Multi-channel Input Research Articles

A Novel Temporal Fusion Channel Network with Multi-Channel Hybrid Attention for the Remaining Useful Life Prediction of Rolling Bearings

The remaining useful life (RUL) prediction of rolling bearings is crucial for optimizing maintenance schedules, reducing downtime, and extending machinery lifespan. However, existing multi-channel feature fusion methods do not fully capture the correlations between channels and time points in multi-dimensional sensor data. To address the above problems, this paper proposes a multi-channel feature fusion algorithm based on a hybrid attention mechanism and temporal convolutional networks (TCNs), called MCHA-TFCN. The model employs a dual-channel hybrid attention mechanism, integrating self-attention and channel attention to extract spatiotemporal features from multi-channel inputs. It uses causal dilated convolutions in TCNs to capture long-term dependencies and incorporates enhanced residual structures for global feature fusion, effectively extracting high-level spatiotemporal degradation information. The experimental results on the PHM2012 dataset show that MCHA-TFCN achieves excellent performance, with an average Root-Mean-Square Error (RMSE) of 0.091, significantly outperforming existing methods like the DANN and CNN-LSTM.

Analysis of spatial filtering in neural spatiospectral filters and its dependence on training target characteristics

AbstractMask-based multichannel speech enhancement methods based on artificial neural networks estimate a mask that is applied to the multichannel input signal or a reference channel to obtain the estimated desired signal. For the estimation, both spectral and spatial cues from the multichannel input can be used. However, the interplay of the two inside the neural network is typically unknown. In this contribution, we propose a framework to analyze neural spatiospectral filters (NSSFs) with respect to their capabilities to extract and represent spatial information. We explicitly take the characteristics of the training target signal into account and analyze its effect on the functionality of the NSSF. Using two conceptually different NSSFs as example, we show that not all NSSFs use spatial information under all circumstances and that the training target signal has a significant influence on the spatial filtering behavior of an NSSF. These insights help to assess the signal processing capabilities of neural networks and allow to make informed decisions when configuring, training, and deploying NSSFs.

A Feature Integration Network for Multi-Channel Speech Enhancement.

Multi-channel speech enhancement has become an active area of research, demonstrating excellent performance in recovering desired speech signals from noisy environments. Recent approaches have increasingly focused on leveraging spectral information from multi-channel inputs, yielding promising results. In this study, we propose a novel feature integration network that not only captures spectral information but also refines it through shifted-window-based self-attention, enhancing the quality and precision of the feature extraction. Our network consists of blocks containing a full- and sub-band LSTM module for capturing spectral information, and a global-local attention fusion module for refining this information. The full- and sub-band LSTM module integrates both full-band and sub-band information through two LSTM layers, while the global-local attention fusion module learns global and local attention in a dual-branch architecture. To further enhance the feature integration, we fuse the outputs of these branches using a spatial attention module. The model is trained to predict the complex ratio mask (CRM), thereby improving the quality of the enhanced signal. We conducted an ablation study to assess the contribution of each module, with each showing a significant impact on performance. Additionally, our model was trained on the SPA-DNS dataset using a circular microphone array and the Libri-wham dataset with a linear microphone array, achieving competitive results compared to state-of-the-art models.

LIDeepDet: Deepfake Detection via Image Decomposition and Advanced Lighting Information Analysis

The proliferation of AI-generated content (AIGC) has empowered non-experts to create highly realistic Deepfake images and videos using user-friendly software, posing significant challenges to the legal system, particularly in criminal investigations, court proceedings, and accident analyses. The absence of reliable Deepfake verification methods threatens the integrity of legal processes. In response, researchers have explored deep forgery detection, proposing various forensic techniques. However, the swift evolution of deep forgery creation and the limited generalizability of current detection methods impede practical application. We introduce a new deep forgery detection method that utilizes image decomposition and lighting inconsistency. By exploiting inherent discrepancies in imaging environments between genuine and fabricated images, this method extracts robust lighting cues and mitigates disturbances from environmental factors, revealing deeper-level alterations. A crucial element is the lighting information feature extractor, designed according to color constancy principles, to identify inconsistencies in lighting conditions. To address lighting variations, we employ a face material feature extractor using Pattern of Local Gravitational Force (PLGF), which selectively processes image patterns with defined convolutional masks to isolate and focus on reflectance coefficients, rich in textural details essential for forgery detection. Utilizing the Lambertian lighting model, we generate lighting direction vectors across frames to provide temporal context for detection. This framework processes RGB images, face reflectance maps, lighting features, and lighting direction vectors as multi-channel inputs, applying a cross-attention mechanism at the feature level to enhance detection accuracy and adaptability. Experimental results show that our proposed method performs exceptionally well and is widely applicable across multiple datasets, underscoring its importance in advancing deep forgery detection.

Estimation of CO 2 saturation maps from synthetic seismic data using a deep-learning method with a multi-scale approach

In the context of Carbon Capture and Storage (CCS), monitoring the behaviour of CO 2 within subsurface reservoirs is pivotal for efficient and safe storage operations. This study proposes a new approach to enhance the accuracy of predicting CO 2 saturation maps directly from shot gathers, using a combination of deep learning (DL) and feature extraction methods. We employ a U-Net model, specifically tailored to solve regression tasks, and two-dimensional continuous wavelet transform (2D-CWT) to analyse shot gathers at different scales. We introduce a novel approach using multi-channel input data for the DL model, combining shot gathers and CWT images. The DL model was trained and tested using both single channel and multichannel input. The single channel datasets included shot gathers and CWT images at the first scale, while the multi-channel datasets integrated shot gathers with either two or four scales of CWT. We conducted a sensitivity analysis on the number of training epochs to compare the performance of the model with multichannel and single channel input. Additionally, a Transfer Learning approach was implemented to improve performance in noisy conditions by leveraging knowledge from a pre-trained model on noise-free images. Monte Carlo dropout was used to predict pixel-scale variability and provided a prediction variability with a standard deviation ranging from 0.019 to 0.194, enhancing the robustness of CO 2 saturation estimates. Our results suggest that combining feature extraction using 2D-CWT with the U-Net architecture improves the prediction performance of CO 2 saturation in a synthetic CCS model. Additionally, using multi-channel input instead of single channel is more efficient as it requires less training and prediction time to achieve comparable results. This confirms the effectiveness of 2D-CWT as a pre-processing technique, contributing valuable insights and advances in the field of data-driven geophysical inversion problems.

A noise-enhanced machine fault diagnosis method with multi-channel information fusion

ABSTRACT Condition health monitoring and fault diagnosis is able to safeguard against potential threats to machines. Nevertheless, the acquired signals by using a plenty of sensors are often accompanied by noise and meanwhile the useful information contained in the signals is not both complete and reliable, thereby together reducing the accuracy and stability of the fault diagnosis. Therefore, this paper proposes a noise-enhanced intelligent fault diagnosis method with multi-channel information fusion. Here, it would inject the varying levels of Gaussian noise into multi-channel input data as well as intelligent diagnostic models, with a view to elucidating the benefits of noise. Furthermore, the proposed method is validated through three experiments. The experimental results demonstrate that the diagnostic accuracy of the hydraulic motor and the two different bearings reaches 99.85%, 99.89%, and 99.85% by using the proposed method, respectively, which are increased by 2.50%, 4.09%, and 2.50%, respectively, compared with one without the injection of noise. In addition, by comparing the diagnostic accuracy of ones with single-channel and multi-channel information fusion under the optimal noise level in both cases, it is found that one with multi-channel information fusion shows an average of 3.20% increase than the single-channel one.

A method for predicting remaining useful life using enhanced Savitzky–Golay filter and improved deep learning framework

Ensuring operational integrity in large-scale equipment hinges on effective fault prediction and health management. Prognostics and health management (PHM) face the challenge of accurately predicting remaining useful life (RUL) using multivariate sensor data. Traditional methods often require extensive prior knowledge for indicator construction and processing. Deep learning offers a promising alternative. This study presents a multi-channel multi-scale deep learning approach. Initially, an improved Savitzky‒Golay filter (ISG) addresses challenges posed by large and rapidly changing data volumes, enhancing data preprocessing. Subsequently, a framework integrates convolutional neural networks (CNNs) with long short-term memory (LSTM) to capture hierarchical signal information and make integrated predictions. The CNN extracts spatial features from multi-channel input data, while the LSTM captures temporal dependencies. By fusing outputs from both components, the framework enhances predictive accuracy and robustness for complex operational datasets. Experimental validation on the C-MAPSS dataset tests various fusion methods and CNN depths, determining parameters and evaluating filtering effectiveness. Comparative analyses show promising performance, particularly under dynamic conditions. While not optimal for predicting multiple fault types, it outperforms classical algorithms, especially in single fault type prediction tasks.

Sparse-View Spectral CT Reconstruction and Material Decomposition Based on Multi-Channel SGM.

In medical applications, the diffusion of contrast agents in tissue can reflect the physiological function of organisms, so it is valuable to quantify the distribution and content of contrast agents in the body over a period. Spectral CT has the advantages of multi-energy projection acquisition and material decomposition, which can quantify K-edge contrast agents. However, multiple repetitive spectral CT scans can cause excessive radiation doses. Sparse-view scanning is commonly used to reduce dose and scan time, but its reconstructed images are usually accompanied by streaking artifacts, which leads to inaccurate quantification of the contrast agents. To solve this problem, an unsupervised sparse-view spectral CT reconstruction and material decomposition algorithm based on the multi-channel score-based generative model (SGM) is proposed in this paper. First, multi-energy images and tissue images are used as multi-channel input data for SGM training. Secondly, the organism is multiply scanned in sparse views, and the trained SGM is utilized to generate multi-energy images and tissue images driven by sparse-view projections. After that, a material decomposition algorithm using tissue images generated by SGM as prior images for solving contrast agent images is established. Finally, the distribution and content of the contrast agents are obtained. The comparison and evaluation of this method are given in this paper, and a series of mouse scanning experiments are carried out to verify the effectiveness of the method.

FA-Net: A hierarchical feature fusion and interactive attention-based network for dose prediction in liver cancer patients

Dose prediction is a crucial step in automated radiotherapy planning for liver cancer. Several deep learning-based approaches for dose prediction have been proposed to enhance the design efficiency and quality of radiotherapy plan. However, these approaches usually take CT images and contours of organs at risk (OARs) and planning target volume (PTV) as a multi-channel input and is thus difficult to extract sufficient feature information from each input, which results in unsatisfactory dose distribution. In this paper, we propose a novel dose prediction network for liver cancer based on hierarchical feature fusion and interactive attention. A feature extraction module is first constructed to extract multi-scale features from different inputs, and a hierarchical feature fusion module is then built to fuse these multi-scale features hierarchically. A decoder based on attention mechanism is designed to gradually reconstruct the fused features into dose distribution. Additionally, we design an autoencoder network to generate a perceptual loss during training stage, which is used to improve the accuracy of dose prediction. The proposed method is tested on private clinical dataset and obtains HI and CI of 0.31 and 0.87, respectively. The experimental results are better than those by several existing methods, indicating that the dose distribution generated by the proposed method is close to that approved in clinics. The codes are available at https://github.com/hired-ld/FA-Net.

Multi-Scale Spatial Attention-Based Multi-Channel 2D Convolutional Network for Soil Property Prediction.

Visible near-infrared spectroscopy (VNIR) is extensively researched for obtaining soil property information due to its rapid, cost-effective, and environmentally friendly advantages. Despite its widespread application and significant achievements in soil property analysis, current soil prediction models continue to suffer from low accuracy. To address this issue, we propose a convolutional neural network model that can achieve high-precision soil property prediction by creating 2D multi-channel inputs and applying a multi-scale spatial attention mechanism. Initially, we explored two-dimensional multi-channel inputs for seven soil properties in the public LUCAS spectral dataset using the Gramian Angular Field (GAF) method and various preprocessing techniques. Subsequently, we developed a convolutional neural network model with a multi-scale spatial attention mechanism to improve the network's extraction of relevant spatial contextual information. Our proposed model showed superior performance in a statistical comparison with current state-of-the-art techniques. The RMSE (R²) values for various soil properties were as follows: organic carbon content (OC) of 19.083 (0.955), calcium carbonate content (CaCO3) of 24.901 (0.961), nitrogen content (N) of 0.969 (0.933), cation exchange capacity (CEC) of 6.52 (0.803), pH in H2O of 0.366 (0.927), clay content of 4.845 (0.86), and sand content of 12.069 (0.789). Our proposed model can effectively extract features from visible near-infrared spectroscopy data, contributing to the precise detection of soil properties.

Robust deep learning estimation of cortical bone porosity from MR T1-weighted images for individualized transcranial focused ultrasound planning.

Transcranial focused ultrasound (tFUS) is an emerging neuromodulation approach that has been demonstrated in animals but is difficult to translate to humans because of acoustic attenuation and scattering in the skull. Optimal dose delivery requires subject-specific skull porosity estimates which has traditionally been done using CT. We propose a deep learning (DL) estimation of skull porosity from T1-weighted MRI images which removes the need for radiation-inducing CT scans. We evaluate the impact of different DL approaches, including network architecture, input size and dimensionality, multichannel inputs, data augmentation, and loss functions. We also propose back-propagation in the mask (BIM), a method whereby only voxels inside the skull mask contribute to training. We evaluate the robustness of the best model to input image noise and MRI acquisition parameters and propagate porosity estimation errors in thousands of beam propagation scenarios. Our best performing model is a cGAN with a ResNet-9 generator with 3D 64×64×64 inputs trained with L1 and L2 losses. The model achieved a mean absolute error of 6.9% in the test set, compared to 9.5% with the pseudo-CT of Izquierdo et al. (38% improvement) and 9.4% with the generic pixel-to-pixel image translation cGAN pix2pix (36% improvement). Acoustic dose distributions in the thalamus were more accurate with our approach than with the pseudo-CT approach of both Burgos et al. and Izquierdo et al, resulting in near-optimal treatment planning and dose estimation at all frequencies compared to CT (reference). Our DL approach porosity estimates with ~7% error, is robust to input image noise and MRI acquisition parameters (sequence, coils, field strength) and yields near-optimal treatment planning and dose estimates for both central (thalamus) and lateral brain targets (amygdala) in the 200-1000 kHz frequency range.

Deep learning-based measurement of split glomerular filtration rate with 99mTc-diethylenetriamine pentaacetic acid renal scan

PurposeTo develop a deep learning (DL) model for generating automated regions of interest (ROIs) on 99mTc-diethylenetriamine pentaacetic acid (DTPA) renal scans for glomerular filtration rate (GFR) measurement.MethodsManually-drawn ROIs retrieved from a Picture Archiving and Communications System were used as ground-truth (GT) labels. A two-dimensional U-Net convolutional neural network architecture with multichannel input was trained to generate DL ROIs. The agreement between GFR values from GT and DL ROIs was evaluated using Lin’s concordance correlation coefficient (CCC) and slope coefficients for linear regression analyses. Bias and 95% limits of agreement (LOA) were assessed using Bland-Altman plots.ResultsA total of 24,364 scans (12,822 patients) were included. Excellent concordance between GT and DL GFR was found for left (CCC 0.982, 95% confidence interval [CI] 0.981–0.982; slope 1.004, 95% CI 1.003–1.004), right (CCC 0.969, 95% CI 0.968–0.969; slope 0.954, 95% CI 0.953–0.955) and both kidneys (CCC 0.978, 95% CI 0.978–0.979; slope 0.979, 95% CI 0.978–0.979). Bland-Altman analysis revealed minimal bias between GT and DL GFR, with mean differences of − 0.2 (95% LOA − 4.4–4.0), 1.4 (95% LOA − 3.5–6.3) and 1.2 (95% LOA − 6.5–8.8) mL/min/1.73 m² for left, right and both kidneys, respectively. Notably, 19,960 scans (81.9%) showed an absolute difference in GFR of less than 5 mL/min/1.73 m².ConclusionOur DL model exhibited excellent performance in the generation of ROIs on 99mTc-DTPA renal scans. This automated approach could potentially reduce manual effort and enhance the precision of GFR measurement in clinical practice.

Classification of osteoarthritic and healthy cartilage using deep learning with Raman spectra

Raman spectroscopy is a rapid method for analysing the molecular composition of biological material. However, noise contamination in the spectral data necessitates careful pre-processing prior to analysis. Here we propose an end-to-end Convolutional Neural Network to automatically learn an optimal combination of pre-processing strategies, for the classification of Raman spectra of superficial and deep layers of cartilage harvested from 45 Osteoarthritis and 19 Osteoporosis (Healthy controls) patients. Using 6-fold cross-validation, the Multi-Convolutional Neural Network achieves comparable or improved classification accuracy against the best-performing Convolutional Neural Network applied to either the raw or pre-processed spectra. We utilised Integrated Gradients to identify the contributing features (Raman signatures) in the network decision process, showing they are biologically relevant. Using these features, we compared Artificial Neural Networks, Decision Trees and Support Vector Machines for the feature selection task. Results show that training on fewer than 3 and 300 features, respectively, for the disease classification and layer assignment task provide performance comparable to the best-performing CNN-based network applied to the full dataset. Our approach, incorporating multi-channel input and Integrated Gradients, can potentially facilitate the clinical translation of Raman spectroscopy-based diagnosis without the need for laborious manual pre-processing and feature selection.

Image synthesis of interictal SPECT from MRI and PET using machine learning.

Cross-modality image estimation can be performed using generative adversarial networks (GANs). To date, SPECT image estimation from another medical imaging modality using this technique has not been considered. We evaluate the estimation of SPECT from MRI and PET, and additionally assess the necessity for cross-modality image registration for GAN training. We estimated interictal SPECT from PET and MRI as a single-channel input, and as a multi-channel input to the GAN. We collected data from 48 individuals with epilepsy and converted them to 3D isotropic images for consistence across the modalities. Training and testing data were prepared in native and template spaces. The Pix2pix framework within the GAN network was adopted. We evaluated the addition of the structural similarity index metric to the loss function in the GAN implementation. Root-mean-square error, structural similarity index, and peak signal-to-noise ratio were used to assess how well SPECT images were able to be synthesised. High quality SPECT images could be synthesised in each case. On average, the use of native space images resulted in a 5.4% percentage improvement in SSIM than the use of images registered to template space. The addition of structural similarity index metric to the GAN loss function did not result in improved synthetic SPECT images. Using PET in either the single channel or dual channel implementation led to the best results, however MRI could produce SPECT images close in quality. Synthesis of SPECT from MRI or PET can potentially reduce the number of scans needed for epilepsy patient evaluation and reduce patient exposure to radiation.

Predicting brain age using Tri-UNet and various MRI scale features

In the process of human aging, significant age-related changes occur in brain tissue. To assist individuals in assessing the degree of brain aging, screening for disease risks, and further diagnosing age-related diseases, it is crucial to develop an accurate method for predicting brain age. This paper proposes a multi-scale feature fusion method called Tri-UNet based on the U-Net network structure, as well as a brain region information fusion method based on multi-channel input networks. These methods address the issue of insufficient image feature learning in brain neuroimaging data. They can effectively utilize features at different scales of MRI and fully leverage feature information from different regions of the brain. In the end, experiments were conducted on the Cam-CAN dataset, resulting in a minimum Mean Absolute Error (MAE) of 7.46. The results demonstrate that this method provides a new approach to feature learning at different scales in brain age prediction tasks, contributing to the advancement of the field and holding significance for practical applications in the context of elderly education.

Synthetic photonic lantern characterization and power regulation with intensity superposition mode decomposition.

Photonic lanterns (PLs) enable to convert the spatial modes, making it possible to simply add/drop modes in multimode division multiplexing (MDM) systems. The parameters of PLs have a significant impact on the modal purity, signal quality, transmission distance, and efficiency. This paper proposes a method for the synthetic characterization and power regulation of PLs with multi-channel intensity superposition mode decomposition (MD). The observation reveals that the single channel on the charge-coupled device (CCD) comprises a combination of the fundamental and higher mode fields. However, when the multi-channel is input together, the intensity of each few-mode spot is superimposed on the CCD, and the parameters that cannot be acquired by a single channel can be obtained through the detection of multi-channel input. This offers a novel avenue for optimizing the PLs. Thus, the insertion loss (IL), modal dependent loss (MDL), mode conversion efficiency (MCE), crosstalk, and mode multiplexing power ratio difference (MMPRD) can be obtained. Furthermore, the power regulation is implemented based on the multi-channel intensity superposition MD. The difference between the two modes before power control at the output of the PL is 0.28 dB. However, after implementing power control measurements, this difference was reduced to 0.05 dB, facilitating power equalization throughout the entire communication system.

The effects of back-projection variants in BPF-like TOF PET reconstruction using CNN filtration - Based on simulated and clinical brain data.

The back-projection strategies such as confidence weighting (CW) and most likely annihilation position (MLAP) have been adopted into back-projection-and-filtering-like (BPF-like) deep reconstruction model and shown great potential on fast and accurate PET reconstruction. Although the two methods degenerate to an identical model at the time resolution of 0ps, they represent two distinct approaches at the realistic time resolutions of current commercial systems. There is a lack of a systematic and fair assessment on these differences. This work aims to analyze the impact of back-projection variants on CNN-based PET image reconstruction to find the most effective back-projection model, and ultimately contribute to accurate PET reconstruction. Different back-projection strategies (CW and MLAP) and different angular view processing methods (view-summed and view-grouped) were considered, leading to the comparison of four back-projection variants integrated with the same CNN filtration model. Meanwhile, we investigated two strategies of physical effect compensation, either introducing pre-corrected data as the input or adding a channel of attenuation map to the CNN model. After training models separately on Monte-Carlo-simulated BrainWeb phantoms with full dose (events=3×107), we tested them on both simulated phantoms and clinical brain scans with two dosage levels. For the performance assessment, peak signal-to-noise ratio (PSNR) and root mean square error (RMSE) were used to evaluate the pixel-wise error, structural similarity index (SSIM) to evaluate the structural similarity, and contrast recovery coefficient (CRC) in manually selected ROI to compare the region recovery. Compared to two MLAP-based histo-image reconstruction models, two CW-based back-projected image methods produced clearer, sharper, and more detailed images, from both simulated and clinical data. For angular view processing methods, view-grouped histo-image improved image quality, while view-grouped cwbp-image showed no advantage except for contrast recovery. Quantitative analysis on simulated data demonstrated that the view-summed cwbp-image model achieved the best PSNR, RMSE, SSIM, while the 8-view cwbp-image model achieved the best CRC in lesions and the white matter. Additionally, the multi-channel input model including the back-projection image and attenuation map was proved to be the most efficient and simplest method for compensating for physical effects for brain data. Applying Gaussian blur to the histo-image yielded images with limited improvement. All above results hold for both the half-dose and the full-dose cases. For brain imaging, the evaluation based on metrics PSNR, RMSE, SSIM, and CRC indicates that the view-summed CW-based back-projection variant is the most effective input for the BPF-like reconstruction model using CNN filtration, which can involve the attenuation map through an additional channel to effectively compensate for physical effects.

Automatic segmentation and classification of frontal sinuses for sex determination from CBCT scans using a two-stage anatomy-guided attention network

Sex determination is essential for identifying unidentified individuals, particularly in forensic contexts. Traditional methods for sex determination involve manual measurements of skeletal features on CBCT scans. However, these manual measurements are labor-intensive, time-consuming, and error-prone. The purpose of this study was to automatically and accurately determine sex on a CBCT scan using a two-stage anatomy-guided attention network (SDetNet). SDetNet consisted of a 2D frontal sinus segmentation network (FSNet) and a 3D anatomy-guided attention network (SDNet). FSNet segmented frontal sinus regions in the CBCT images and extracted regions of interest (ROIs) near them. Then, the ROIs were fed into SDNet to predict sex accurately. To improve sex determination performance, we proposed multi-channel inputs (MSIs) and an anatomy-guided attention module (AGAM), which encouraged SDetNet to learn differences in the anatomical context of the frontal sinus between males and females. SDetNet showed superior sex determination performance in the area under the receiver operating characteristic curve, accuracy, Brier score, and specificity compared with the other 3D CNNs. Moreover, the results of ablation studies showed a notable improvement in sex determination with the embedding of both MSI and AGAM. Consequently, SDetNet demonstrated automatic and accurate sex determination by learning the anatomical context information of the frontal sinus on CBCT scans.

A Multichannel CT and Radiomics-Guided CNN-ViT (RadCT-CNNViT) Ensemble Network for Diagnosis of Pulmonary Sarcoidosis.

Pulmonary sarcoidosis is a multisystem granulomatous interstitial lung disease (ILD) with a variable presentation and prognosis. The early accurate detection of pulmonary sarcoidosis may prevent progression to pulmonary fibrosis, a serious and potentially life-threatening form of the disease. However, the lack of a gold-standard diagnostic test and specific radiographic findings poses challenges in diagnosing pulmonary sarcoidosis. Chest computed tomography (CT) imaging is commonly used but requires expert, chest-trained radiologists to differentiate pulmonary sarcoidosis from lung malignancies, infections, and other ILDs. In this work, we develop a multichannel, CT and radiomics-guided ensemble network (RadCT-CNNViT) with visual explainability for pulmonary sarcoidosis vs. lung cancer (LCa) classification using chest CT images. We leverage CT and hand-crafted radiomics features as input channels, and a 3D convolutional neural network (CNN) and vision transformer (ViT) ensemble network for feature extraction and fusion before a classification head. The 3D CNN sub-network captures the localized spatial information of lesions, while the ViT sub-network captures long-range, global dependencies between features. Through multichannel input and feature fusion, our model achieves the highest performance with accuracy, sensitivity, specificity, precision, F1-score, and combined AUC of 0.93 ± 0.04, 0.94 ± 0.04, 0.93 ± 0.08, 0.95 ± 0.05, 0.94 ± 0.04, and 0.97, respectively, in a five-fold cross-validation study with pulmonary sarcoidosis (n = 126) and LCa (n = 93) cases. A detailed ablation study showing the impact of CNN + ViT compared to CNN or ViT alone, and CT + radiomics input, compared to CT or radiomics alone, is also presented in this work. Overall, the AI model developed in this work offers promising potential for triaging the pulmonary sarcoidosis patients for timely diagnosis and treatment from chest CT.

Deep learning neural networks with input processing for vibration-based bearing fault diagnosis under imbalanced data conditions

Deep learning (DL) networks, such as convolutional neural networks (CNNs) and long short-term memory (LSTM), have gained popularity for bearing fault diagnosis utilizing raw vibration signals. However, their accuracy and stability are compromised when facing imbalanced real-world datasets. This research investigates the impact of imbalanced datasets and explores the potential of signal processing techniques on network inputs compared to the direct use of raw vibration signals. The DL techniques studied include LSTM, one-dimensional CNN, and two-dimensional (2D) CNN, and a novel hybrid 2DCNNLSTM algorithm, incorporating signal processing methods such as Fourier transform and continuous wavelet transform while maintaining nearly equal parameters and the same base architecture. The proposed hybrid 2DCNNLSTM algorithm combines the strengths of LSTM and CNN, allowing for improved bearing diagnosis by capturing both spatial and temporal information in vibration signals. The proposed 2DCNNLSTM algorithm also considers multi-channel input augmenting raw vibration signal, mean, and variance channels to extract meaningful features and enhance classification efficiency. The publicly available Case Western Reserve University benchmark-bearing test rig dataset with ten fault classes, the Paderborn University dataset with three fault classes, and NASA Centre for Intelligent Maintenance Systems bearing datasets with five fault classes are utilized to test the proposed deep learning networks’ accuracy, effectiveness, robustness, and stability. The studies reveal that the hybrid 2DCNNLSTM-based networks outperform both CNN and LSTM networks, even without input processing. Further, utilizing multi-channel input by augmenting the 2D raw signal with mean and variance value channels proves to be more efficient in handling imbalanced and complex datasets while employing a 2DCNNLSTM-based network.