = 1.592, p = .235). We provide a novel model-based method for estimating brain conductivity from SEEG low-intensity pulse stimulations. The brain tissue conductivity is lower in EZ as compared to non-EZ. Conductivity also varies significantly across anatomical brain regions. Involved pathophysiological processes may include changes in the extracellular space (especially volume or tortuosity) in epileptic tissue.

Enhancing Armenian Automatic Speech Recognition Performance: A Comprehensive Strategy for Speed, Accuracy, and Linguistic Refinement

This research introduces a comprehensive strategy to enhance the performance of an existing automatic speech recognition (ASR) model, which has been previously documented in published articles. The study sets out to achieve several objectives. Firstly, it concentrates on updating the ASR model by retraining it with new datasets. This involves integrating samples from the latest Common Voice corpus release and data collected independently via the armspeech.com web application. Another key focus lies in optimizing the ASR model for near-real-time processing, intending to improve its speed and efficiency. The proposed adjustments to the model’s architecture aim to balance accuracy and processing speed, which is essential for applications requiring prompt speech recognition. Furthermore, the research explores the integration of Transformer models into the post-processing pipeline to introduce punctuation and capitalization into the ASR output. This step not only enhances the linguistic quality of transcriptions but also improves their readability and usability. In tandem with these advancements, the research presents a systematic approach to gathering, annotating, and storing datasets specifically tailored for punctuation and capitalization tasks. The methodology outlines the acquisition and organization of a dataset conducive to training Transformer models for these linguistic tasks. This comprehensive approach, which encompasses dataset enrichment, architectural modifications, and post-processing enhancements, aims to elevate the ASR model’s accuracy, speed, and linguistic refinement, with a particular focus on addressing the intricacies of the Armenian language. The research contributes valuable insights into the optimization of ASR systems, tackling both language-specific challenges and broader issues related to linguistic post-processing.

Automatic image-based tracking of gadolinium-filled balloon wedge catheters for MRI-guided cardiac catheterization using deep learning.

Magnetic Resonance Imaging (MRI) is a promising alternative to standard x-ray fluoroscopy for the guidance of cardiac catheterization procedures as it enables soft tissue visualization, avoids ionizing radiation and provides improved hemodynamic data. MRI-guided cardiac catheterization procedures currently require frequent manual tracking of the imaging plane during navigation to follow the tip of a gadolinium-filled balloon wedge catheter, which unnecessarily prolongs and complicates the procedures. Therefore, real-time automatic image-based detection of the catheter balloon has the potential to improve catheter visualization and navigation through automatic slice tracking. In this study, an automatic, parameter-free, deep-learning-based post-processing pipeline was developed for real-time detection of the catheter balloon. A U-Net architecture with a ResNet-34 encoder was trained on semi-artificial images for the segmentation of the catheter balloon. Post-processing steps were implemented to guarantee a unique estimate of the catheter tip coordinates. This approach was evaluated retrospectively in 7 patients (6M and 1F, age = 7 ± 5 year) who underwent an MRI-guided right heart catheterization procedure with all images acquired in an orientation unseen during training. The overall accuracy, specificity and sensitivity of the proposed catheter tracking strategy over all 7 patients were 98.4 ± 2.0%, 99.9 ± 0.2% and 95.4 ± 5.5%, respectively. The computation time of the deep-learning-based segmentation step was ∼10 ms/image, indicating its compatibility with real-time constraints. Deep-learning-based catheter balloon tracking is feasible, accurate, parameter-free, and compatible with real-time conditions. Online integration of the technique and its evaluation in a larger patient cohort are now warranted to determine its benefit during MRI-guided cardiac catheterization.

Efficient simulations of ionized ISM emission lines: a detailed comparison between the FIRE high-redshift suite and observations

ABSTRACT The Atacama Large Millimeter/Submillimeter Array (ALMA) in the submillimetre and the JWST in the infrared have achieved robust spectroscopic detections of emission lines from the interstellar medium (ISM) in some of the first galaxies. These unprecedented measurements provide valuable information regarding the ISM properties, stellar populations, galaxy morphologies, and kinematics in these high-redshift galaxies and, in principle, offer powerful tests of state of the art galaxy formation models, as implemented in hydrodynamical simulations. To facilitate direct comparisons between simulations and observations, we develop a fast post-processing pipeline to predict line emission from the H ii regions around simulated star particles, accounting for spatial variations in the surrounding gas density, metallicity, and incident radiation spectrum. Our ISM line emission model currently captures H α, H β, and all of the [O iii] and [O ii] lines targeted by ALMA and JWST at z &amp;gt; 6. We illustrate the power of this approach by applying our line emission model to the publicly available Feedback In Realistic Environments (FIRE) high-z simulation suite and perform a detailed comparison with current observations. We show that the FIRE mass–metallicity relation is in 1σ agreement with ALMA/JWST measurements after accounting for the inhomogeneities in the ISM properties. We also quantitatively validate the description of the one-zone model, which is widely used for interpreting [O iii] and H β line luminosity measurements. This model is publicly available and can be implemented on top of a broad range of galaxy formation simulations for comparison with JWST and ALMA measurements.

On the equivalence of binary phase masks optimized for localization or detection in extended depth-of-field localization microscopy.

Binary annular masks have recently been proposed to extend the depth of field (DoF) of single-molecule localization microscopy. A strategy for designing optimal masks has been introduced based on maximizing the emitter localization accuracy, expressed in terms of Fisher information, over a targeted DoF range. However, the complete post-processing pipeline to localize a single emitter consists of two successive steps: detection, where the regions containing emitters are determined, and localization, where the sub-pixel position of each detected emitter is estimated. Phase masks usually optimize only this second step. The presence of a phase mask also affecting detection, the purpose of this paper is to quantify and mitigate this effect. Using a rigorous framework built from a detection-oriented information theoretical criterion (Bhattacharyya distance), we demonstrate that in most cases of practical significance, annular binary phase masks maximizing Fisher information also maximize the detection probability. This result supports the common design practice consisting of optimizing a phase mask by maximizing Fisher information only.

Comparison of different neurite density metrics with brain asymmetry evaluation

The standard diffusion MRI model with intra- and extra-axonal water pools offers a set of microstructural parameters describing brain white matter architecture. However, non-linearities in the standard model and diffusion data contamination by noise and imaging artefacts make estimation of diffusion metrics challenging. In order to develop reliable diffusion approaches and to avoid computational model degeneracy, additional theoretical assumptions allowing stable numerical implementations are required. Advanced diffusion approaches allow for estimation of intra-axonal water fraction (AWF), describing a key structural characteristic of brain tissue. AWF can be interpreted as an indirect measure or proxy of neurite density and has a potential as useful clinical biomarker. Established diffusion approaches such as white matter tract integrity, neurite orientation dispersion and density imaging (NODDI), and spherical mean technique provide estimates of AWF within their respective theoretical frameworks. In the present study, we estimated AWF metrics using different diffusion approaches and compared measures of brain asymmetry between the different metrics in a sub-sample of 182 subjects from the UK Biobank. Multivariate decomposition by mean of linked independent component analysis revealed that the various AWF proxies derived from the different diffusion approaches reflect partly non-overlapping variance of independent components, with distinct anatomical distributions and sensitivity to age. Further, voxel-wise analysis revealed age-related differences in AWF-based brain asymmetry, indicating less apparent left-right hemisphere difference with higher age. Finally, we demonstrated that NODDI metrics suffer from a quite strong dependence on used numerical algorithms and post-processing pipeline. The analysis based on AWF metrics strongly depends on the used diffusion approach and leads to poorly reproducible results.

Post-Processing Fairness Evaluation of Federated Models: An Unsupervised Approach in Healthcare.

Modern Healthcare cyberphysical systems have begun to rely more and more on distributed AI leveraging the power of Federated Learning (FL). Its ability to train Machine Learning (ML) and Deep Learning (DL) models for the wide variety of medical fields, while at the same time fortifying the privacy of the sensitive information that are present in the medical sector, makes the FL technology a necessary tool in modern health and medical systems. Unfortunately, due to the polymorphy of distributed data and the shortcomings of distributed learning, the local training of Federated models sometimes proves inadequate and thus negatively imposes the federated learning optimization process and in extend in the subsequent performance of the rest Federated models. Badly trained models can cause dire implications in the healthcare field due to their critical nature. This work strives to solve this problem by applying a post-processing pipeline to models used by FL. In particular, the proposed work ranks the model by finding how fair they are by discovering and inspecting micro-Manifolds that cluster each neural model's latent knowledge. The produced work applies a completely unsupervised both model and data agnostic methodology that can be leveraged for general model fairness discovery. The proposed methodology is tested against a variety of benchmark DL architectures and in the FL environment, showing an average 8.75% increase in Federated model accuracy in comparison with similar work.

Diffusion MRI of the facial-vestibulocochlear nerve complex: a prospective clinical validation study

ObjectivesSurgical planning of vestibular schwannoma surgery would benefit greatly from a robust method of delineating the facial-vestibulocochlear nerve complex with respect to the tumour. This study aimed to optimise a multi-shell readout-segmented diffusion-weighted imaging (rs-DWI) protocol and develop a novel post-processing pipeline to delineate the facial-vestibulocochlear complex within the skull base region, evaluating its accuracy intraoperatively using neuronavigation and tracked electrophysiological recordings.MethodsIn a prospective study of five healthy volunteers and five patients who underwent vestibular schwannoma surgery, rs-DWI was performed and colour tissue maps (CTM) and probabilistic tractography of the cranial nerves were generated. In patients, the average symmetric surface distance (ASSD) and 95% Hausdorff distance (HD-95) were calculated with reference to the neuroradiologist-approved facial nerve segmentation. The accuracy of patient results was assessed intraoperatively using neuronavigation and tracked electrophysiological recordings.ResultsUsing CTM alone, the facial-vestibulocochlear complex of healthy volunteer subjects was visualised on 9/10 sides. CTM were generated in all 5 patients with vestibular schwannoma enabling the facial nerve to be accurately identified preoperatively. The mean ASSD between the annotators’ two segmentations was 1.11 mm (SD 0.40) and the mean HD-95 was 4.62 mm (SD 1.78). The median distance from the nerve segmentation to a positive stimulation point was 1.21 mm (IQR 0.81–3.27 mm) and 2.03 mm (IQR 0.99–3.84 mm) for the two annotators, respectively.Conclusionsrs-DWI may be used to acquire dMRI data of the cranial nerves within the posterior fossa.Clinical relevance statementReadout-segmented diffusion-weighted imaging and colour tissue mapping provide 1–2 mm spatially accurate imaging of the facial-vestibulocochlear nerve complex, enabling accurate preoperative localisation of the facial nerve. This study evaluated the technique in 5 healthy volunteers and 5 patients with vestibular schwannoma.Key Points• Readout-segmented diffusion-weighted imaging (rs-DWI) with colour tissue mapping (CTM) visualised the facial-vestibulocochlear nerve complex on 9/10 sides in 5 healthy volunteer subjects.• Using rs-DWI and CTM, the facial nerve was visualised in all 5 patients with vestibular schwannoma and within 1.21–2.03 mm of the nerve’s true intraoperative location.• Reproducible results were obtained on different scanners.

Downfield proton MRSI with whole-brain coverage at 3T.

To develop a 3D downfield (DF) MRSI protocol with whole brain coverage and post-processing pipeline for creation of metabolite maps. A 3D, circularly phase-encoded version of the previously developed 2D DF MRSI sequence with spectral-spatial excitation and frequency selective refocusing was implemented and tested in five healthy volunteers at 3T. The DF metabolite maps with a nominal spatial resolution of 0.7 cm3 were recorded in eight slices at 3T in a scan time of 22 m 40 s. An MRSI post-processing pipeline was developed to create DF metabolite maps. Metabolite concentrations and uncertainty estimates were compared between region differences for nine DF peaks. LCModel analysis showed Cramer Rao lower bounds average values of 3%-4% for protein amide resonances in the three selected regions (anterior cingulate, dorsolateral prefrontal cortex, and centrum semiovale); Cramer Rao lower bounds were somewhat higher for individual peaks but for the most part were less than 20%. While DF concentration maps were visually quite homogeneous throughout the brain, general linear regression analysis corrected for multiple comparisons found significant differences between centrum semiovale and dorsolateral prefrontal cortex for peaks at 7.09 ppm (p = 0.014), 7.90 ppm (p = 0.009), 8.18 ppm (p = 0.009), combined amides (p = 0.009), and between anterior cingulate and dorsolateral prefrontal cortex for the 7.30 ppm peak (p = 0.020). Cramer Rao lower bounds values were not significantly different between brain regions for any of the DF peaks. The 3D DF MRSI of the human brain at 3T with wide spatial coverage for the mapping of exchangeable amide and other resonances is feasible at a nominal spatial resolution of 0.7 cm3 .

Full-Spectrum CARS Microscopy of Cells and Tissues with Ultrashort White-Light Continuum Pulses.

Coherent anti-Stokes Raman scattering (CARS) microscopy is an emerging nonlinear vibrational imaging technique that delivers label-free chemical maps of cells and tissues. In narrowband CARS, two spatiotemporally superimposed picosecond pulses, pump and Stokes, illuminate the sample to interrogate a single vibrational mode. Broadband CARS (BCARS) combines narrowband pump pulses with broadband Stokes pulses to record broad vibrational spectra. Despite recent technological advancements, BCARS microscopes still struggle to image biological samples over the entire Raman-active region (400-3100 cm-1). Here, we demonstrate a robust BCARS platform that answers this need. Our system is based on a femtosecond ytterbium laser at a 1035 nm wavelength and a 2 MHz repetition rate, which delivers high-energy pulses used to produce broadband Stokes pulses by white-light continuum generation in a bulk YAG crystal. Combining such pulses, pre-compressed to sub-20 fs duration, with narrowband pump pulses, we generate a CARS signal with a high (<9 cm-1) spectral resolution in the whole Raman-active window, exploiting both the two-color and three-color excitation mechanisms. Aided by an innovative post-processing pipeline, our microscope allows us to perform high-speed (≈1 ms pixel dwell time) imaging over a large field of view, identifying the main chemical compounds in cancer cells and discriminating tumorous from healthy regions in liver slices of mouse models, paving the way for applications in histopathological settings.

Post-processing methodology for the multimodal study of the left atrium during atrial fibrillation

In the context of a prospective study for the multimodal characterization of the left atrium in patients undergoing atrial fibrillation ablation, we developed a systematic pipeline to allow the co-registration and analysis of three different imaging modalities: LGE-CMR, 3D catheter-based invasive voltage maps and FDG uptake measured by PET. The software SLICER3D was choosed as the primary building block. As a free, open source software, it allows for the quick building and deployment of custom solutions thanks to its python wrapper. Several software solutions were developped in python using both the SLICER3D innate capabilities and the Vizualisation toolkit library (VTK) powerful post-processing power. Building on a previous work from Hohmann S. we developed a program allowing the importation in SLICER3D of CARTOv7 (biosense) voltage maps. Using built-in tools, 3D meshes were created from LGE-CMR and PET studies. Unto those 3D meshes, a custom program calculated a maximum intensity projection of the atrial wall along normal vectors, allowing for the creation of LGE and SUV 3D maps. Lastly, iterative point cloud registration permitted the fusion of all 3 mappings. Point-by-point test of association between low voltage, low FDG uptake and high-LGE was used. We were able to fusion and do a point-by-point analysis of three different imaging studies of the left atrium in atria fibrillation: voltage maps, LGE-CMR, and PET. Software solutions developed for this post-processing pipeline are generic and could be adapted for other multimodalities studies. A prospective study using this methodology and involving patients in AF undergoing ablation is underway and will be presented at the ESC congress 2023 (Fig. 1).

Influence of Residual Quadrupolar Interaction on Quantitative Sodium Brain Magnetic Resonance Imaging of Patients With Multiple Sclerosis.

The purpose of this work was to evaluate the influence of residual quadrupolar interaction on the determination of human brain apparent tissue sodium concentrations (aTSCs) using quantitative sodium magnetic resonance imaging ( 23 Na MRI) in healthy controls (HCs) and patients with multiple sclerosis (MS). Especially, it was investigated if the more detailed examination of residual quadrupolar interaction effects enables further analysis of the observed 23 Na MRI signal increase in MS patients. 23 Na MRI with a 7 T MR system was performed on 21 HC and 50 MS patients covering all MS subtypes (25 patients with relapsing-remitting MS, 14 patients with secondary progressive MS, and 11 patients with primary progressive MS) using 2 different 23 Na pulse sequences for quantification: a commonly used standard sequence (aTSC Std ) as well as a sequence with shorter excitation pulse length and lower flip angle for minimizing signal loss resulting from residual quadrupolar interactions (aTSC SP ). Apparent tissue sodium concentration was determined using the same postprocessing pipeline including correction of the receive profile of the radiofrequency coil, partial volume correction, and relaxation correction. Spin dynamic simulations of spin-3/2 nuclei were performed to aid in the understanding of the measurement results and to get deeper insight in the underlying mechanisms. In normal-appearing white matter (NAWM) of HC and all MS subtypes, the aTSC SP values were approximately 20% higher than the aTSC Std values ( P < 0.001). In addition, the ratio aTSC SP /aTSC Std was significantly higher in NAWM than in normal-appearing gray matter (NAGM) for all subject cohorts ( P < 0.002). In NAWM, aTSC Std values were significantly higher in primary progressive MS compared with HC ( P = 0.01) as well as relapsing-remitting MS ( P = 0.03). However, in contrast, no significant differences between the subject cohorts were found for aTSC SP . Spin simulations assuming the occurrence of residual quadrupolar interaction in NAWM were in good accordance with the measurement results, in particular, the ratio aTSC SP /aTSC Std in NAWM and NAGM. Our results showed that residual quadrupolar interactions in white matter regions of the human brain have an influence on aTSC quantification and therefore must be considered, especially in pathologies with expected microstructural changes such as loss of myelin in MS. Furthermore, the more detailed examination of residual quadrupolar interactions may lead to a better understanding of the pathologies themselves.

Computational Optics for Mobile Terminals in Mass Production.

Correcting the optical aberrations and the manufacturing deviations of cameras is a challenging task. Due to the limitation on volume and the demand for mass production, existing mobile terminals cannot rectify optical degradation. In this work, we systematically construct the perturbed lens system model to illustrate the relationship between the deviated system parameters and the spatial frequency response (SFR) measured from photographs. To further address this issue, an optimization framework is proposed based on this model to build proxy cameras from the machining samples' SFRs. Engaging with the proxy cameras, we synthetic data pairs, which encode the optical aberrations and the random manufacturing biases, for training the learning-based algorithms. In correcting aberration, although promising results have been shown recently with convolutional neural networks, they are hard to generalize to stochastic machining biases. Therefore, we propose a dilated Omni-dimensional dynamic convolution (DOConv) and implement it in post-processing to account for the manufacturing degradation. Extensive experiments which evaluate multiple samples of two representative devices demonstrate that the proposed optimization framework accurately constructs the proxy camera. And the dynamic processing model is well-adapted to manufacturing deviations of different cameras, realizing perfect computational photography. The evaluation shows that the proposed method bridges the gap between optical design, system machining, and post-processing pipeline, shedding light on the joint of image signal reception (lens and sensor) and image signal processing (ISP).

Deep Learning to Detect Pancreatic Cystic Lesions on Abdominal Computed Tomography Scans: Development and Validation Study.

Pancreatic cystic lesions (PCLs) are frequent and underreported incidental findings on computed tomography (CT) scans and can evolve to pancreatic cancer-the most lethal cancer, with less than 5 months of life expectancy. The aim of this study was to develop and validate an artificial deep neural network (attention gate U-Net, also named "AGNet") for automated detection of PCLs. This kind of technology can help radiologists to cope with an increasing demand of cross-sectional imaging tests and increase the number of PCLs incidentally detected, thus increasing the early detection of pancreatic cancer. We adapted and evaluated an algorithm based on an attention gate U-Net architecture for automated detection of PCL on CTs. A total of 335 abdominal CTs with PCLs and control cases were manually segmented in 3D by 2 radiologists with over 10 years of experience in consensus with a board-certified radiologist specialized in abdominal radiology. This information was used to train a neural network for segmentation followed by a postprocessing pipeline that filtered the results of the network and applied some physical constraints, such as the expected position of the pancreas, to minimize the number of false positives. Of 335 studies included in this study, 297 had a PCL, including serous cystadenoma, intraductal pseudopapillary mucinous neoplasia, mucinous cystic neoplasm, and pseudocysts . The Shannon Index of the chosen data set was 0.991 with an evenness of 0.902. The mean sensitivity obtained in the detection of these lesions was 93.1% (SD 0.1%), and the specificity was 81.8% (SD 0.1%). This study shows a good performance of an automated artificial deep neural network in the detection of PCL on both noncontrast- and contrast-enhanced abdominal CT scans.

Post-processing CHARIS integral field spectrograph data with pyklip

Abstract We present the pyKLIP-CHARIS post-processing pipeline, a Python library that reduces high contrast imaging data for the CHARIS integral field spectrograph used with the SCExAO project on the Subaru Telescope. The pipeline is a part of the pyklip package, a Python library dedicated to the reduction of direct imaging data of exoplanets, brown dwarfs, and discs. For PSF subtraction, the pyKLIP-CHARIS post-processing pipeline relies on the core algorithms implemented in pyklip but uses image registration and calibrations that are unique to CHARIS. We describe the pipeline procedures, calibration results, and capabilities in processing imaging data acquired via the angular differential imaging and spectral differential imaging observing techniques. We showcase its performance on extracting spectra of injected synthetic point sources as well as compare the extracted spectra from real data sets on HD 33632 and HR 8799 to results in the literature. The pipeline is a python-based complement to the SCExAO project supported, widely used (and currently IDL-based) CHARIS data post-processing pipeline (CHARIS DPP) and provides an additional approach to reducing CHARIS data and extracting calibrated planet spectra.

Clinical validation of rapid GPU-enabled DTI tractography of the brain

Diffusion tensor imaging (DTI) is a non-invasive magnetic resonance imaging (MRI) modality used to map white matter fiber tracts for a variety of clinical applications; one of which is aiding preoperative assessments for tumor patients. DTI requires numerical computations on multiple diffusion weighted images to calculate diffusion tensors at each voxel and probabilistic tracking&lt;sup&gt;1&lt;/sup&gt; to construct fiber tracts, or tractography. Greater accuracy in tractography is possible with larger, more advanced imaging and reconstruction algorithms. However, larger scans and advanced reconstruction is often computationally intensive. The post-processing pipeline involves significant computational resources and time and requires up to 40 minutes of computation time on state-of-the-art hardware. Parallel GPU computations can improve time for the resource-intensive tractography. A collaborative team from DIPY, NVIDIA, and UCSF recently developed a tool, GPUStreamlines, for GPU-enabled tractography&lt;sup&gt;2&lt;/sup&gt; which has been expanded to support the constant solid angle (CSA) reconstruction algorithm&lt;sup&gt;3&lt;/sup&gt;. This GPU-enabled tractography was applied to MRIs of brains with and without presence of lesions, with substantial increases in processing speed. We demonstrate that CSA GPU-enabled tractography in normal controls and patients are comparable to the existing gold standard tractography currently in place at UCSF.