Processing Algorithms Research Articles

Use of ChatGPT Large Language Models to Extract Details of Recommendations for Additional Imaging From Free-Text Impressions of Radiology Reports.

Background: Automated extraction of actionable details of recommendations for additional imaging (RAIs) from radiology reports could facilitate tracking and timely completion of clinically necessary RAIs and thereby potentially reduce diagnostic delays. Purpose: To assess the performance of large-language models (LLMs) in extracting actionable details of RAIs from radiology reports. Methods: This retrospective single-center study evaluated reports of diagnostic radiology examinations performed across modalities and care settings within five subspecialties (abdominal imaging, musculoskeletal imaging, neuroradiology, nuclear medicine, thoracic imaging) in August 2023. Of reports identified by a previously validated natural-language processing algorithm to contain an RAI, 250 were randomly selected; 231 of these were confirmed to contain an RAI on manual review and formed the study sample. Twenty-five reports were used to engineer a prompt instructing an LLM, when inputted a report impression containing an RAI, to extract details about the modality, body part, timeframe, and rationale of the RAI; the remaining 206 reports were used for testing the prompt in combination with GPT-3.5 and GPT-4. A 4th-year medical student and radiologist from the relevant subspecialty independently classified the LLM outputs as correct versus incorrect for extracting the four actionable details of RAIs in comparison with the report impressions; a third reviewer assisted in resolving discrepancies. Extraction accuracy was summarized and compared between LLMs. Results: For GPT-3.5 and GPT-4, the two reviewers agreed for classification of LLM output as correct versus incorrect with respect to report impressions for 95.6% and 94.2% for RAI modality, 89.3% and 88.3% for RAI body part, 96.1% and 95.1% for RAI timeframe, and 89.8% and 88.8% for RAI rationale, respectively. Using consensus assessments, GPT-4 was more accurate than GPT-3.5 in extracting RAI modality (94.2% [194/206] vs 85.4% [176/206], p<.001), RAI body part (86.9% [179/206] vs 77.2% [159/206], p=.004), and RAI timeframe (99.0% [204/206] vs 95.6% [197/206], p=.02). Both LLMs had accuracy of 91.7% (189/206) for extracting RAI rationale. Conclusion: LLMs were used to extract actionable details of RAIs from free-text impression sections of radiology reports; GPT-4 outperformed GPT-3.5. Clinical Impact: The technique could represent an innovative method to facilitate timely completion of clinically necessary radiologist recommendations.

Peri-operative Intrastromal Corneal Segmentation after creation of corneal cuts based on Laser Induced Optical Breakdown: A Perspective Study

Significance Assessing refractive correction techniques, such as Lenticule Extraction or LASIK proves instrumental in diagnostics and monitor treatment efficacy. Moreover, these evaluations aid in elucidating potential post-operative complication stemming from imperfect residuals. However, strong noise distribution and low contrast between adjacent faults complicate accurate image segmentation and characteristic inference. This study proposes an image processing algorithm designed to automatically characterize volumes for lenticule extraction or surfaces for flap lift on corneal refractive surgery. Aim The proposed approach fills a current gap in precise reassessment of intrastromal laser cuts where manual expert measurements were seen as the conventional (user-based) standard. Approach The proposed algorithm employs a non-local denoising, the Sobel gradient approach, and Bayesian optimization to delineate and characterize cuts determining their macroscopic parameters. Our analysis involved testing 60 ex-vivo porcine eyes subjected either to flap or lenticule cuts using the SCHWIND ATOS femtosecond laser system. In-situ image analysis was conducted and compared with the intended geometries. Results 60 porcine eyes treated either through flap or lenticule were characterized with our proposed approach exceeding conventional manual measurements. The reproducibility and reliability of the algorithm enabled inferential analysis which might not be achieved with doubted user-base marking. Conclusions Our effort was concentrated on demonstrating the potential of the proposed work, particularly its ability to achieve comprehensive intrastromal characterization—a capability, to the best of our knowledge, being showcased for the first time.

TLSynth: A Novel Blender Add-On for Real-Time Point Cloud Generation from 3D Models

Point clouds are a crucial element in the process of scanning and reconstructing 3D environments, such as buildings or heritage sites. They allow for the creation of 3D models that can be used in a wide range of applications. In some cases, however, only the 3D model of an environment is available, and it is necessary to obtain point clouds with the same characteristics as those captured by a laser scanner. For instance, point clouds may be required for surveys, performance optimization, site scan planning, or validation of point cloud processing algorithms. This paper presents a new terrestrial laser scanner (TLS) simulator, designed as a Blender add-on, that produces synthetic point clouds from 3D models in real time. The simulator allows users to adjust a set of parameters to replicate real-world scanning conditions, such as noise generation, ensuring the synthetic point clouds closely mirror those produced by actual laser scanners. The target meshes may be derived from either a real-world scan or 3D designs created using design software. By replicating the spatial distributions and attributes of real laser scanner outputs and supporting real-time generation, the simulator serves as a valuable tool for scan planning and the development of synthetic point cloud repositories, advancing research and practical applications in 3D computer vision.

Framework for Analyzing Spatial Interference in Vehicle-to-Vehicle Communication Networks with Positional Errors

This paper introduces a novel framework for evaluating vehicle-to-vehicle (V2V) communication systems, employing beamforming and null steering techniques. Addressing challenges such as electromagnetic interference from other vehicles within the network as well as diverse road conditions, the framework defines the simulation environment for V2V networks across different traffic scenarios to assess system reliability. The analytical components of the proposed framework are structured as follows: a comprehensive framework is developed, serving as the basis for implementing a simulator that leverages advanced spatial signal processing algorithms to evaluate V2V networks across specific scenarios, accounting for the effects of positional errors. The framework integrates multiple blocks, including road modeling, system modeling, and performance evaluation, providing adaptability for different algorithms or configurations. Positional inaccuracies are examined, highlighting their effects on system performance, particularly in scenarios where null steering accuracy is imperfect, thus underscoring the need for enhanced interference management strategies.

Research on Interactive Teaching of College English Based on Deep Learning and Virtual Reality

As digital technologies advance, the need for interactive teaching methodologies in higher education has become crucial, particularly in the context of English language learning. This study explores innovative approaches utilizing deep learning and virtual reality to enhance student engagement and learning outcomes in college environments. The primary aim is to investigate the effectiveness of an Intelligent Coral Reef Optimization-driven Redefined Long Short-Term Memory (ICR-RLSTM) model, designed to facilitate a more dynamic and personalized learning experience in college English courses. The dataset comprises diverse English language materials, including texts, audio and interactive exercises, sourced from various educational platforms. Data pre-processing involves standardization techniques and the application of Natural Language Processing (NLP) algorithms to ensure the information is clean and structured for optimal analysis. The proposed interactive teaching model integrates gamification elements and virtual environments to provide students with immersive learning experiences. The ICR-RLSTM algorithm enhances predictive capabilities by optimizing long-term memory retention through intelligent coral reef optimization techniques, adapting to individual learner progress. Preliminary results indicate that the proposed model significantly improves student engagement and knowledge retention compared to traditional methods. The findings suggest that integrating deep learning and virtual reality into English language teaching can create adaptive and effective learning environments. This research contributes valuable insights into enhancing interactive teaching strategies and informs future educational practices.

Application of Multidimensional Data Analysis in Network Intrusion Detection System

It is important to understand the behavior of an information network and its features. In this research, we explore this idea by applying a multidimensional data analysis system, a system that is significant in enhancing data intrusion detection. To accomplish this, we gather data from various sources, like network traffic, user behavior, and system logs. Using the examples mentioned, the proposed system detects and prevents cyber threats more accurately. Artificial Intelligence techniques including deep learning, clustering, and principal component analysis (PCA), are used. These are essential in analyzing complex patterns within the data and enabling the detection of sophisticated and evolving intrusion techniques. Multidimensional data allows for the capture of intricate, non-linear relationships. This improves the system’s ability to differentiate between normal and abnormal activities. Real-time data processing and AI-driven algorithms enhance detection speed and enable faster responses to potential intrusions. We tested this system on benchmark datasets. Results showed significant improvements in detection rates and a reduction in false positives compared to traditional NIDS. The integration of AI gave us a more adaptive and scalable approach to intrusion detection. Also, it allowed the system to learn from new attack patterns and continuously refine its capabilities.

Hyperspectral analysis to assess gametocytogenesis stage progression in malaria-infected human erythrocytes.

Developments of anti-gametocyte drugs have been delayed due to insufficient understanding of gametocyte biology. We report a systematic workflow of data processing algorithms to quantify changes in the absorption spectrum and cell morphology of single malaria-infected erythrocytes. These changes may serve as biomarkers instrumental for the future development of antimalarial strategies, especially for anti-gametocyte drug design and testing. Image-based biomarkers may also be useful for nondestructive, label-free malaria detection and drug efficacy evaluation in resource-limited communities. We extend the application of hyperspectral microscopy to provide detailed insights into gametocyte stage progression through the quantitative analysis of absorbance spectra and cell morphology in malaria-infected erythrocytes. Malaria-infected erythrocytes at asexual and different gametocytogenesis stages were imaged through hyperspectral confocal microscopy. The preprocessing of the hyperspectral data cubes to transform them to color images and spectral angle mapper (SAM) analysis were first used to segment hemoglobin (Hb)- and hemozoin (Hz)-abundant areas within the host erythrocytes. Correlations between changes in cell morphology and increasing Hz-abundant areas of the infected erythrocytes were then examined to test their potential as optical biomarkers to determine the progression of infection, involving transitions from asexual to various gametocytogenesis stages. Following successful segmentation of Hb- and Hz-abundant areas in malaria-infected erythrocytes through SAM analysis, a modest correlation between the segmented Hz-abundant area and cell shape changes over time was observed. A significant increase in both the areal fraction of Hz and the ellipticity of the cell confirms that the Hz fraction change correlates with the progression of gametocytogenesis. Our workflow enables the quantification of changes in host cell morphology and the relative contents of Hb and Hz at various parasite growth stages. The quantified results exhibit a trend that both the segmented areal fraction of intracellular Hz and the ellipticity of the host cell increase as gametocytogenesis progresses, suggesting that these two metrics may serve as useful biomarkers to determine the stage of gametocytogenesis.

Self-homodyne 2-OTDM for doubling the baud rate in low-energy optical interconnect.

As a low-energy method to increase the data rate of optical links in data centers, we propose self-homodyne Nyquist optical time division multiplexing (OTDM). In Nyquist OTDM, spectrally efficient high-baud rate signals can be generated exceeding the limit of electronic signal processing. However, full integration of OTDM systems has not been reported, mainly because of the complicated signal detection scheme, which involves demultiplexing and clock recovery. In our proposal, the Nyquist pulse train is transmitted to the receiver as a local oscillator (LO) to leverage self-homodyne detection, which allows using large linewidth lasers and a simplified digital signal processing (DSP) algorithm. As the transmitted pulse train serves as an optical clock, demultiplexing and detection of the OTDM signal can be performed without using power-intensive high-bandwidth electronics and DSP. In this method, the LO pulse train needs to enter the coherent detector in exact synchronization with the OTDM signal for detecting the individual tributary correctly. For this purpose, we present a pulse delay control method suitable for photonic integration. A Nyquist pulse train with m carriers enables m-time multiplexing of optical signals. We explain and demonstrate the proposed concept in the case of m = 2, as it is the most feasible implementation. In the O-band where the chromatic dispersion (CD) is negligible, DSP-free operation can be achieved using the QPSK format. At the band edges where CD is non-negligible, it can be compensated by the DSP as in the conventional coherent detection. We verify this numerically and in an experiment involving the transmission of a 64-Gbaud QPSK signal at 1550 nm over a single-mode fiber. In terms of low energy, self-homodyne Nyquist OTDM is advantageous in wavelength division multiplexing (WDM). Taking it into consideration, we perform 4-channel WDM transmission of the 64-Gbaud QPSK signal over a 1-km dispersion shifted fiber without CD compensation. The results demonstrate a data rate of 512 Gb/s with a BER of <1 × 10-10.

Automated Detection of Cancer-Suspicious Findings in Japanese Radiology Reports with Natural Language Processing: A Multicenter Study.

Missed critical imaging findings, particularly those indicating cancer, are a common issue that can result in delays in patient follow-up and treatment. To address this, we developed a rule-based natural language processing (NLP) algorithm to detect cancer-suspicious findings from Japanese radiology reports. The dataset used consisted of chest and abdomen CT reports from six institutions. Reports from our institution were used for algorithm development and internal evaluation, while reports from the other five institutions were used for external evaluation. To create the gold standard, reports were annotated by two experienced physicians. Data were statistically analyzed using precision, recall and F1 score with 1000 bootstrap iterations. BERT was used as a baseline deep learning model, and its performance was compared with the proposed rule-based method. At the report level of detection, the overall precision, recall, and F-1 score were 0.886, 0.886, and 0.883, respectively, for the rule-based algorithm, which were higher than those of the deep learning algorithm (0.851, 0.679, and 0.733). The overall results include both internal and external validation data. For the internal validation set, the precision, recall, and F-1 score were 0.929, 0.929, and 0.927, respectively. For the external validation set, the precision, recall, and F-1 score were 0.875, 0.879, and 0.873, demonstrating generalizability. In conclusion, we show the rule-based NLP algorithm exhibited a high performance in detecting cancer-suspicious findings from multi-institutional CT reports.

Longitudinal evolution of the transdiagnostic prodrome to severe mental disorders: a dynamic temporal network analysis informed by natural language processing and electronic health records.

Modelling the prodrome to severe mental disorders (SMD), including unipolar mood disorders (UMD), bipolar mood disorders (BMD) and psychotic disorders (PSY), should consider both the evolution and interactions of symptoms and substance use (prodromal features) over time. Temporal network analysis can detect causal dependence between and within prodromal features by representing prodromal features as nodes, with their connections (edges) indicating the likelihood of one feature preceding the other. In SMD, node centrality could reveal insights into important prodromal features and potential intervention targets. Community analysis can identify commonly occurring feature groups to define SMD at-risk states. This retrospective (2-year) cohort study aimed to develop a global transdiagnostic SMD network of the temporal relationships between prodromal features and to examine within-group differences with sub-networks specific to UMD, BMD and PSY. Electronic health records (EHRs) from South London and Maudsley(SLaM) NHS Foundation Trust were included from 6462 individuals with SMD diagnoses (UMD:2066; BMD:740; PSY:3656). Validated natural language processing algorithms extracted the occurrence of 61 prodromal features every three months from two years to six months before SMD onset. Temporal networks of prodromal features were constructed using generalised vector autoregression panel analysis, adjusting for covariates. Edge weights (partial directed correlation coefficients, z) were reported in autocorrelative, unidirectional and bidirectional relationships. Centrality was calculated as the sum of (non-autoregressive) connections leaving (out-centrality, cout) or entering (in-centrality, cin) a node. The three sub-networks (UMD, BMD, PSY) were compared using permutation analysis, and community analysis was performed using Spinglass. The SMD network revealed strong autocorrelations (0.04 ≤ z ≤ 0.10), predominantly positive connections, and identified aggression (cout = 0.103) and tearfulness (cin = 0.134) as the most central features. Sub-networks for UMD, BMD, and PSY showed minimal differences, with 3.5% of edges differing between UMD and PSY, 0.8% between UMD and BMD, and 0.4% between BMD and PSY. Community analysis identified one positive psychotic community (delusional thinking-hallucinations-paranoia) and two behavioural communities (aggression-cannabis use-cocaine use-hostility, aggression-agitation-hostility) as the most common. This study represents the most extensive temporal network analysis conducted on the longitudinal interplay of SMD prodromal features. The findings provide further evidence to support transdiagnostic early detection services across SMD, refine assessments to detect individuals at risk and identify central features as potential intervention targets.

Toward Efficient Edge Detection: A Novel Optimization Method Based on Integral Image Technology and Canny Edge Detection

The traditional SIFT (Scale Invariant Feature Transform) registration algorithm is highly regarded in the field of image processing due to its scale invariance, rotation invariance, and robustness to noise. However, it faces challenges such as a large number of feature points, high computational demand, and poor real-time performance when dealing with large-scale images. A novel optimization method based on integral image technology and canny edge detection is presented in this paper, aiming to maintain the core advantages of the SIFT algorithm while reducing the complexity involved in image registration computations, enhancing the efficiency of the algorithm for real-time image processing, and better adaption to the needs of large-scale image handling. Firstly, Gaussian separation techniques were used to simplify Gaussian filtering, followed by the application of integral image techniques to accelerate the construction of the entire pyramid. Additionally, during the feature point detection phase, an innovative feature point filtering strategy was introduced by combining Canny edge detection with dilation operations alongside the traditional SIFT approach, aiming to reduce the number of feature points and thereby lessen the computational load. The method proposed in this paper takes 0.0134 s for Image type a, 0.0504 s for Image type b, and 0.0212 s for Image type c. In contrast, the traditional method takes 0.1452 s for Image type a, 0.5276 s for Image type b, and 0.2717 s for Image type c, resulting in reductions of 0.1318 s, 0.4772 s, and 0.2505 s, respectively. A series of comparative experiments showed that the time taken to construct the Gaussian pyramid using our proposed method was consistently lower than that required by the traditional method, indicating greater efficiency and stability regardless of image size or type.

Performance of K-Nearest Neighbor Algorithm and C4.5 Algorithm in Classifying Citizens Eligible to Receive Direct Cash Assistance in Bandar Mahligai Village

Direct Cash Assistance, commonly called BLT, is one of the many programs the Indonesian government held to reduce the poverty rate of the Indonesian population. This study compares the KNN and C4.5 methods to determine the eligibility of residents eligible to receive Direct Cash Assistance in Bandar Mahligai Village. This study began with collecting resident data from the Bandar Mahligai village office. Then, the data obtained was taken into several attributes to be used in the classification process, namely the name of the head of the family, KK number, NIK, number of dependents, occupation, income, and monthly expenses. After the data is collected, the data will be classified using the KNN and C4.5 algorithms. There is a significant difference between the two algorithms in the classification process; the KNN algorithm by looking for the nearest neighbor data value, in this study, the K value = 9, while the C4.5 algorithm by building a decision tree from the attribute values taken based on resident data used as training data. The classification results of the two methods will be compared using a confusion matrix to obtain a higher accuracy technique. The results of testing using a confusion matrix for both algorithms are the accuracy produced by the KNN and C4.5 algorithms in classifying residents eligible for Direct Cash Assistance (BLT) of 90% in the system that has been built. The results of comparing the KNN and C4.5 algorithms for this study show that the KNN algorithm is better because the accuracy level reaches 90% in manual and system calculations. While the C4.5 method only gets 85% for the accuracy of its manual calculations, it receives an accuracy level of 90% in the system that has been built.

Production planning optimization with fuzzy analytic hierarchy process and genetic algorithm

Proper production planning is essential for improving productivity and lowering resource (material, energy, employees) related costs in the highly competitive business world. Dealing with the challenges of asymmetric setup times—where the time required to switch between manufacturing different products varies —makes this task much more difficult. Conventional planning techniques frequently ignore these articulations and produce sub-optimal schedules. This paper proposes a novel approach to tackle the following challenge: optimizing production planning using the Fuzzy Analytic Hierarchy Process (FAHP) with asymmetric setup times and Genetic Algorithm (GA). The proposed methodology involves a step-by-step process. The first stage defines key objectives: makespan, total waste cost, and maximum weighted tardiness. Decision-makers compare the relative importance of each criterion within its hierarchy level using fuzzy numbers. The consistency of these comparisons is assessed using fuzzy consistency ratio computations. At the same time, the overall priority weights for each production planning alternative are determined by summing fuzzy judgments across the hierarchy. In the second stage, the production plan is optimized using GA, considering sequence and lot size variables and asymmetric setup times, by applying the computed weights. The comparisons are performed using the proposed approach with the optimum solution.

Image Processing using VITIS Model Composer

This study outlines a method for image processing utilizing the Vitis Model Composer. AMD's Vitis Model Composer includes comprehensive libraries that support a variety of algorithms. In this work, it is integrated within the MATLAB Simulink environment. A model-based design approach is employed to implement various image processing algorithms. To ensure accuracy, hardware co-simulation is conducted for result verification. The study covers several image processing algorithms, including those for converting RGB images to grayscale, creating image negatives, enhancing images, subtracting backgrounds, applying thresholding, and performing morphological operations such as erosion, dilation, and masking, using AMD's provided blocks. Key Words: optics, photonics, light, lasers, templates, journals

Field-Programmable Gate Array (FPGA)-Based Lock-In Amplifier System with Signal Enhancement: A Comprehensive Review on the Design for Advanced Measurement Applications.

Lock-in amplifiers (LIAs) are critical tools in precision measurement, particularly for applications involving weak signals obscured by noise. Advances in signal processing algorithms and hardware synthesis have enabled accurate signal extraction, even in extremely noisy environments, making LIAs indispensable in sensor applications for healthcare, industry, and other services. For instance, the electrical impedance measurement of the human body, organs, tissues, and cells, known as bioelectrical impedance, is commonly used in biomedical and healthcare applications because it is non-invasive and relatively inexpensive. Also, due to its portability and miniaturization capabilities, it has great potential for the development of new point-of-care and portable testing devices. In this document, we highlight existing techniques for high-frequency resolution and precise phase detection in LIA reference signals from field-programmable gate array (FPGA) designs. A comprehensive review is presented under the key requirements and techniques for single- and dual-phase digital LIA architectures, where relevant insights are provided to address the LIAs' digital precision in measurement system configurations. Furthermore, the document highlights a novel method to enhance the spurious-free dynamic range (SFDR), thereby advancing the precision and effectiveness of LIAs in complex measurement environments. Finally, we summarize the diverse applications of impedance measurement, highlighting the wide range of fields that can benefit from the design of high performance in modern measurement technologies.

Advanced IoT-integrated parking systems with automated license plate recognition and payment management

Urban parking management is a growing challenge with increasing vehicle numbers and limited parking space. Traditional methods often fail during peak hours, leading to inefficiencies, unauthorized usage, and revenue losses. For instance, a parking lot designed for 300 vehicles often exceeds 90% occupancy during peak times, creating congestion and billing inaccuracies. This research proposes an automated system integrating sensors, image processing, and database management to address these issues. A single camera monitors multiple parking slots, with predefined coordinates linked to IR sensors for dual verification. Image processing algorithms, including Optical Character Recognition (OCR), enable accurate license plate recognition. Testing under real-world conditions showed 95% accuracy in daylight, 90% in low light, and 93% for plates at 45-degree angles. Detection accuracy reached 88% at distances of 1.5–3 m, ensuring reliable operation even at the camera’s range limits. Occupancy tracking achieved less than a 5% error margin compared to manual methods, while the fare calculation module reduced billing errors by 90%, enhancing efficiency and revenue. The system’s scalable design supports applications in parking management, toll collection, and traffic monitoring. By improving vehicle detection, occupancy tracking, and billing accuracy, this solution addresses critical challenges in urban parking and contributes to smarter city infrastructure.

On the hidden layer-to-layer topology of the representations of reality realised within neural networks

PurposeConsider an information processing algorithm that is designed to process an input data object onto an output data object via a number of successive internal {\it layers} and mappings between them. The possible activation state within each layer can be represented as a cube within Euclidean space of a high dimension (e.g. equal to the number of artificial neurons at that level). Multiple instances of such input objects produce a point cloud within each layer’s cube: this is the “representation of the reality” at that layer, as sampled by the set of input objects.Design/methodology/approachMost neural networks reduce the dimension of each layer’s cube from layer to successive layer. This gives the false impression of refining the inner representations of reality, distilling it down to fewer dimensions from which to discriminate or to infer outcomes (whatever is the aim). However, the representation of reality realised within each layer’s cube is a manifold, a curved subset embedded within it and of much lower dimension. Investigations show that such manifolds may not always be reducing in their local dimension. Instead, the manifold may become folded over and over, filling up further dimensions and creating non-realistic (unforeseeable) proximities.FindingsWe discuss some of the likely consequences of these relatively unforeseen characteristics and, in particular, the possible vulnerability of such algorithms to non-realistic perturbations. We consider a possible response to this issue.Practical implicationsNew forms of calibration are necessary, using geometric/topological loss functions, as opposed to simple (variation-limiting) regularisation terms.Originality/valueWe apply persistent homology methods to understand how the images of the point cloud (representing the sampled reality) change as they pass from layer to layer.

Object Detection Post Processing Accelerator Based on Co-Design of Hardware and Software

Deep learning significantly advances object detection. Post processes, a critical component of this process, select valid bounding boxes to represent the true targets during inference and assign boxes and labels to these objects during training to optimize the loss function. However, post processes constitute a substantial portion of the total processing time for a single image. This inefficiency primarily arises from the extensive Intersection over Union (IoU) calculations required between numerous redundant bounding boxes in post processing algorithms. To reduce these redundant IoU calculations, we introduce a classification prioritization strategy during both training and inference post processes. Additionally, post processes involve sorting operations that contribute to their inefficiency. To minimize unnecessary comparisons in Top-K sorting, we have improved the bitonic sorter by developing a hybrid bitonic algorithm. These improvements have effectively accelerated the post processing. Given the similarities between the training and inference post processes, we unify four typical post processing algorithms and design a hardware accelerator based on this framework. Our accelerator achieves at least 7.55 times the speed in inference post processing compared to that of recent accelerators. When compared to the RTX 2080 Ti system, our proposed accelerator offers at least 21.93 times the speed for the training post process and 19.89 times for the inference post process, thereby significantly enhancing the efficiency of loss function minimization.

Identifying Metabolic Dysfunction Associated Steatotic Liver Disease Using Natural Language Processing in a U.S National Cohort

Background and Aims: Although metabolic dysfunction-associated steatotic liver disease (MASLD) and MASLD with increased alcohol intake (MetALD) are identified as clinical entities, tools to identify patients from electronic health records (EHRs) to perform large outcome studies are lacking. Methods: In this retrospective study of participants from the Veterans Analysis of Liver Disease (VALID) cohort assembled from 1/1/2013 to 12/31/2022, a rule-based natural language processing (NLP) algorithm searched EHRs for imaging evidence of hepatic steatosis. This was combined with identification of cardiometabolic risk factors (CMRF) and harmful alcohol use. Algorithm-derived diagnoses of MASLD, MetALD, alcohol-associated steatotic liver disease (ALD), and no steatotic liver disease (SLD) were validated using a blinded review of randomly selected charts. Results: Among 817,657 eligible Veterans, SLD was present in over half (n=438,209, 53.5%), including MASLD in 299,259 (36.5%), 99,163 with MetALD (12.1%), 38,552 (4.7%) with ALD. The NLP algorithm had a high correlation with steatosis on chart review, with a kappa of 0.86 (95% CI 0.82-0.90), sensitivity of 0.96 and specificity of 0.90. Classification of MASLD, MetALD, ALD and no SLD by the algorithm also showed high correlation with chart review, with a kappa of 0.87 (95% CI 0.82-0.91). This algorithm identified 299,259 (36.5%) of the study cohort with MASLD, compared to 23,218 (2.8%) patients identified using ICD 9/10 codes. Conclusions and Relevance: An algorithm combining rule-based NLP with CMRFs and alcohol use from EHRs accurately identifies and classifies SLD and can be applied in large epidemiologic studies of SLD in the VHA.

Binaural speech intelligibility for combinations of noise, reverberation, and hearing-aid signal processing.

Binaural speech intelligibility in rooms is a complex process that is affected by many factors including room acoustics, hearing loss, and hearing aid (HA) signal processing. Intelligibility is evaluated in this paper for a simulated room combined with a simulated hearing aid. The test conditions comprise three spatial configurations of the speech and noise sources, simulated anechoic and concert hall acoustics, three amounts of multitalker babble interference, the hearing status of the listeners, and three degrees of simulated HA processing provided to compensate for the noise and/or hearing loss. The impact of these factors and their interactions is considered for normal-hearing (NH) and hearing-impaired (HI) listeners for sentence stimuli. Both listener groups showed a significant reduction in intelligibility as the signal-to-noise ratio (SNR) decreased, and showed a reduction in intelligibility in reverberation when compared to anechoic listening. There was no significant improvement in intelligibility for the NH group for the noise suppression algorithm used here, and no significant improvement in intelligibility for the HI group for more advanced HA processing algorithms as opposed to linear amplification in either of the two acoustic spaces or at any of the three SNRs.