Skip to main content
Search

# Convolutional Neural Network

15 related articles

Subscribe to updates
  1. medRxiv

    Focal dystonia gene hunt gets a new entry point: DystoniaGeneNet flags familial MRI signatures, then three new genes

    Isolated focal dystonia has fewer than five known causal genes, and small families with incomplete penetrance make linkage analysis hard. A 3D convolutional network, DystoniaGeneNet, judged familial status from T1-weighted brain MRI and correctly classified 18 of 21 independent familial cases (85.7%), then guided exome variant prioritization and segregation analysis in 214 individuals. Rare variants segregating with disease emerged in RAP1GAP, TNN and ITSN2, all converging on perisynaptic extracellular matrix–synapse signalling.

  2. Glioblastoma invasion gets a live-brain-slice tracker: thousands of cells followed as standard therapy enriches the fastest branching state

    Glioblastoma cells switch among branching, diffuse translocation, locomotion and perivascular invasion modes, but these switches were hard to observe at scale over days in intact brain tissue. GlioTrace links whole-specimen confocal imaging, convolutional-network morphology classification and a covariate hidden Markov model to track thousands of cells across six patient-derived xenograft brain slices. Invasion plasticity proved patient-specific and spatially organized, modulated by vascular proximity and microglial contact; targeted drugs shifted phenotype composition and transition probabilities, while radiation plus temozolomide enriched the fastest, highly connected branching state.

  3. bioRxiv

    One stretch of DNA, four evolutionary forces: multiscale haplotype images lift classification accuracy to 90.85%

    Telling neutrality, selective sweeps, balancing selection and adaptive introgression apart is hard because the haplotype footprints these processes leave overlap. SHADES renders haplotype frequencies, differences between frequency distributions, and sequence differences between ranked haplotypes as three image types, trains convolutional networks at window sizes of 11, 21 and 51 SNPs, and stacks the three scales for four-class classification. On simulated test replicates, the deeper SHADES[M2] reached 90.85% overall accuracy, against 67.83% for SURFDAWave, 68.30% for diploS/HIC, 65.20% for TrIdent, 63.20% for FAST-NN and 79.43% for FASTER-NN on the same test set.

  4. Multi-pinhole SPECT drops the CT attenuation map: one network separates overlapping projections and corrects attenuation and scatter

    Multi-pinhole SPECT gains sensitivity but stacks overlapping projections with attenuation, scatter, quantum noise and aperture effects, and attenuation correction usually needs an attenuation coefficient map from X-ray CT. A U-Net++ convolutional network takes Monte Carlo-simulated overlapping projections as input and ray-traced ideal projections as target, separating overlaps and correcting the degradations together before maximum-likelihood expectation-maximization reconstruction. On 60 public brain cases under quasi-stationary acquisition, it reached NMSE 0.18±0.02, PSNR 18.42±0.69 dB and SSIM 0.81±0.04, better than 11-pinhole reconstruction and fan-beam reconstruction with 20-fold longer acquisition.

  5. ECG models break on filter phase, not filtering: a causal all-pass shift drops conduction-block AUROC from 0.91 to 0.62

    Real-time ECG monitoring must use causal filters, whose nonlinear phase misaligns the frequency components of the cardiac waveform. Across 36,583 twelve-lead records from three independent clinical databases, changing phase alone while holding spectral magnitude fixed cut ResNet-18 macro-AUROC by 0.1244 on PTB-XL, with conduction-block detection falling from above 0.91 to about 0.62. A training-time regularizer that randomly applies all-pass phase perturbations recovered 86.9% to 90.9% of that loss across cohorts without changing the architecture.

  6. Low-dose CT denoising drops the clean-image pairs: align adjacent slices, and cone-beam CT works in image domain

    Low-dose CT denoising can now be trained from adjacent slices alone: a matching network estimates inter-slice deformation and corrects the denoiser output, reaching 40.08 dB PSNR on Mayo 2016 and, for the first time, image-domain self-supervised low-dose cone-beam CT denoising.

  7. arXiv

    Ship track forecasts move to Transformers: 0.18 km average error, prediction interval in one pass

    Ship trajectory forecasting no longer needs recurrent networks or repeated sampling. Earlier AIS-based models either used only the target vessel's history or relied on clustering other vessels, and approximated probability intervals through multiple forward passes. Here the vessel's past positions and a local KDE heat map are fed to a Transformer trained with quantile loss, reaching about 0.18 km average displacement error across three Danish water areas while emitting a prediction interval in a single forward pass.

  8. arXiv

    T0 detector doubles as beam monitor: 32 SiPM channels regress 2D position, about 2.4 mm in experiment

    The T0 detector of the HIRFL-CSR external-target experiment previously supplied only the event start time; its 32 SiPM over-threshold times and TOT values are reorganized into a geometry-preserving tensor, and a CNN regresses the incident coordinates directly. Residual standard deviations are about 1.11 mm and 1.04 mm for uniform planar simulation, about 2.36 mm and 1.34 mm for the laser-scanning geometry in simulation, and about 2.42 mm and 1.41 mm in experiment. The trained model reconstructed a two-dimensional beam spot from CEE zero-field beam data taken at the Institute of Modern Physics, Lanzhou.

  9. arXiv

    Communication signals double as radar: multistatic terahertz echoes identify object shape, holding at three-quarters occlusion

    A multistatic terahertz link treats ordinary communication traffic scattered off an object as the observation, identifying shape at orientations never seen in training, with real gains from more transmitters and accuracy that holds when three quarters of the surface is hidden.

  10. arXiv

    Ground-state electron density as machine-learning input predicts molecular absorption spectra more accurately than molecular geometry

    Using 6874 molecules selected from the QM7 dataset, with ground-state densities computed by DFT and absorption spectra by linear-response TDDFT, this work compares a 3D convolutional neural network taking volumetric electron density as input against graph neural networks (SchNet, DimeNet++, MACE) taking atom-centred geometry as input for predicting normalized absorption spectra, finding that the density CNN achieves a validation correlation of 0.9926 versus 0.9795 for the best geometry-based graph model, reducing residual decorrelation by about two-thirds.