Ground Truth Research Articles

Bolus tracking from pulsed x-ray projections: A feasibility study using a five-dimensional cardiac CT contrast dynamics model.

Cardiac computed tomography (CT) exams are some of the most complex CT exams due to the need to carefully time the scan when the heart chambers are near the peak contrast concentration. With current "bolus tracking" and "timing bolus" techniques, after contrast medium is injected, a target vessel or chamber is scanned periodically, and images are reconstructed to monitor the opacification. Both techniques have opportunities for improvement, such as reducing the contrast medium volume, the exam time, the number of manual steps, and improving the robustness of correctly timing the peak opacification. The objective of our study is to (1) develop a novel autonomous cardiac CT clinical workflow to track contrast bolus dynamics directly from pulsed x-ray projections, (2) develop a new five-dimensional virtual cardiac CT data generation tool with programmable cardiac profiles and bolus dynamics, and (3) demonstrate the feasibility of projection-domain prospective bolus tracking using a neural network trained and tested with the virtual data to find the contrast peak. In our proposed workflow, pulsed mode projections (PMPs) are acquired with a wide-open collimator under sparse view conditions (monitoring phase). Each time a new PMP is acquired, the neural network is used to estimate the contrast enhancement inside the target chambers. To train such a network, we introduce a new approach to generate clinically realistic virtual scan data based on a five-dimensional cardiac model, by synthesizing user-defined contrast bolus dynamics and patient electrocardiogram profiles. In this study, we investigated a scenario with one single PMP per rotation. To find the optimal PMP view angle, 20 angles were explored. For each angle, 300 virtual exams were generated from 115 human subject datasets and divided into training, validation, and testing groups. Twenty neural networks were trained and evaluated in total to find the optimal network. Finally, a simple bolus peak time estimation algorithm was developed and evaluated by comparing to the ground truth bolus peak time. To evaluate the accuracy of a bolus time-intensity curve estimated by the network, the cosine similarity between the estimation and the ground truth was computed. The cosine similarity was larger than 0.97 for all projection angles. A view angle corresponding to the x-ray tube at 30 degrees from vertical (left-anterior of subject) showed the lowest errors. The amplitude of the estimated bolus curves (in Hounsfield Units) was not always correctly predicted, but the shape was accurately predicted. This resulted in an RMSE of 1.23s for the left chambers and 0.78s for the right chambers in the contrast peak time estimation. In this study, we proposed an innovative real-time way to predict the contrast bolus peak in cardiac CT as well as an innovative approach to train a neural network using virtual but clinically realistic data. Our trained network successfully estimated the shape of the time-intensity curve for the target chambers, which led to accurate bolus peak time estimation. This technique could be used for autonomous diagnostic cardiac CT to trigger a diagnostic scan for optimal contrast enhancement.

AnEEG: leveraging deep learning for effective artifact removal in EEG data

In neuroscience and clinical diagnostics, electroencephalography (EEG) is a crucial instrument for capturing neural activity. However, this signal is polluted by different artifacts like muscle activity, eye blinks, environmental interference, etc., which makes it more difficult to retrieve important information from the signal. Deep learning methods have demonstrated the potential to lower these artifacts and enhance the EEG’s quality in recent years. In this work, a novel deep learning method,“AnEEG” is presented for eliminating artifacts from EEG signal. The quantitative matrices NMSE, RMSE, CC, SNR and SAR are calculated to confirm the effectiveness of the proposed model. Through this process, it was found that the suggested model outperformed wavelet decomposition techniques. The model achieves lower NMSE and RMSE values, which indicates better agreement with the original signal. Achieving higher CC values means stronger linear agreement with the ground truth signals. Additionally, the model shows improvements in both SNR and SAR values. Overall, this suggested approach showcases promising results in improving the quality of EEG data by utilizing deep learning.

Anatomic Interpretability in Neuroimage Deep Learning: Saliency Approaches for Typical Aging and Traumatic Brain Injury.

The black box nature of deep neural networks (DNNs) makes researchers and clinicians hesitant to rely on their findings. Saliency maps can enhance DNN explainability by suggesting the anatomic localization of relevant brain features. This study compares seven popular attribution-based saliency approaches to assign neuroanatomic interpretability to DNNs that estimate biological brain age (BA) from magnetic resonance imaging (MRI). Cognitively normal (CN) adults ( N = 13,394, 5,900 males; mean age: 65.82 ± 8.89 years) are included for DNN training, testing, validation, and saliency map generation to estimate BA. To study saliency robustness to the presence of anatomic deviations from normality, saliency maps are also generated for adults with mild traumatic brain injury (mTBI, \(\:N\) = 214, 135 males; mean age: 55.3 ± 9.9 years). We assess saliency methods' capacities to capture known anatomic features of brain aging and compare them to a surrogate ground truth whose anatomic saliency is known a priori . Anatomic aging features are identified most reliably by the integrated gradients method, which outperforms all others through its ability to localize relevant anatomic features. Gradient Shapley additive explanations, input × gradient, and masked gradient perform less consistently but still highlight ubiquitous neuroanatomic features of aging (ventricle dilation, hippocampal atrophy, sulcal widening). Saliency methods involving gradient saliency, guided backpropagation, and guided gradient-weight class attribution mapping localize saliency outside the brain, which is undesirable. Our research suggests the relative tradeoffs of saliency methods to interpret DNN findings during BA estimation in typical aging and after mTBI.

Feasibility of using two generative AI models for teeth reconstruction

ObjectivesThis feasibility study investigates the application of artificial intelligence (AI) models, specifically transformer-based (TM) and diffusion-based (DM) models, for the reconstruction of single and multiple missing teeth. MethodsA dataset of 129 digitized models was utilized to create both TM and DM models. Single and multiple missing teeth were artificially generated. Reconstruction accuracy was assessed against ground truth data using Root Mean Square (RMS) and mean absolute error (MAE) across various artificially generated teeth. Paired t-tests were used for analyzing differences between the two models (p < 0.05). ResultsBoth TM and DM models demonstrated similar accuracy in the reconstruction of single and multiple missing teeth. The greatest disparity occurred in the reconstruction of all remaining teeth, with the exception of 33 and 43 for both models (RMS TM: 0.37; DM: 0.43). TM exhibited the highest precision in reconstructing tooth 34 (RMS: 0.21), whereas DM demonstrated superior accuracy in reconstructing tooth 21 (RMS: 0.19). Despite there was no significant difference between the models. ConclusionsAI-based TM and DM models demonstrate promising results in reconstructing missing teeth, with superior accuracy in single-tooth compared to multiple-tooth edentulous spaces. Despite the need for additional refining and larger datasets, including antagonistic teeth, these models have the potential to streamline and improve the dental restoration processes, potentially leading to cost savings and enhanced clinical outcomes. SignificanceThis study demonstrates the feasibility and potential of transformer- and diffusion-based AI models to accurately reconstruct missing teeth, offering a novel approach that could streamline and enhance the precision of implant planning.

Spatial Downscaling of Sea Surface Temperature Using Diffusion Model

In recent years, advancements in high-resolution digital twin platforms or artificial intelligence marine forecasting have led to the increased requirements of high-resolution oceanic data. However, existing sea surface temperature (SST) products from observations often fail to meet researchers’ resolution requirements. Deep learning models serve as practical techniques for improving the spatial resolution of SST data. In particular, diffusion models (DMs) have attracted widespread attention due to their ability to generate more vivid and realistic results than other neural networks. Despite DMs’ potential, their application in SST spatial downscaling remains largely unexplored. Hence we propose a novel DM-based spatial downscaling model, called DIFFDS, designed to obtain a high-resolution version of the input SST and to restore most of the meso scale processes. Experimental results indicate that DIFFDS is more effective and accurate than baseline neural networks, its downscaled high-resolution SST data are also visually comparable to the ground truth. The DIFFDS achieves an average root-mean-square error of 0.1074 °C and a peak signal-to-noise ratio of 50.48 dB in the 4× scale downscaling task, which shows its accuracy.

A Machine Learning Framework for Assessing Experts’ Decision Quality

Expert workers make non-trivial decisions with significant implications. Experts’ decision accuracy is, thus, a fundamental aspect of their judgment quality, key to both management and consumers of experts’ services. Yet, in many important settings, transparency in experts’ decision quality is rarely possible because ground truth data for evaluating the experts’ decisions is costly and available only for a limited set of decisions. Furthermore, different experts typically handle exclusive sets of decisions, and thus, prior solutions that rely on the aggregation of multiple experts’ decisions for the same instance are inapplicable. We first formulate the problem of estimating experts’ decision accuracy in this setting and then develop a machine–learning–based framework to address it. Our method effectively leverages both abundant historical data on workers’ past decisions and scarce decision instances with ground truth labels. Using both semi-synthetic data based on publicly available data sets and purposefully compiled data sets on real workers’ decisions, we conduct extensive empirical evaluations of our method’s performance relative to alternatives. The results show that our approach is superior to existing alternatives across diverse settings, including settings that involve different data domains, experts’ qualities, and amounts of ground truth data. To our knowledge, this paper is the first to posit and address the problem of estimating experts’ decision accuracies from historical data with scarce ground truth, and it is the first to offer comprehensive results for this problem setting, establishing the performances that can be achieved across settings as well as the state-of-the-art performance on which future work can build. This paper was accepted by Anindya Ghose, information systems. Funding: T. Geva acknowledges research grants from the Jeremy Coller Foundation and from the Henry Crown Institute for Business Research. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2021.03357 .

Assessment of the Diagnostic Performance of a Commercially Available Artificial Intelligence Algorithm for Risk Stratification of Thyroid Nodules on Ultrasound.

Background: Thyroid nodules are challenging to accurately characterize on ultrasound (US), though the emergence of risk stratification systems and more recently artificial intelligence (AI) algorithms has improved nodule classification. The purpose of this study was to evaluate the performance of a recent Food and Drug Administration (FDA)-cleared AI tool for detection of malignancy in thyroid nodules on US. Methods: One year of consecutive thyroid US with ≥1 nodule from Duke University Hospital and its affiliate community hospital (649 nodules from 347 patients) were retrospectively evaluated. Included nodules had ground truth diagnoses by surgical pathology, fine needle aspiration (FNA), or three-year follow-up US showing stability. An FDA-cleared AI tool (Koios DS Thyroid) analyzed each nodule to generate (i) American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS) descriptors, scores, and follow-up recommendations and (ii) an AI-adapter score to further adjust risk assessments and recommendations. Four groups were then compared: (i) Koios with AI-adapter, (ii) Koios without AI-adapter, (iii) clinical radiology report, and (iv) radiology report combined with AI-adapter. Performance of the final recommendations (FNA or no FNA) was determined based on ground truth, and comparison between the four groups was made using sensitivity, specificity, and receiver-operating-curve analysis. Results: Of 649 nodules, 32 were malignant and 617 were benign. Performance of Koios with AI-adapter enabled was similar to radiologists (area under the curve [AUC] 0.70 for both, [CI 0.60-0.81] and [0.60-0.79], respectively). Koios with AI-adapter had improved specificity compared to radiologists (0.63 [CI: 0.59-0.67] versus 0.43 [CI: 0.38-0.48]) but decreased sensitivity (0.69 [CI: 0.50-0.83) versus 0.81 [CI: 0.61, 0.92]). Highest performance was seen when the radiology interpretation was combined with the AI-adapter (AUC 0.76, [CI: 0.67-0.85]). Combined with the AI-adapter, radiologist specificity improved from 0.43 ([CI: 0.38-0.48]) to 0.53 ([CI: 0.49-0.58]) (McNemar's test p < 0.001), resulting in 17% fewer FNA recommendations, with unchanged sensitivity (0.81, p = 1). Conclusion: Koios DS demonstrated standalone performance similar to radiologists, though with lower sensitivity and higher specificity. Performance was best when radiologist interpretations were combined with the AI-adapter component, with improved specificity and reduced unnecessary FNA recommendations.

An Adaptive SCG-ECG Multimodal Gating Framework for Cardiac CTA.

Cardiovascular disease (CVD) is the leading cause of death worldwide. Coronary artery disease (CAD), a prevalent form of CVD, is typically assessed using catheter coronary angiography (CCA), an invasive, costly procedure with associated risks. While cardiac computed tomography angiography (CTA) presents a less invasive alternative, it suffers from limited temporal resolution, often resulting in motion artifacts that degrade diagnostic quality. Traditional ECG-based gating methods for CTA inadequately capture cardiac mechanical motion. To address this, we propose a novel multimodal approach that enhances CTA imaging by predicting cardiac quiescent periods using seismocardiogram (SCG) and ECG data, integrated through a weighted fusion (WF) approach and artificial neural networks (ANNs). We developed a regression-based ANN framework (r-ANN WF) designed to improve prediction accuracy and reduce computational complexity, which was compared with a classification-based framework (c-ANN WF), ECG gating, and US data. Our results demonstrate that the r-ANN WF approach improved overall diastolic and systolic cardiac quiescence prediction accuracy by 52.6% compared to ECG-based predictions, using ultrasound (US) as the ground truth, with an average prediction time of 4.83 ms. Comparative evaluations based on reconstructed CTA images show that both r-ANN WF and c-ANN WF offer diagnostic quality comparable to US-based gating, underscoring their clinical potential. Additionally, the lower computational complexity of r-ANN WF makes it suitable for real-time applications. This approach could enhance CTA's diagnostic quality, offering a more accurate and efficient method for CVD diagnosis and management.

Deep Learning-Enhanced Paper-Based Vertical Flow Assay for High-Sensitivity Troponin Detection Using Nanoparticle Amplification.

Successful integration of point-of-care testing (POCT) into clinical settings requires improved assay sensitivity and precision to match laboratory standards. Here, we show how innovations in amplified biosensing, imaging, and data processing, coupled with deep learning, can help improve POCT. To demonstrate the performance of our approach, we present a rapid and cost-effective paper-based high-sensitivity vertical flow assay (hs-VFA) for quantitative measurement of cardiac troponin I (cTnI), a biomarker widely used for measuring acute cardiac damage and assessing cardiovascular risk. The hs-VFA includes a colorimetric paper-based sensor, a portable reader with time-lapse imaging, and computational algorithms for digital assay validation and outlier detection. Operating at the level of a rapid at-home test, the hs-VFA enabled the accurate quantification of cTnI using 50 μL of serum within 15 min per test and achieved a detection limit of 0.2 pg/mL, enabled by gold ion amplification chemistry and time-lapse imaging. It also achieved high precision with a coefficient of variation of <7% and a very large dynamic range, covering cTnI concentrations over 6 orders of magnitude, up to 100 ng/mL, satisfying clinical requirements. In blinded testing, this computational hs-VFA platform accurately quantified cTnI levels in patient samples and showed a strong correlation with the ground truth values obtained by a benchtop clinical analyzer. This nanoparticle amplification-based computational hs-VFA platform can democratize access to high-sensitivity point-of-care diagnostics and provide a cost-effective alternative to laboratory-based biomarker testing.

Using UAV RGB Images for Assessing Tree Species Diversity in Elevation Gradient of Zao Mountains

Vegetation biodiversity in mountainous regions is controlled by altitudinal gradients and their corresponding microclimate. Higher temperatures, shorter snow cover periods, and high variability in the precipitation regime might lead to changes in vegetation distribution in mountains all over the world. In this study, we evaluate vegetation distribution along an altitudinal gradient (1334–1667 m.a.s.l.) in the Zao Mountains, northeastern Japan, by means of alpha diversity indices, including species richness, the Shannon index, and the Simpson index. In order to assess vegetation species and their characteristics along the mountain slope selected, fourteen 50 m × 50 m plots were selected at different altitudes and scanned with RGB cameras attached to Unmanned Aerial Vehicles (UAVs). Image analysis revealed the presence of 12 dominant tree and shrub species of which the number of individuals and heights were validated with fieldwork ground truth data. The results showed a significant variability in species richness along the altitudinal gradient. Species richness ranged from 7 to 11 out of a total of 12 species. Notably, species such as Fagus crenata, despite their low individual numbers, dominated the canopy area. In contrast, shrub species like Quercus crispula and Acer tschonoskii had high individual numbers but covered smaller canopy areas. Tree height correlated well with canopy areas, both representing tree size, which has a strong relationship with species diversity indices. Species such as F. crenata, Q. crispula, Cornus controversa, and others have an established range of altitudinal distribution. At high altitudes (1524–1653 m), the average shrubs’ height is less than 4 m, and the presence of Abies mariesii is negligible because of high mortality rates caused by a severe bark beetle attack. These results highlight the complex interactions between species abundance, canopy area, and altitude, providing valuable insights into vegetation distribution in mountainous regions. However, species diversity indices vary slightly and show some unusually low values without a clear pattern. Overall, these indices are higher at lower altitudes, peak at mid-elevations, and decrease at higher elevations in the study area. Vegetation diversity indices did not show a clear downward trend with altitude but depicted a vegetation composition at different altitudes as controlled by their surrounding environment. Finally, UAVs showed their significant potential for conducting large-scale vegetation surveys reliably and in a short time, with low costs and low manpower.

Human Fall Motion Prediction: Fall Motion Forecasting and Detection with GRU

The human fall motion prediction system is a preventive tool aimed at reducing the risk of falls. In our research, we developed a deep learning model that utilizes pose estimation to track human body posture and integrated this with a Gated Recurrent Unit (GRU) to forecast human motion and predict falls. GRU, an enhancement of Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) models offers improved memorization and more efficient memory usage and performance. Our study presents the human fall motion prediction, which combines the forecasting and classification of potential falls.The CAUCAFall dataset is used as the benchmark of this study, which contains the image sequences of single human motion with ten actions conducted by ten actors. We employed the YOLOv8 Pose model to track the 2D human body pose as the input in our system. A thorough evaluation of the CAUCAFall dataset highlights the effectiveness of our proposed system. Evaluation using the CAUCAFall dataset demonstrates that the model achieved a Mean Per Joint Position Error (MPJPE) of 4.65 pixels from the ground truth, with a 70% accuracy rate in fall prediction. However, the model also exhibited a Mean Relative Error (MRE) of 0.3, indicating that 30% of the predictions were incorrect. These findings underscore the potential of the GRU-based system in fall prevention

External Validation of a Commercial Artificial Intelligence Algorithm on a Diverse Population for Detection of False Negative Breast Cancers.

There are limited data on the application of artificial intelligence (AI) on nonenriched, real-world screening mammograms. This work aims to evaluate the ability of AI to detect false negative cancers not detected at the time of screening when reviewed by the radiologist alone. A commercially available AI algorithm was retrospectively applied to patients undergoing screening full-field digital mammography (FFDM) or digital breast tomosynthesis (DBT) at a single institution from 2010 to 2019. Ground truth was established based on 1-year follow-up data. Descriptive statistics were performed with attention focused on AI detection of false negative cancers within these subsets. A total of 26 694 FFDM and 3183 DBT examinations were analyzed. Artificial intelligence was able to detect 7/13 false negative cancers (54%) in the FFDM cohort and 4/10 (40%) in the DBT cohort on the preceding screening mammogram that was interpreted as negative by the radiologist. Of these, 4 in the FFDM cohort and 4 in the DBT cohort were identified in breast densities of C or greater. False negative cancers detected by AI were predominantly luminal A invasive malignancies (9/11, 82%). Artificial intelligence was able to detect these false negative cancers a median time of 272 days sooner in the FFDM cohort and 248 days sooner in the DBT cohort compared to the radiologist. Artificial intelligence was able to detect cancers at the time of screening that were missed by the radiologist. Prospective studies are needed to evaluate the synergy of AI and the radiologist in real-world settings, especially on DBT examinations.

Estimating Contribution Quality in Online Deliberations Using a Large Language Model

Deliberation involves participants exchanging knowledge, arguments, and perspectives and has been shown to be effective at addressing polarization. The Stanford Online Deliberation Platform facilitates large-scale deliberations. It enables video-based online discussions on a structured agenda for small groups without requiring human moderators. This paper's data comes from various deliberation events, including one conducted in collaboration with Meta in 32 countries, and another with 38 post-secondary institutions in the US. Estimating the quality of contributions in a conversation is crucial for assessing feature and intervention impacts. Traditionally, this is done by human annotators, which is time-consuming and costly. We use a large language model (LLM) alongside eight human annotators to rate contributions based on justification, novelty, expansion of the conversation, and potential for further expansion, with scores ranging from 1 to 5. Annotators also provide brief justifications for their ratings. Using the average rating from other human annotators as the ground truth, we find the model outperforms individual human annotators. While pairs of human annotators outperform the model in rating justification and groups of three outperform it on all four metrics, the model remains competitive. We illustrate the usefulness of the automated quality rating by assessing the effect of nudges on the quality of deliberation. We first observe that individual nudges after prolonged inactivity are highly effective, increasing the likelihood of the individual requesting to speak in the next 30 seconds by 65%. Using our automated quality estimation, we show that the quality ratings for statements prompted by nudging are similar to those made without nudging, signifying that nudging leads to more ideas being generated in the conversation without losing overall quality.

Sinogram-characteristic-informed network for efficient restoration of low-dose SPECT projection data.

Single Photon Emission Computed Tomography (SPECT) sinogram restoration for low-dose imaging is a critical challenge in medical imaging. Existing methods often overlook the characteristics of the sinograms, necessitating innovativeapproaches. In this study, we introduce the Sinogram-characteristic-informed network (SCI-Net) to address the restoration of low-dose SPECT sinograms. Our aim is to build and train a model based on the characteristics of sinograms, including continuity, periodicity, multi-scale properties of lines in sinograms, and others, to enhance the model's understanding of the restorationprocess. SCI-Net incorporates several novel mechanisms tailored to exploit the inherent characteristics of sinograms. We implement a channel attention module with a decay mechanism to leverage continuity across adjacent sinograms, while a position attention module captures global correlations within individual sinograms. Additionally, we propose a multi-stage progressive integration mechanism to balance local detail and overall structure. Multiple regularization terms, customized to sinogram image characteristics, are embedded into the loss function for modeltraining. The experimental evaluations are divided into two parts: simulation data evaluation and clinical evaluation. The simulation data evaluation is conducted on a dataset comprising ten organ types, generated by the SIMIND Monte Carlo program from extended cardiac-torso (XCAT) anatomical phantoms. The dataset includes a total of SPECT sinograms with low-dose as input data and normal-dose as ground truth, consisting of 3881 sinograms in the training dataset and 849 sinograms in the testing set. When comparing the restoration of low-dose sinograms to normal-dose references, SCI-Net effectively improves performance. Specifically, the peak signal-to-noise ratio (PSNR) and the structural similarity (SSIM) on sinograms increase from 15.72 to 34.66 ( 0.001) and 0.6297 to 0.9834 ( 0.001), respectively, and on reconstructed images, reconstructed by maximum likelihood-expectation maximization (ML-EM), the PSNR and the SSIM improve from 21.95 to 33.14 ( 0.001) and 0.9084 to 0.9866 ( 0.001), respectively. We compared SCI-Net with existing methods, including baseline models, traditional reconstruction algorithms, end-to-end methods, sinogram restoration methods, and image post-processing methods. The experimental results and visual examples demonstrate that SCI-Net surpasses these existing methods in SPECT sinogram restoration. The clinical evaluation is conducted on clinical data of low-dose SPECT sinograms for spleen, thyroid, skull, and bone. These SPECT projection data are obtained from Discovery NM/CT670 scans. We compare the reconstructed images from the SCI-Net restored sinograms, the reconstructed images from the original low-dose sinograms, and the reconstructed images using the built-in algorithm of the Discovery NM/CT670. The results show that our method effectively reduces the coefficient of variation (COV) in the regions of interest (ROI) of the reconstructed images, thereby enhancing the quality of the reconstructed images through SPECT sinogramrestoration. Our proposed SCI-Net exhibits promising performance in the restoration of low-dose SPECT projection data. In the SCI-Net, we have implemented three mechanisms based on distinct forms, which are advantageous for the model to more effectively leverage the characteristics of sinograms and achieve commendable restorationoutcomes.

Development of an accelerometer-based wearable sensor approach for alcohol consumption detection.

Alcohol is a commonly used substance associated with significant public health consequences. Treatment is often stigmatized and limited with regard to both access and affordability, demonstrating the need for innovations in alcohol treatment. Accelerometer sensors can detect drinking without user input and are widely incorporated into wearable devices, increasing accessibility and affordability. We compared a distributional and random forest classification approach to detect and evaluate sensor-based drinking data. Data were collected at a local state fair (n = 194), where participants drank water at specified intervals interspersed with confounding behaviors (e.g., touching nose, rubbing forehead, or yawning) while wearing an Android-based smartwatch for 10 min. Participants were randomized to receive one of three drinking container shapes: pint, martini, or wine. The random forest model achieved an overall testing accuracy of 93% (sensitivity = 0.32; specificity = 0.99; positive predictive value = 0.74). The distributional algorithm achieved an overall accuracy of 95% (sensitivity = 0.76; specificity = 0.97; positive predictive value = 0.72). The distributional algorithm had a significantly greater accuracy (t(193) = 7.73, p < 0.001, d = 0.56) and sensitivity (t(193) = 24.5, p < 0.001, d = 1.76). Equivalency testing demonstrated significant equivalency to the ground truth for sip duration (tlower(193) = 16.92, p < 0.001; tupper(193) = -9.85, p < 0.001) and between-sip interval (tlower(193) = 1.72, p = 0.044; thigher(193) = -3.96, p < 0.001). However, the random forest did not have significant equivalency to the ground truth for between-sip interval (tlower(193) = 1.98, p = 0.025; thigher(193) = 0.160, p = 0.564). Overall, the results indicated that consumer-grade smartwatches can be utilized to detect and measure alcohol use behavior using machine learning and distributional algorithms. This work provides the methodological foundation for future research to analyze the behavioral pharmacology of alcohol use and develop accessible just-in-time clinical interventions.

A Crowded Object Counting System with Self-Attention Mechanism

Traditional object counting systems use object detection methods to count objects. However, when objects are small, crowded, and dense, object detection may fail, leading to inaccuracies in counting. To address this issue, we propose a crowded object counting system based on density map estimation. While most density map estimation models employ encoder–decoder or multi-branch approaches to generate feature maps at different scales for obtaining an accurate density map, improving the accuracy of crowded object counting remains a challenge. In this paper, we propose a novel model that can generate more accurate density maps, utilizing the context-aware network as the primary structure and integrating the self-attention mechanism. There are three main contributions in this paper. Firstly, the self-attention mechanism is employed to improve the accuracy of density map estimation. Secondly, the missing vehicle labels in the TRANCOS database are relabeled, ensuring that the ground truth data are more complete than the original TRANCOS database, thus enabling the proposed novel model to have higher crowded object counting accuracy. Thirdly, the parameters of the self-attention mechanism are analyzed to obtain the optimum parameter combination. The experimental results demonstrate that the accuracy of crowded object counting can reach 85.9%, 90.0%, 83.4%, and 92.6% for the TRANCOS, relabeled TRANCOS, ShanghaiTech Part A, and Part B datasets, respectively. Furthermore, the ablation study for the context-aware network with self-attention mechanism analyzes the optimum parameter combination.

High-Fidelity Sensing Modality for Anomaly Detection in Inkjet Printing

Abstract Inkjet three-dimensional (3D) printing has emerged as a transformative manufacturing technique, finding applications in diverse fields such as biomedical, metal fabrication, and functional materials production. It involves precise deposition of materials onto a moving substrate through a nozzle, achieving submillimeter scale resolution. However, the dynamic nature of droplet deposition introduces uncertainties, challenging consistent quality control. Current process monitoring, relying on image-based techniques, is slow and limited, hindering real-time feedback in erratic droplet ejection. In response to these challenges, we present the zero-dimensional ultrafast sensing (0-DUS) system, a novel, cost-effective, in situ monitoring tool designed to assess the quality of drop-on-demand inkjet printing. The 0-DUS system leverages the sensitivity of the light-beam field interference effect to rapidly and precisely detect and analyze localized droplets. Two core technical advancements drive this innovation: first, the exploration of integral sensing of the computational light-beam field, which allows for efficient extraction of temporal and spatial information about droplet evolution, introducing a novel in situ sensing modality; second, the establishment of a robust mapping mechanism that aligns sensor data with image-based data, facilitating accurate estimation of droplet characteristics. We successfully implemented the 0-DUS system within a commercial inkjet printer and conducted a comparative analysis with ground truth data. Our experimental results demonstrate a detection accuracy exceeding 95%, even at elevated speeds, allowing for an impressive in situ certification throughput of up to 500 Hz. Consequently, our proposed 0-DUS system meets the stringent quality assurance requirements, thereby expanding the potential applications of inkjet printing across a wide spectrum of industrial sectors.

VP-net: an end-to-end deep learning network for elastic wave velocity prediction in human skin in vivo using optical coherence elastography.

Acne vulgaris, one of the most common skin conditions, affects up to 85% of late adolescents, currently no universally accepted assessment system. The biomechanical properties of skin provide valuable information for the assessment and management of skin conditions. Wave-based optical coherence elastography (OCE) quantitatively assesses these properties of tissues by analyzing induced elastic wave velocities. However, velocity estimation methods require significant expertise and lengthy image processing times, limiting the clinical translation of OCE technology. Recent advances in machine learning offer promising solutions to simplify velocity estimation process. In this study, we proposed a novel end-to-end deep-learning model, named velocity prediction network (VP-Net), aiming to accurately predict elastic wave velocity from raw OCE data of in vivo healthy and abnormal human skin. A total of 16,424 raw phase slices from 1% to 5% agar-based tissue-mimicking phantoms, 28,270 slices from in vivo human skin sites including the palm, forearm, back of the hand from 16 participants, and 580 slices of facial closed comedones were acquired to train, validate, and test VP-Net. VP-Net demonstrated highly accurate velocity prediction performance compared to other deep-learning-based methods, as evidenced by small evaluation metrics. Furthermore, VP-Net exhibited low model complexity and parameter requirements, enabling end-to-end velocity prediction from a single raw phase slice in 1.32ms, enhancing processing speed by a factor of ∼100 compared to a conventional wave velocity estimation method. Additionally, we employed gradient-weighted class activation maps to showcase VP-Net's proficiency in discerning wave propagation patterns from raw phase slices. VP-Net predicted wave velocities that were consistent with the ground truth velocities in agar phantom, two age groups (20s and 30s) of multiple human skin sites and closed comedones datasets. This study indicates that VP-Net could rapidly and accurately predict elastic wave velocities related to biomechanical properties of in vivo healthy and abnormal skin, offering potential clinical applications in characterizing skin aging, as well as assessing and managing the treatment of acne vulgaris.

An alloy-agnostic machine learning framework for process mapping in laser powder bed fusion

PurposeThis study aims to introduce an image-based method to determine the processing window for a given alloy system using laser powder bed fusion equipment based on achieving the desired melting mode across multiple materials for powder-free specimens. The method uses a convolutional neural network trained to classify different track morphologies across different alloy systems to select appropriate printing settings. This method is intended for the development of new alloy systems, where the powder feedstock may be unavailable, or prohibitively expensive to manufacture.Design/methodology/approachA convolutional neural network is designed from scratch to identify the 4 key melting modes that are observed in laser powder bed fusion additive manufacturing across different alloy systems. To increase the prediction accuracy and generalisation accuracy across different materials, the network is trained using a novel hybrid data set that combines fully unsupervised learning with semi-supervised learning.FindingsThis study demonstrates that our convolutional network with a novel hybrid training approach can be generalised across different materials, and k-fold validation shows that the model retains good accuracy with changing training conditions. The model can predict the processing maps for the different alloys with an accuracy of up to 96% in some cases. It is also shown that powder-free single-track experiments are a useful indicator for predicting the final print quality of a component.Originality/valueThe “invariant information clustering” (IIC) approach is applied to process optimisation for additive manufacturing, and a novel hybrid data set construction approach that accounts for uncertainty in the ground truth data, enables the trained convolutional model to perform across a range of different materials and most importantly, generalise to materials outside of the training data set. Compared to the traditional cross-sectioning approach, this method considers the whole length of the single track when determining the melting mode.

Integrating color histogram analysis and convolutional neural networks for skin lesion classification

The color of skin lesions is a crucial diagnostic feature for identifying malignant melanoma and other skin diseases. Typical colors associated with melanocytic lesions include tan, brown, black, red, white, and blue-gray. This study introduces a novel feature: the number of colors present in lesions, which can indicate the severity of skin diseases and help distinguish melanomas from benign lesions. We propose a color histogram analysis, a traditional image processing technique, to analyze the pixels of skin lesions from three publicly available datasets: PH2, ISIC2016, and Med-Node, which include dermoscopic and non-dermoscopic images. While the PH2 dataset contains ground truth about skin lesion colors, the ISIC2016 and Med-Node datasets lack such annotations; our algorithm establishes this ground truth using the color histogram analysis based on the PH2 dataset. We then design and train a 19-layer Convolutional Neural Network (CNN) with different skip connections of residual blocks to classify lesions into three categories based on the number of colors present. The DeepDream algorithm is utilized to visualize the learned features of different layers, and multiple configurations of the proposed CNN are tested, achieving the highest weighted F1-score of 75.00 % on the test set. LIME is subsequently applied to identify the most important features influencing the model's decision-making. The findings demonstrate that the number of colors in lesions is a significant feature for describing skin conditions. The proposed CNN, particularly with three skip connections, shows strong potential for clinical application in diagnosing melanoma, supporting its use alongside traditional diagnostic methods.