Manual Refinement Research Articles

Parasagittal dural volume correlates with cerebrospinal fluid volume and developmental delay in children with autism spectrum disorder

BackgroundThe parasagittal dura, a tissue that lines the walls of the superior sagittal sinus, acts as an active site for immune-surveillance, promotes the reabsorption of cerebrospinal fluid, and facilitates the removal of metabolic waste products from the brain. Cerebrospinal fluid is important for the distribution of growth factors that signal immature neurons to proliferate and migrate. Autism spectrum disorder is characterized by altered cerebrospinal fluid dynamics.MethodsIn this retrospective study, we investigated potential correlations between parasagittal dura volume, brain structure volumes, and clinical severity scales in young children with autism spectrum disorder. We employed a semi-supervised two step pipeline to extract parasagittal dura volume from 3D-T2 Fluid Attenuated Inversion Recovery sequences, based on U-Net followed by manual refinement of the extracted parasagittal dura masks.ResultsHere we show that the parasagittal dura volume does not change with age but is significantly correlated with cerebrospinal fluid (p-value = 0.002), extra-axial cerebrospinal fluid volume (p-value = 0.0003) and severity of developmental delay (p-value = 0.024).ConclusionsThese findings suggest that autism spectrum disorder children with severe developmental delay may have a maldeveloped parasagittal dura that potentially perturbs cerebrospinal fluid dynamics.

A deep learning-based method enables the automatic and accurate assembly of chromosome-level genomes.

The application of high-throughput chromosome conformation capture (Hi-C) technology enables the construction of chromosome-level assemblies. However, the correction of errors and the anchoring of sequences to chromosomes in the assembly remain significant challenges. In this study, we developed a deep learning-based method, AutoHiC, to address the challenges in chromosome-level genome assembly by enhancing contiguity and accuracy. Conventional Hi-C-aided scaffolding often requires manual refinement, but AutoHiC instead utilizes Hi-C data for automated workflows and iterative error correction. When trained on data from 300+species, AutoHiC demonstrated a robust average error detection accuracy exceeding 90%. The benchmarking results confirmed its significant impact on genome contiguity and error correction. The innovative approach and comprehensive results of AutoHiC constitute a breakthrough in automated error detection, promising more accurate genome assemblies for advancing genomics research.

Main challenges on the curation of large scale datasets for pancreas segmentation using deep learning in multi-phase CT scans: Focus on cardinality, manual refinement, and annotation quality

Accurate segmentation of the pancreas in computed tomography (CT) holds paramount importance in diagnostics, surgical planning, and interventions. Recent studies have proposed supervised deep-learning models for segmentation, but their efficacy relies on the quality and quantity of the training data. Most of such works employed small-scale public datasets, without proving the efficacy of generalization to external datasets. This study explored the optimization of pancreas segmentation accuracy by pinpointing the ideal dataset size, understanding resource implications, examining manual refinement impact, and assessing the influence of anatomical subregions. We present the AIMS-1300 dataset encompassing 1,300 CT scans. Its manual annotation by medical experts required 938 h. A 2.5D UNet was implemented to assess the impact of training sample size on segmentation accuracy by partitioning the original AIMS-1300 dataset into 11 smaller subsets of progressively increasing numerosity. The findings revealed that training sets exceeding 440 CTs did not lead to better segmentation performance. In contrast, nnU-Net and UNet with Attention Gate reached a plateau for 585 CTs. Tests on generalization on the publicly available AMOS-CT dataset confirmed this outcome. As the size of the partition of the AIMS-1300 training set increases, the number of error slices decreases, reaching a minimum with 730 and 440 CTs, for AIMS-1300 and AMOS-CT datasets, respectively. Segmentation metrics on the AIMS-1300 and AMOS-CT datasets improved more on the head than the body and tail of the pancreas as the dataset size increased. By carefully considering the task and the characteristics of the available data, researchers can develop deep learning models without sacrificing performance even with limited data. This could accelerate developing and deploying artificial intelligence tools for pancreas surgery and other surgical data science applications.

Precise ablation zone segmentation on CT images after liver cancer ablation using semi‐automatic CNN‐based segmentation

AbstractBackgroundAblation zone segmentation in contrast‐enhanced computed tomography (CECT) images enables the quantitative assessment of treatment success in the ablation of liver lesions. However, fully automatic liver ablation zone segmentation in CT images still remains challenging, such as low accuracy and time‐consuming manual refinement of the incorrect regions.PurposeTherefore, in this study, we developed a semi‐automatic technique to address the remaining drawbacks and improve the accuracy of the liver ablation zone segmentation in the CT images.MethodsOur approach uses a combination of a CNN‐based automatic segmentation method and an interactive CNN‐based segmentation method. First, automatic segmentation is applied for coarse ablation zone segmentation in the whole CT image. Human experts then visually validate the segmentation results. If there are errors in the coarse segmentation, local corrections can be performed on each slice via an interactive CNN‐based segmentation method. The models were trained and the proposed method was evaluated using two internal datasets of post‐interventional CECT images ( = 22, = 145; 62 patients in total) and then further tested using an external benchmark dataset ( = 12; 10 patients).ResultsTo evaluate the accuracy of the proposed approach, we used Dice similarity coefficient (DSC), average symmetric surface distance (ASSD), Hausdorff distance (HD), and volume difference (VD). The quantitative evaluation results show that the proposed approach obtained mean DSC, ASSD, HD, and VD scores of 94.0%, 0.4 mm, 8.4 mm, 0.02, respectively, on the internal dataset, and 87.8%, 0.9 mm, 9.5 mm, and −0.03, respectively, on the benchmark dataset. We also compared the performance of the proposed approach to that of five well‐known segmentation methods; the proposed semi‐automatic method achieved state‐of‐the‐art performance on ablation segmentation accuracy, and on average, 2 min are required to correct the segmentation. Furthermore, we found that the accuracy of the proposed method on the benchmark dataset is comparable to that of manual segmentation by human experts ( = 0.55, ‐test).ConclusionsThe proposed semi‐automatic CNN‐based segmentation method can be used to effectively segment the ablation zones, increasing the value of CECT for an assessment of treatment success. For reproducibility, the trained models, source code, and demonstration tool are publicly available at https://github.com/lqanh11/Interactive_AblationZone_Segmentation.

A cooperative coevolutionary genetic programming hyper-heuristic for multi-objective makespan and cost optimization in cloud workflow scheduling

This study presents a novel multi-objective approach for NP-hard workflow scheduling in cloud computing environments. Traditional rule-based heuristics offer flexibility but lack consistent superiority across all criteria and scenarios. The development of specific rules usually requires tedious manual refinement by experts. To overcome this limitation, our method leverages evolutionary computation and simulation to automatically derive new rules. Moreover, workflow scheduling involves two crucial and related aspects: task scheduling and the allocation of virtual machines. Our contribution includes three distinct algorithms: two that address each decision separately and a third approach that coevolves priority rules for both decisions simultaneously. Computational tests demonstrate superior performance, with an exploration of rules yielding a 72.91% larger hypervolume for optimizing makespan and costs compared to benchmark heuristics from the literature. The validation on unseen instances shows a 90.26% improvement in hypervolume performance, highlighting the robustness of our approach.

AAontology: An Ontology of Amino Acid Scales for Interpretable Machine Learning

Amino acid scales are crucial for protein prediction tasks, many of them being curated in the AAindex database. Despite various clustering attempts to organize them and to better understand their relationships, these approaches lack the fine-grained classification necessary for satisfactory interpretability in many protein prediction problems.To address this issue, we developed AAontology—a two-level classification for 586 amino acid scales (mainly from AAindex) together with an in-depth analysis of their relations—using bag-of-word-based classification, clustering, and manual refinement over multiple iterations. AAontology organizes physicochemical scales into 8 categories and 67 subcategories, enhancing the interpretability of scale-based machine learning methods in protein bioinformatics. Thereby it enables researchers to gain a deeper biological insight. We anticipate that AAontology will be a building block to link amino acid properties with protein function and dysfunctions as well as aid informed decision-making in mutation analysis or protein drug design.

Retrieval of dominant methane (CH4) emission sources, the first high-resolution (1–2 m) dataset of storage tanks of China in 2000–2021

Abstract. Methane (CH4) is a significant greenhouse gas in exacerbating climate change. Approximately 25 % of CH4 is emitted from storage tanks. It is crucial to spatially explore the CH4 emission patterns from storage tanks for efficient strategy proposals to mitigate climate change. However, due to the lack of publicly accessible storage tank locations and distributions, it is difficult to ascertain the CH4 emission spatial pattern over a large-scale area. To address this problem, we generated a storage tank dataset (STD) by implementing a deep learning model with manual refinement based on 4403 high-spatial-resolution images (1–2 m) from the Gaofen-1, Gaofen-2, Gaofen-6, and Ziyuan-3 satellites over city regions in China with officially reported numerous storage tanks in 2021. STD is the first storage tank dataset for over 92 typical city regions in China. The dataset can be accessed at https://doi.org/10.5281/zenodo.10514151 (Chen et al., 2024). It provides a detailed georeferenced inventory of 14 461 storage tanks wherein each storage tank is validated and assigned the construction year (2000–2021) by visual interpretation of the collected high-spatial-resolution images, historical high-spatial-resolution images of Google Earth, and field survey. The inventory comprises storage tanks with various distribution patterns in different city regions. Spatial consistency analysis with the CH4 emission product shows good agreement with storage tank distributions. The intensive construction of storage tanks significantly induces CH4 emissions from 2005 to 2020, underscoring the need for more robust measures to curb CH4 release and aid in climate change mitigation efforts. Our proposed dataset, STD, will foster the accurate estimation of CH4 released from storage tanks for CH4 control and reduction and ensure more efficient treatment strategies are proposed to better understand the impact of storage tanks on the environment, ecology, and human settlements.

Pelvic Palpatory Tests in Manual Therapy and Osteopathy: A Critical Review of the Literature and Suggestions for New Research.

Manual therapists apply physical interventions to the entire structure of the body to promote healing, prevent pathologies, and improve patient health. In osteopathic practice, palpatory evaluation is considered a fundamental clinical practice requirement for identifying somatic dysfunction. Most of the articles published in this area have failed to demonstrate a level of reproducibility that supports palpation in evidence-based clinical practice. When considering the poor reliability of the palpatory tests highlighted in the literature, a discrepancy is noted with what is known about the tactile sensitivity of human hands. For static touch, the minimum size that can be detected, in the absence of applied movement or vibration, is approximately 0.2 mm. Yet, it seems that this high level of precision is insufficient to ensure reliability in the tests used to evaluate osteopathic somatic dysfunction. The purpose that underscores this article is to determine how these two contradictory elements, high sensitivity and low reliability, can be present in palpatory tests. The article reports the literature findings regarding palpatory tests of pelvic, which is an important structure for clinical purposes. Additionally, a critical review of how these studies were conducted is provided to identify any elements that may justify the obtained results. Following recent accredited guidelines present in the literature, we propose suggestions on vision training methods, manual perception refinement training, the search for anatomical markers, and the position of the examiner in relation to the examineethat may be useful for future studies on the topic covered by the article.

Accuracy of artificial intelligence‐based segmentation of the mandibular canal in CBCT

AbstractObjectivesTo investigate the accuracy of artificial intelligence (AI)‐based segmentation of the mandibular canal, compared to the conventional manual tracing, implementing implant planning software.Materials and methodsLocalization of the mandibular canals was performed for 104 randomly selected patients. A localization was performed by three experienced clinicians in order to serve as control. Five tracings were performed: One from a clinician with a moderate experience with a manual tracing (I1), followed by the implementation of an automatic refinement (I2), one manual from a dental student (S1), and one from the experienced clinician, followed by an automatic refinement (E). Subsequently, two fully automatic AI‐driven segmentations were performed (A1,A2). The accuracy between each method was measured using root mean square error calculation.ResultsThe discrepancy among the models of the mandibular canals, between the experienced clinicians and each investigated method ranged from 0.21 to 7.65 mm with a mean of 3.5 mm RMS error. The analysis of each separate mandibular canal's section revealed that mean RMS error was higher in the posterior and anterior loop compared to the middle section. Regarding time efficiency, tracing by experienced users required more time compared to AI‐driven segmentation.ConclusionsThe experience of the clinician had a significant influence on the accuracy of mandibular canal's localization. An AI‐driven segmentation of the mandibular canal constitutes a time‐efficient and reliable procedure for pre‐operative implant planning. Nevertheless, AI‐based segmentation results should always be verified, as a subsequent manual refinement of the initial segmentation may be required to avoid clinical significant errors.

Antimicrobial resistance and self-reported hand hygiene awareness before and after an infection prevention and control programme: A 7-year analysis in a small animal veterinary teaching hospital

Infection prevention and control (IPC) in veterinary medicine is crucial to protect patients, owners, staff, and the public. An IPC programme is recommended for every animal hospital. The objective of this retrospective longitudinal study was to describe the changes in bacterial and multidrug-resistant (MDR) bacterial isolates and self-reported hand hygiene awareness and practices after an IPC programme to assess the long-term effect of this programme in small animal veterinary medicine. The IPC programme was implemented at our veterinary teaching hospital in April 2018, which included the establishment of an infection control task force, regular IPC lectures and poster campaigns, infrastructure improvement, and manual refinement. Laboratory-based surveillance was retrospectively conducted before and after the programme (January 2016–December 2022). Level and slope changes in bacterial isolates were evaluated using interrupted time-series analysis. Self-reported hand hygiene awareness and practices were assessed using an annual questionnaire. Additionally, hygiene product purchases during the study period were investigated.The monthly number of total and MDR bacterial isolates decreased significantly after the programme (MDR level change: −0.426; 95% confidence interval: −0.744, −0.109; P = 0.009; and MDR slope change: −0.035; 95% confidence interval: −0.058, −0.011; P = 0.003). Additionally, awareness of hand hygiene before touching animals improved after the programme. Overall self-reported hand hygiene practices improved, and hygiene product purchases significantly increased. These results suggested that the IPC programme may have long-term effects regarding reducing total and MDR bacterial isolates and improving hand hygiene awareness in veterinary medicine.

Large language model based framework for automated extraction of genetic interactions from unstructured data.

Extracting biological interactions from published literature helps us understand complex biological systems, accelerate research, and support decision-making in drug or treatment development. Despite efforts to automate the extraction of biological relations using text mining tools and machine learning pipelines, manual curation continues to serve as the gold standard. However, the rapidly increasing volume of literature pertaining to biological relations poses challenges in its manual curation and refinement. These challenges are further compounded because only a small fraction of the published literature is relevant to biological relation extraction, and the embedded sentences of relevant sections have complex structures, which can lead to incorrect inference of relationships. To overcome these challenges, we propose GIX, an automated and robust Gene Interaction Extraction framework, based on pre-trained Large Language models fine-tuned through extensive evaluations on various gene/protein interaction corpora including LLL and RegulonDB. GIX identifies relevant publications with minimal keywords, optimises sentence selection to reduce computational overhead, simplifies sentence structure while preserving meaning, and provides a confidence factor indicating the reliability of extracted relations. GIX's Stage-2 relation extraction method performed well on benchmark protein/gene interaction datasets, assessed using 10-fold cross-validation, surpassing state-of-the-art approaches. We demonstrated that the proposed method, although fully automated, performs as well as manual relation extraction, with enhanced robustness. We also observed GIX's capability to augment existing datasets with new sentences, incorporating newly discovered biological terms and processes. Further, we demonstrated GIX's real-world applicability in inferring E. coli gene circuits.

Optimization of the on-power refueling strategy for the slightly enriched uranium core of Atucha-II NPP

The development of an on-power refueling strategy requires meticulous consideration due to its impact on relevant reactor parameters during operation. The rapid increase in computational power promotes the use of heuristic (or metaheuristic) methods for optimizing systems with numerous core calculations, such as designing a refueling strategy. Building on the previous success of optimizing the on-power refueling strategy for the Atucha-I core using the Simulated Annealing algorithm, the approach was extended to design an enhanced refueling strategy for the future use of slightly enriched uranium fuel in Atucha-II. Previous efforts manually achieved an acceptable refueling strategy, consistent with 3 irradiation paths, each having 3 burnup zones. The primary goal of the optimization was to reduce the usage of the refueling machine, a critical component for plant availability and performance. To achieve this, the number of burnup zones for the optimization was set at 2 per irradiation path. An acceptable optimized refueling strategy was obtained and, subsequently, manual refinements were introduced to improve the time-average channel power distribution. To assess the feasibility of the obtained 2-burnup-zones per path refueling strategy, a simulation of the reactor operation including on-power refueling consistent with this refueling strategy was conducted for 4 years at full power. The simulation predicts a 25.6 % reduction in the refueling machine’s use at full power and savings of 2.9 fuel assemblies per year compared to operation with the 3-burnup-zones per path refueling strategy. The compelling advantages in the performance of the 2-burnup-zones per path refueling strategy have led to its formal adoption as the new official refueling strategy for the design phase of the use of slightly enriched uranium fuel in the Atucha-II core.

Rapid Creation of Knowledge-Balanced Student Groups Using ChatGPT4.

We demonstrated use of ChatGPT4 for efficient group formation in undergraduate medical education. ChatGPT4 created balanced groups considering students' backgrounds in microbiology, physiology, genetics, and immunology considerably faster than manual efforts. Manual refinements included gender balance and discipline distribution. Improvements included ChatGTP's ability to further incorporate GPA and MCAT scores.

Relaxation Along a Fictitious Field, continuous wave T1rho, adiabatic T1rho and adiabatic T2rho imaging of human gliomas at 3T: A feasibility study.

In pre-clinical models of brain gliomas, Relaxation Along a Fictitious Field in second rotating frame (TRAFF2), continues wave T1rho (T1ρcw), adiabatic T1rho (T1ρadiab), and adiabatic T2rho (T2ρadiab) relaxation time mappings have demonstrated potential to non-invasively characterize brain gliomas. Our aim was to evaluate the feasibility and potential of 4 different spin lock methods at 3T to characterize primary brain glioma. 22 patients (26-72 years) with suspected primary glioma. T1ρcw was performed using pulse peak amplitude of 500Hz and pulse train durations of 40 and 80 ms while the corresponding values for T1ρadiab, T2ρadiab, TRAFF2 were 500/500/500Hz and 48 and 96, 64 and 112, 45 and 90 ms, respectively. The parametric maps were calculated using a monoexponential model. Molecular profiles were evaluated from tissue specimens obtained during the resection. The lesion regions-of-interest were segmented from high intensity FLAIR using automatic segmentation with manual refinement. Statistical descriptors from the voxel intensity values inside each lesion and radiomic features (Pyrad MRC package) were calculated. From extracted radiomics, mRMRe R package version 2.1.0 was used to select 3 features in each modality for statistical comparisons. Of the 22 patients, 10 were found to have IDH-mutant gliomas and of those 5 patients had 1p/19q codeletion group comparisons. Following correction for effects of age and gender, at least one statistical descriptor was able to differentiate between IDH and 1p/19q codeletion status for all the parametric maps. In the radiomic analysis, corner-edge detector features with Harris-Stephens filtered signal showed significant group differences in IDH and 1p/19q codeletion groups. Spin lock imaging at 3T of human glioma was feasible and various qualitative parameters derived from the parametric maps were found to have potential to differentiate IDH and 1p19q codeletion status. Future larger prospective clinical trials are warranted to evaluate these methods further.

Enhancing cryo-EM structure prediction with DeepTracer and AlphaFold2 integration.

Understanding the protein structures is invaluable in various biomedical applications, such as vaccine development. Protein structure model building from experimental electron density maps is a time-consuming and labor-intensive task. To address the challenge, machine learning approaches have been proposed to automate this process. Currently, the majority of the experimental maps in the database lack atomic resolution features, making it challenging for machine learning-based methods to precisely determine protein structures from cryogenic electron microscopy density maps. On the other hand, protein structure prediction methods, such as AlphaFold2, leverage evolutionary information from protein sequences and have recently achieved groundbreaking accuracy. However, these methods often require manual refinement, which is labor intensive and time consuming. In this study, we present DeepTracer-Refine, an automated method that refines AlphaFold predicted structures by aligning them to DeepTracers modeled structure. Our method was evaluated on 39 multi-domain proteins and we improved the average residue coverage from 78.2 to 90.0% and average local Distance Difference Test score from 0.67 to 0.71. We also compared DeepTracer-Refine with Phenixs AlphaFold refinement and demonstrated that our method not only performs better when the initial AlphaFold model is less precise but also surpasses Phenix in run-time performance.

Building an adverse outcome pathway network for estrogen-, androgen- and steroidogenesis-mediated reproductive toxicity.

Introduction: Adverse Outcome Pathways (AOPs) can support both testing and assessment of endocrine disruptors (EDs). There is, however, a need for further development of the AOP framework to improve its applicability in a regulatory context. Here we have inventoried the AOP-wiki to identify all existing AOPs related to mammalian reproductive toxicity arising from disruption to the estrogen, androgen, and steroidogenesis modalities. Core key events (KEs) shared between relevant AOPs were also identified to aid in further AOP network (AOPN) development. Methods: A systematic approach using two different methods was applied to screen and search the entire AOP-wiki library. An AOPN was visualized using Cytoscape. Manual refinement was performed to remove AOPS devoid of any KEs and/or KERs. Results: Fifty-eight AOPs relevant for mammalian reproductive toxicity were originally identified, with 42 AOPs included in the final AOPN. Several of the KEs and KE relationships (KERs) described similar events and were thus merged to optimize AOPN construction. Sixteen sub-networks related to effects on hormone levels or hormone activity, cancer outcomes, male and female reproductive systems, and overall effects on fertility and reproduction were identified within the AOPN. Twenty-six KEs and 11 KERs were identified as core blocks of knowledge in the AOPN, of which 19 core KEs are already included as parameters in current OECD and US EPA test guidelines. Discussion: The AOPN highlights knowledge gaps that can be targeted for further development of a more complete AOPN that can support the identification and assessment of EDs.

Rapid and automated design of two-component protein nanomaterials using ProteinMPNN

The design of protein-protein interfaces using physics-based design methods such as Rosetta requires substantial computational resources and manual refinement by expert structural biologists. Deep learning methods promise to simplify protein-protein interface design and enable its application to a wide variety of problems by researchers from various scientific disciplines. Here, we test the ability of a deep learning method for protein sequence design, ProteinMPNN, to design two-component tetrahedral protein nanomaterials and benchmark its performance against Rosetta. ProteinMPNN had a similar success rate to Rosetta, yielding 13 new experimentally confirmed assemblies, but required orders of magnitude less computation and no manual refinement. The interfaces designed by ProteinMPNN were substantially more polar than those designed by Rosetta, which facilitated in vitro assembly of the designed nanomaterials from independently purified components. Crystal structures of several of the assemblies confirmed the accuracy of the design method at high resolution. Our results showcase the potential of deep learning-based methods to unlock the widespread application of designed protein-protein interfaces and self-assembling protein nanomaterials in biotechnology.

Genebe.net: Implementation and validation of an automatic ACMG variant pathogenicity criteria assignment.

We present GeneBe, an online platform streamlining the automated application of American College of Medical Genetics and Genomics (ACMG), Association for Molecular Pathology (AMP), and the College of American Pathologists (CAP) criteria for assessment of pathogenicity of genetic variants. GeneBe utilizes automated algorithms that evaluate 17 criteria from 28, closely aligning with current guidelines and leveraging data from diverse sources, including ClinVar. The user-friendly web interface enables manual refinement of assignments for specific criteria based on site-collected data. Our algorithm demonstrates a high correlation (r = 0.90) of assigned pathogenicity scores compared to expert assessments from the ClinGen Evidence Repository and substantial concordance with ClinVar verdict assignments (κ = 0.69). Comparative analysis with other published tools reveals that GeneBe performs similarly to VarSome while being superior over TAPES and InterVar. In contrast to some other tools, GeneBe's web implementation is tracker-free and third-party request-free, safeguarding user privacy. Additionally, GeneBe offers an Application Programming Interface (API) for enhanced flexibility and integration into existing workflows and is provided free of charge for research purposes. GeneBe is available at https://genebe.net.

Pneumonia prediction on chest x-ray images using deep learning approach

<span lang="EN-US">Coronavirus disease 2019 (COVID-19) is an infectious disease with first symptoms similar to the flu. In many cases, this disease causes pneumonia. Since pulmonary infections can be observed through radiography images, this paper investigates deep learning methods for automatically analyzing query chest x-ray images. In deep learning, computers can automatically identify useful features for the model, directly from the raw data, bypassing the difficult step of manual information refinement. The main part of the deep learning method is the focus on automatically learning data representations. Visual geometry group 16 (VGG16) and DenseNet121 are methods in deep learning. The data used is a chest x-ray of pneumonia. Data is divided into training, testing, and validation. The best method for this research case is VGG16 with 93% accuracy training and 90% accuracy testing. In this study, DenseNet121 obtained accuracy below VGG16, with 92% accuracy in training and 88% for accuracy testing. Parameters have a significant influence on the accuracy of each model, and with the parameters that have been used, the VGG16 is a method that has high accuracy and can be used to predict chest x-ray images aimed at checking pneumonia in patients.</span>

Simultaneous Automatic Picking and Manual Picking Refinement for First-Break

Simultaneous Automatic Picking and Manual Picking Refinement for First-Break