Training Model Research Articles

A Deep Learning-Based Method for Rapid 3D Whole-Heart Modeling in Congenital Heart Disease.

This study aims to develop a deep learning-based method for generating three-dimensional heart mesh models for patients with congenital heart disease by integrating medical imaging and clinical diagnostic information. A deep learning model was trained using CT and cardiac MRI (CMR) images, along with clinical data from 110 patients. The web-based platform automatically outputs STL files for 3D printing and Unity 3D OBJ files for virtual reality (VR) applications upon uploading the medical images and diagnostic information. The models were tested on three congenital heart disease cases, with corresponding 3D printed and VR heart models generated. The 3D printed and VR heart models received high praise from professional doctors for their anatomical accuracy and clarity. Evaluations indicated that the proposed method effectively and rapidly reconstructs complex congenital heart disease structures, proving useful for preoperative planning and diagnostic support. The 3D modeling approach has the potential to enhance the precision of surgical planning and diagnosis for congenital heart disease. Future studies should explore larger datasets and training models for different types of congenital heart disease to validate the model's broad applicability.

Intelligent diagnostic method for beam pumping units based on torque indicator cards

An innovative approach is introduced to the intelligent diagnosis of torque indicator card conditions using electric parameters, effectively addressing the high costs associated with acquiring polished rod indicator cards and the convergence difficulties encountered when converting electrical parameters. The proposed method begins by establishing a library of 15 typical operating conditions, followed by an engineering feature analysis. Key to this approach is the use of the AlexNet model for intelligent recognition training, which successfully enhances model convergence and stability, even with varying sample sizes. This allows for automated recognition and classification of operating conditions through model training. The research outcomes demonstrate that the model achieved a fault diagnosis accuracy of over 95% for beam pumping units. This not only improves the accuracy and efficiency of fault diagnosis in oil well pumping units but also presents a novel and effective approach for intelligent diagnostic applications in oil field extraction, offering substantial practical value.

Neuroendoscopy Training

AbstractNeuroendoscopy can be learnt by assisting or doing live human surgery, cadaver dissection with or without augmented pulsatile vessel and cerebrospinal fluid (CSF) perfusion, and practicing on live animal, dead animal model, synthetic models, three-dimensional printing model with or without augmentation with animal, cadaver tissue, pulsatile vessel and reconstructed CSF model, virtual reality (VR) simulator, and hybrid simulators (combined physical model and VR model). Neurosurgery skill laboratory with basic and advanced learning should be there in all teaching hospitals. Skills can be transferred from simulation model or VR to cadaver to live surgery. Staged learning (first with simple model to learn basic endoscopic technique, then animal model, and then augmented cadavers) is the preferred method of learning. Although most surveys favor live surgery and practice on animal models and cadavers as the most preferred training model now, in future VR may also become a favored method of learning. This article is based on our experience in over 10,000 neuroendoscopic surgeries, and feedback from over 950 neuroendoscopic fellows or consultants who attended workshops conducted every 6 monthly since 2010. A literature search was done on PubMed and Google Scholar using (neuroendoscopy) AND (learning), and (neuroendoscopy) AND (training), which resulted in 121 and 213 results, respectively. Out of them, 77 articles were finally selected for this article. Most of the training programs typically focus on microneurosurgical training. There is lack of learning facilities for neuroendoscopy in most centers. Learning of neuroendoscopy differs greatly from microneurosurgery; switching from microneurosurgery to neuroendoscopy can be challenging. Postgraduate training centers should have well-equipped neuroendoscopy skill laboratory and the surgical educational curriculum should include neuroendoscopy training. Learning endoscopy is about taking advantages of the technique and overcoming the limitations of endoscopy by continuous training.

The use of the objective structured clinical examination to evaluate paediatric cardiopulmonary resuscitation skills in medical students and measures to improve training

BackgroundThe main objectives of the study were to analyse the use of the Objective Structured Clinical Examination (OSCE) to evaluate the skills of medical students in paediatric basic life support (PBLS), to compare two resuscitation training models and to evaluate the measures to improve the teaching program.MethodsComparative, prospective, observation study with intervention in two hospitals, one undergoing a PILS course (Paediatric Immediate Life Support) and another PBLS. The study was performed over three phases. 1º. PBLS OSCE in 2022 three months after the resuscitation training 2. Measures to improve the training program in 2023 3. PBLS OSCE in 2023. Overall results were analysed and comparison between both sites and those for 2022 and 2023 were made.ResultsA total of 210 and 182 students took part in the OSCE in 2022 and 2023, respectively. The overall mean score out of 100 was 83.2 (19), 77.8 (19.8) in 2022 and 89.5 (15.9) and 2023, P < 0.001. Overall cardiopulmonary resuscitation (CPR) effectiveness was adequate in 79.4% and 84.6% of students in 2022 and 2023, respectively. The results of hospital students undergoing a PILS course were better (86.4 (16.6) than those undergoing a PBLS. 80.2 (20.6) p < 0.001. The results from both hospitals improved significantly in 2023.ConclusionsThe OSCE is a valid instrument to evaluate PBLS skills in medical students and to compare different training methods and program improvements. Medical students who receive a PILS attain better PBLS skills than those who undergo a PBLS course.

The development of an intradisciplinary staff training intervention on the optimal management of behavioural and psychological symptoms of dementia: A qualitative study

Introduction: Behavioural and psychological symptoms of dementia (BPSDs) are a group of highly prevalent symptoms in residents with a neurocognitive disorder (NCD), including agitation and depressive symptoms. Caregiving staff often mention knowledge and skills deficits regarding optimal BPSD management. While several staff trainings on BPSDs management exist internationally, their transferability to other clinical contexts is limited, owing to methodological challenges. Therefore, to address this implementation gap, there is a strong need for trainings based on high-quality research to strengthen existing evidence, ensure feasibility and reproducibility. Methods: This qualitative study, part of a larger research project, occurred in 2022 on a long-term care (LTC) center unit and an alternate level of care (ALT) hospital unit located in the Province of Quebec, Canada. This study aimed to: 1) evaluate the needs and perceptions of staff caregivers regarding BPSD management, 2) identify the training content and modalities to prioritize according to experts, 3) develop an intradisciplinary training on BPSD management, and 4) pretest the preliminarily version of the training. Objectives one and two were evaluated using focus groups and objective four using individual cognitive interviews. Qualitative interview data was audio-recorded with participants’ consent, transcribed verbatim, and thematically content analyzed using an established method. Results: Overall, thirteen caregivers participated in objective one (8 from the LTC unit, and 6 from the ALT unit). The main staff training need identified on each site was the management of resident’s aggressive behaviours. Staff verbalized a preference for a virtual training. Objective 2 involved a panel of eight experts. Experts recommended the use of an online training platform, and certain training models and indicators. Based on caregivers and experts’ input, five interactive online staff training capsules lasting from 20-25 minutes each and an algorithm guiding the evaluation and management of BPSDs were created. The training capsules and the algorithm were iteratively improved following cognitive interviews with 4 caregivers from the participating sites. Conclusion: An interactive virtual staff training on BPSD management was created based on staff and expert consultation. The next step in investigation will be to evaluate the feasibility and acceptability of the training.

Методы оптимизации процесса обучения и тонкой настройки больших языковых моделей

The main problematic issues in the development and specialization of LLM are: catastrophic forgetting, the risk of overfitting, hallucinations, incorrect interpretations, incorrect processing of exceptional situations as well as exceptionally high performance requirements for the computing tools used in this case. The purpose of the study is to select and develop methods for optimizing the training and fine-tuning process LLM, providing a significant reduction in the computing resources required for this. To achieve this goal, it is proposed to use the following methods of optimizing LLMs and their learning algorithms: LoRA and QLoRA, Batch size choice, Gradient Accumulation, Gradient Checkpoint, Mixed precision training, FlashAttention-2. To obtain a cumulative positive effect when using these methods together, it is necessary to perform a number of practical experiments. When setting up LLM learning hyperparameters, you should first determine which package size gives the best results, and then choose adequate methods to optimize the computing resources used. The application of the presented methods will increase the efficiency of using computing resources when training and fine-tuning large language models and will reduce the time and financial costs necessary for this.

Rapid discrimination and authentication of Lushan Yunwu tea with different harvest periods based on Nano-ESI-MS combined with chemometric analysis

The economic value and flavor of tea with different harvest periods vary greatly. Therefore, rapid discrimination and authentication are important for quality control and commercialization of tea. Nano-electrospray ionization mass spectrometry (Nano-ESI-MS) is an emerging technique in food authenticity and traceability due to the minimal sample preparation required and short analysis time. In this study, Nano-ESI-MS method was established for rapid discrimination and authentication of Lushan Yunwu tea (LYT). The results showed that Nano-ESI-MS could not only effectively distinguishing LYT of different harvest periods but also differentiating Mingqian tea adulterated with 5% Yuhou tea and Autumn tea, respectively, with 58 and 17 compounds were considered as markers, respectively. The decision tree and linear discriminant analysis were used to identify the adulteration of tea with classification accuracy of models of training and validation sets were higher than 99.2% and 86.9%, respectively. Moreover, Nano-ESI-MS was used to track the changes of these compounds during tea processing, including amino acids, caffeine, and catechins. The Nano-ESI-MS method could be regarded as a powerful analytical tool for rapid discrimination and authentication of tea.

A Federated Transfer Fault Diagnosis Method for Cross-Domain and Incomplete Data

In fact, multiple clients often use similar devices and collect fault data separately, so joint multi-client collaborative fault diagnosis modeling can solve the problem of data scarcity, but this poses great challenges to data privacy protection. In this paper, we propose a federated transfer fault diagnosis method based on federated learning for cross-domain incomplete data. The proposed method only exchanges the parameters of the local training model, which achieves the privacy protection of the client’s local data. We construct a multi-client collaborative learning framework to address the problem of weak generalization ability caused by the lack of terms in single client training samples. We also propose a targeted semi-supervised fine-tuning strategy based on relative distance to reduce the probability of negative fine-tuning of out-of-distribution samples and improve the accuracy of diagnostic models. The results of cross-condition and cross-equipment experiments demonstrate that the proposed method has obvious advantages over the existing fault diagnosis methods.

Training Model for Extended Career Athletes: A Narrative Review.

Today's elite and professional sports tend to feature older, more seasoned athletes, who have longer sporting careers. As advancing age can potentially limit peak performance, balancing training load is necessary to maintain an optimal state of performance and extend their sports career. To describe an appropriate training model for extended career athletes. Medline (PubMed), SPORTDiscus, ScienceDirect, Web of Science, and Google Scholar. A search of the literature between January 1, 2015 and November 22, 2023 was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Narrative review. Level 4. Data were extracted from studies related to the management of training and performance of athletes with extended and long careers. A total of 21 articles related to extended careers were found. Key themes from these papers included: expertise, biological maturation, and specificity; epidemiology and health; athlete monitoring; strength training; load management and detraining; success management. A training model for extended career athletes should balance the deleterious effects of age with the athletes' knowledge of, and expertise within, the sport. Designing specific training that accommodates previous injuries, training load intolerances, and caters for quality of life after retirement should be key considerations. Load management strategies for athletes with extended careers should include strength training adaptations to minimize pain, load-response monitoring, a broad range of movement, recovery and intensity activities, and the avoidance of large training load peaks and periods of inactivity.

Two-Stage Optimization Model Based on Neo4j-Dueling Deep Q Network

To alleviate the power flow congestion in active distribution networks (ADNs), this paper proposes a two-stage load transfer optimization model based on Neo4j-Dueling DQN. First, the Neo4j graph model was established as the training environment for Dueling DQN. Meanwhile, the power supply paths from the congestion point to the power source point were obtained using the Cypher language built into Neo4j, forming a load transfer space that served as the action space. Secondly, based on various constraints in the load transfer process, a reward and penalty function was formulated to establish the Dueling DQN training model. Finally, according to the ε−greedy action selection strategy, actions were selected from the action space and interacted with the Neo4j environment, resulting in the optimal load transfer operation sequence. In this paper, Python was used as the programming language, TensorFlow open-source software library was used to form a deep reinforcement network, and Py2neo toolkit was used to complete the linkage between the python platform and Neo4j. We conducted experiments on a real 79-node system, using three power flow congestion scenarios for validation. Under the three power flow congestion scenarios, the time required to obtain the results was 2.87 s, 4.37 s and 3.45 s, respectively. For scenario 1 before and after load transfer, the line loss, voltage deviation and line load rate were reduced by about 56.0%, 76.0% and 55.7%, respectively. For scenario 2 before and after load transfer, the line loss, voltage deviation and line load rate were reduced by 41.7%, 72.9% and 56.7%, respectively. For scenario 3 before and after load transfer, the line loss, voltage deviation and line load rate were reduced by 13.6%, 47.1% and 37.7%, respectively. The experimental results show that the trained model can quickly and accurately derive the optimal load transfer operation sequence under different power flow congestion conditions, thereby validating the effectiveness of the proposed model.

Development of an endurance training model in athletic performance: a case study on long distance running in Indonesia

This research aims to create innovative and effective endurance training materials tailored for long-distance runners aged 15-18. Utilizing the research and development (RnD) approach with the Borg and Gall model, the study incorporates feedback from expert validators and athletes along with their coaches from Sorong, West Papua. The research procedure encompasses ten stages, starting with problem identification and culminating in the implementation of the final product, involving 80 athletes. These stages include initial analysis, planning, initial product development, preliminary field testing, primary product revision, main field testing, operational product revision, operational field testing, final product revision, and dissemination and implementation. The involvement of both experts and practitioners ensures the training materials are theoretically sound and practically applicable. Data analysis was conducted using SPSS version 21, employing a normality test, a homogeneity test, and a paired-sample t-test to ensure the reliability and validity of the findings. The normality test confirmed the data followed a normal distribution, while the homogeneity test ensured the sample groups were comparable. The paired-sample t-test results indicated the newly developed book-based endurance training model significantly enhanced athletes' performance (p-value 0.000), showing a more substantial improvement in training outcomes compared to conventional methods (p-value 0.014). This improvement is attributed to the structured and comprehensive nature of the training materials, which include interactive elements and technology integration, making the training process more engaging and effective for young athletes. In conclusion, adopting interactive and technology-based training models can significantly aid in developing endurance skills in young long-distance runners, addressing specific needs and optimizing their training regimen. Keywords: Athletics, Endurance Training, Exercise Media, Physical Education

Determining the chemical ordering in nanoalloys by considering atomic coordination types.

The energetically most favorable chemical ordering of bimetallic nanoparticles can be characterized by combining global optimization algorithms and surrogate energy models. The latter approximate the energy of nanoalloys relying on structural descriptors, training models, and data. Here, we systematically evaluate the performance of highly data-efficient topological descriptors [Kozlov et al., Chem. Sci. 6, 3868 (2015)] for predicting the energies of metal nanoalloys with different chemical orderings. We also introduce a new descriptor based on atomic coordination types, which results in a less data-efficient and interpretable approach, but improves the general accuracy and the quantification of orderings in the inner parts of nanoparticles. The capacity of both the original and new approaches in combination with a basin hopping algorithm is illustrated by generating convex hulls of PdZn nanoalloys and predicting the resulting active surface site distribution as a function of particle composition. Finally, we show how these approaches can be combined with machine-learning adsorption models in electrocatalysis studies for a fast evaluation of the reactivity landscape of targeted nanoalloys.

Life-threatening event in laparoscopic hepatic surgery: Training curriculum on sudden hepatic artery haemorrhage

BackgroundExposure of the hepatic artery is a fundamental step in many surgeries, during which iatrogenic hepatic artery injury may occur. Although the incidence of hepatic artery haemorrhage is low, its occurrence can lead to life-threatening haemorrhage. It is difficult and dangerous to accumulate clinical experience in laparoscopic hepatic artery repair in actual patients, and simulation training models for laparoscopic hepatic artery repair are currently lacking. In this study, a 3D printed model was designed to simulate the training curriculum for sudden hepatic artery haemorrhage, but whether training with the 3D printed model could yield superior skill improvement for surgeons remained to be determined.MethodsA new 3D printed model was designed for this study. Surgeons from the General Surgery Department of Sir Run Run Shaw Hospital participated in this simulation training. The surgical performance of each model was compared, and the authenticity of the model was evaluated and mechanically tested.ResultsExperienced surgeons performed better on the 3D printed model. After repeated training, inexperienced surgeons showed significant improvement of their laparoscopic hepatic artery repair skills. The authenticity of the model was generally satisfactory, but shortcomings persisted in the mechanical testing of artery wall tearing, necessitating further improvement.ConclusionsFew studies have investigated laparoscopic simulation training for sudden hepatic artery haemorrhage. This simulation model distinguishes surgeons with different levels of experience and allows those with less experience to improve their laparoscopic hepatic artery repair skills through training on the model.

Construction of knowledge sharing network indicator system for medication therapy management service training teams based on social network analysis

BackgroundBased on the perspective of social network theory, this study explored the network indicator system that facilitated optimal knowledge sharing effect in Medication Therapy Management Services (MTMS) training teams. The aim was to provide a reference for optimizing MTMS training and improving training quality.MethodsUtilizing social network analysis combined with a questionnaire survey, a knowledge sharing matrix for MTMS training teams was constructed. Knowledge sharing behavior was assessed from three perspectives: individual networks, whole networks, and cohesive subgroups.ResultsIndividual network analysis showed that the knowledge sharing effect within the training team reached its peak when the out-degree centrality was ≥ 3.5, in-degree centrality was ≥ 2.5, eigenvector centrality was ≥ 0.065, and closeness centrality was ≥ 7.86. Whole network analysis indicated that the optimal knowledge sharing effect occurred when the network density of the training team was higher than 0.0343 and the training size was less than 117 participants. Cohesion subgroups analysis demonstrated that knowledge sharing was more effective when members with similar working years participated in training together.ConclusionsThe knowledge sharing indicator system developed for MTMS training teams, based on social network analysis, can assist in optimizing the MTMS training model and improving training effectiveness.

Perspectives of an NCI T32 Training Program Designed to Train a Diverse Workforce in Cancer Health Equity Research.

With cancer health disparities on the rise in the United States (USA), there is an increased need for novel approaches to address these challenges. One such approach that may help address these disparities is increasing diversity in the biomedical research workforce. The Cancer Health Equity and Career Development Program (CHECDP) embodies this approach by recruiting and training underrepresented minorities in cancer research to develop the skills and training needed to be competitive for independent research careers, thus diversifying the biomedical research workforce. The training model that CHECDP employs is unique with its funding through the NCI training mechanism, its strong institutional support, and its participant-driven curriculum. The curriculum includes educational, career, and leadership opportunities that are continuously evaluated for sustained impact. The program has been comprised of mostly under-represented minorities that have been propelled to independent careers with a high rate of funded career development awards. Our T32 program serves as a model of success for other programs seeking to diversify the biomedical research workforce and reduce cancer health disparities.

Educational model and research levels in the professional training of lawyers

Educational model and research levels in the professional training of lawyers

Low-complexity short-time acoustic scene classification based on causal convolutions

Acoustic scene classification is vital for building smart cities. Most acoustic scene classification studies are performed by training convolutional neural networks. The networks constructed based on the general convolutional approach need to consider the temporal-frequency connections in the audio spectrogram. This work proposes a bottleneck structure neural network model for acoustic scene classification based on causal convolution and Sub-spectral normalization to overcome the abovementioned problem. To obtain more sample data, a new mix-spec augmentation data enhancement method is proposed by drawing on the existing classical mixup and spec augmentation methods, and the three are combined for sample augmentation. It was experimentally found that the proposed bottleneck structure helps improve the recognition accuracy and robustness of the model. In addition, the ablation experiment showed that the data augmentation training model combined with the mixup, SpecAugment, and the newly proposed mix-SpecAugment method had the highest classification accuracy. Finally, the near-square audio spectrogram has better training results by employing different shapes of audio spectrograms for the experiments. Compared with the baseline results, it is found that our proposed model significantly improved the accuracy of the baseline and reduced loss.

CTCTime: A New Model for Unidimensional Time Series Classification

One-dimensional time series classification has always been an important research direction, playing an irreplaceable role in various fields. With the rapid development of deep learning, most traditional time series classification methods have gradually been replaced by neural network-based classification methods. At present, in the field of time series classification, models that perform well are mostly based on Convolutional Neural Networks (CNNs), while models with the Transformer architecture, which have shown outstanding performance in natural language processing and computer vision, do not stand out. In order to change this situation, We have investigated the feasibility of training models based on a single Transformer architecture directly on small datasets without additional data processing. Furthermore, we propose a new model called CTCTime, which combines the Transformer architecture with CNNs to address one-dimensional time series classification problems. We compared CTCTime with 13 traditional algorithms on 44 datasets from the UCR archive and with 7 advanced methods on 85 datasets. The UCR datasets are classic datasets for one-dimensional time series classification tasks. The experimental results demonstrate its feasibility, accuracy, and scalability.

Seating Accuracy of Prefabricated Interim Crowns on Immediate and Delayed Implants: A Laboratory Study.

Evaluation of seating accuracy of interim crowns for the immediate restoration of immediate implants (I-Imp) and delayed implants (D-Imp) placed via static computer-assisted implant surgery (sCAIS). A maxillary training model was modified by removing the central incisors and simulating fresh extraction socket in one site and healed ridge on the other site. An I-Imp was planned in the extraction socket and D-Imp was planned in the healed site. The planned implants were used to design sCAIS surgical template and interim crowns for immediate restoration of the implants. Fourteen surgical models received sCAIS implants after which the interim crowns were inserted. Subsequently, the models with the seated crowns were scanned by a laboratory scanner. The planning virtual model was superimposed against every surgical model to measure vertical, horizontal and proximal contact errors of each crown. All the crowns were positioned more incisally than the planned crowns. This was significantly more noticeable for the D-Imp crowns (0.81 mm) than the I-Imp crowns (0.55 mm). The 2 crown groups had similar horizontal errors (I-Imp &#61;0.35 mm, D-Imp &#61; 0.36 mm). The D-Imp crowns had minimal proximal contact error (0.14 mm), but the I-Imp crowns had significantly greater proximal contact error (0.74 mm) in the form of open distal contacts. This pattern of error appears related to the relationship between the socket morphology and the planned implant position. Prefabricated interim I-Imp crowns suffered from greater errors that affected the proximal contact quality than DImp crowns. The observed deviation of the I-Imp crowns can be attributed to the socket morphology and its relation to the planned implant position. The deviations of the I-Imp crowns are clinically significant and will require clinical adjustments. Thus, caution is needed before routine use of prefabricated interim crowns on I-Imp.

A systematic literature review of neuroimaging coupled with machine learning approaches for diagnosis of attention deficit hyperactivity disorder

ProblemAttention deficit hyperactivity disorder (ADHD) is the most commonly found neurodevelopmental condition among children with an estimated 2.5% to 9% global prevalence. While ADHD has been regarded as a lifelong condition, its early diagnosis elevates the probability of recovery and normal life for children. Despite clinical diagnosis being the primary one, substantial developments in ADHD diagnosis have been made during the past decade.AimSeveral imaging-based technologies and approaches have been presented in the existing literature including magnetic resonance imaging (MRI) and functional MRI (fMRI). In addition, the deployment of machine learning and deep learning models has paved the way for automated diagnosis of ADHD which increases both accuracy and robustness of ADHD detection. A comprehensive and systematic literature review (SLR) of imaging technologies and machine learning approaches is highly desired to comprehend the current status of such approaches concerning their potential and challenges to outline future directions. Although a substantial body of literature exists on imaging-based ADHD diagnosis, comprehensive SRL on such approaches is scarce. This SLR aims to provide a comprehensive overview of imaging-based ADHD diagnosis with emphasis on machine learning approaches, reveals their pros and cons, and provides potential future research directions thereby contributing to the scientific community to accelerate further research for ADHD diagnosis.MethodsThis SRL focuses on analyzing recently published studies between 2010 and 2023. For this purpose, preferred reporting items for systematic review and meta-analyses (PRISMA) approach is performed in this study. Five eminent academic databases Web of Science, ACM, Springerlink, Elsevier, and PubMed are selected for article search. The SLR follows a systematic methodology comprising article search, selection based on inclusion and exclusion criteria, and rigorous assessment for categorization.ResultsIt is found that MRI and fMRI are the dominant approaches integrated with machine learning models for ADHD detection and function near-infrared spectroscopy is also adopted by a few studies. Predominantly, the ATHENA preprocessing approach is used to preprocess MRI data before model training. Due to the public availability of the ADHD-200 dataset, it is widely used in the existing literature while a few studies utilized their self-collected datasets. Machine learning models are the choice of the majority of the studies, particularly, the support vector machines model has been widely used for ADHD detection. Feature fusion is observed to be a better choice for obtaining more accurate results.ConclusionMachine learning and deep learning models provide automated ADHD detection with better accuracy and robustness, however, such models are not generalizable and their performance varies concerning the locality of data used for experiments. In addition, the heterogeneity of data collection from various devices is a challenge, and the use of a standard device may provide better solutions. The lack of labeled data also adds difficulties to training models. Besides the use of MRI and fMRI, other novel technologies should be explored for better ADHD detection performance.