Tag: deep learning

Researchers present CADS, a framework pairing a 22,022-scan dataset with complete annotations of 167 anatomical structures and a whole-body CT ...

A new review links mantis shrimp-inspired helicoidal composites, automated fiber placement, AI-based defect control, and self-healing microparticles to superior impact ...

A systematic review finds hybrid CNN-transformer models lead stroke lesion analysis, but explainable AI appears in only 11 percent of ...

Researchers propose SteRGAN, a GAN-based image steganography framework that achieves PSNR values above 41 dB on ImageNet and Microsoft COCO.

A new review argues that deep learning skin lesion analysis must prove itself through external validation, fairness and clinical utility ...

A systematic review finds that vision-based touchless gesture recognition systems, despite laboratory accuracies above 95%, lose 1 to 23 percentage ...

Researchers built a dual-stream deep learning framework that localizes and classifies facial emotions and binary gender labels together, achieving 94.56% ...

A multinational cohort study using temporal deep learning identified deteriorating walking time as the strongest predictor of dual cognitive-physical decline ...

A systematic review and meta-analysis finds pooled sensitivity of 85.4 percent and specificity of 83.5 percent for deep learning-assisted low-dose ...

A retrospective multicentre study reports that a deep learning model combining three time-point chest CT scans predicted MDR-TB treatment outcomes ...

A physics-informed TS-Mixer neural network predicted measured PEM fuel cell stack power with an R-squared of 0.969 on a 497-hour ...

A retrospective deep learning prototype called XceptionAttnV1 processes multiple data types for oral cavity squamous cell carcinoma but has not ...

A music-aware transformer framework combining multi-scale spectrogram learning and adaptive attention optimization reported accuracies of 93.5%, 99.2%, and 97.1% on ...

A transfer learning framework classified neonatal seizures from EEG topographic maps, with ResNet-18 reaching 97.1 percent subject-wise accuracy on unseen ...

A fine-tuned Segment Anything pipeline matched bespoke networks for thigh muscle segmentation on MRI, while radiologist-flagged errors barely changed volume ...

Researchers report a hybrid BiLSTM framework with wavelet feature extraction and modified Archimedes optimization that classifies composite insulator aging states ...

A new hierarchical feature attention network reconstructs high-resolution images from low-resolution inputs while keeping the model below one million parameters.

A new deep learning pipeline raises yellow rust classification accuracy on wheat leaves from 78 to 87 percent on the ...

A hybrid LSTM and CNN model outperformed traditional time series and individual machine learning models in predicting monthly hepatitis B ...

A systematic benchmark of fifteen activation functions across CT, mammography and MRI deblurring finds that the spatially adaptive Funnel ReLU ...