Gun Bhakdisongkhram

Dr. Gun Bhakdisongkhram


Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 December 2021 in IEEE Transactions on Biomedical Engineering
Verified
Journal Article
Closed access
PAPhairot Autthasan
Phairot Autthasan
RCRattanaphon Chaisaen
Rattanaphon Chaisaen
TSThapanun Sudhawiyangkul
Thapanun Sudhawiyangkul
PRPhurin Rangpong
Phurin Rangpong
SKSuktipol Kiatthaveephong
Suktipol Kiatthaveephong
TWTheerawit Wilaiprasitporn
Theerawit Wilaiprasitporn
Objective: Advances in the motor imagery (MI)-based brain-computer interfaces (BCIs) allow control of several applications by decoding neurophysiological phenomena, which are usually recorded by electroencephalography (EEG) using a non-invasive technique. Despite significant advances in MI-based BCI, EEG rhythms are specific to a subject and various changes over time. These issues point to significant challenges to enhance the classification performance, especially in a subject-independent manner. Methods: To overcome these challenges, we propose MIN2Net, a novel end-to-end multi-task learning to tackle this task. We integrate deep metric learning into a multi-task autoencoder to learn a compact and discriminative latent representation from EEG and perform classification simultaneously. Results: This approach reduces the complexity in pre-processing, results in significant performance improvement on EEG classification. Experimental results in a subject-independent manner show that MIN2Net outperforms the state-of-the-art techniques, achieving an F1-score improvement of 6.72% and 2.23% on the SMR-BCI and OpenBMI datasets, respectively. Conclusion: We demonstrate that MIN2Net improves discriminative information in the latent representation. Significance: This study indicates the possibility and practicality of using this model to develop MI-based BCI applications for new users without calibration.

1 March 2023 in Journal of Image and Graphics
Verified
Journal Article
Open access
Sarunya KanjanawattanaSarunya Kanjanawattana
Sarunya Kanjanawattana
WTWorawit Teerawatthanaprapha
Worawit Teerawatthanaprapha
PPPanchalee Praneetpholkrang
Panchalee Praneetpholkrang
GBGun Bhakdisongkhram
Gun Bhakdisongkhram
SWSuchada Weeragulpiriya
Suchada Weeragulpiriya
In Thailand, the pineapple is a valuable crop whose price is determined by its sweetness. An optical refractometer or another technique that requires expert judgment can be used to determine a fruit's sweetness. Furthermore, determining the sweetness of each fruit takes time and effort. This study employed the Alexnet deep learning model to categorize pineapple sweetness levels based on physical attributes shown in images. The dataset was classified into four classes, i.e., M1 to M4, and sorted in ascending order by sweetness level. The dataset was divided into two parts: training and testing datasets. Training accounted for 80% of the dataset while testing accounted for 20%. This study's experiments were repeated five times, each with a distinct epoch and working with data that had been prepared. According to the experiment, the Alexnet model produced the greatest results when trained with balancing data across 10 epochs and 120 figures per class. The model's accuracy and F1 score were 91.78% and 92.31%, respectively.

13 January 2025 in JMIR Medical Education
Verified
Journal Article
Open access
NKNaritsaret Kaewboonlert
Naritsaret Kaewboonlert
JPJiraphon Poontananggul
Jiraphon Poontananggul
NPNatthipong Pongsuwan
Natthipong Pongsuwan
GBGun Bhakdisongkhram
Gun Bhakdisongkhram
Background: Artificial intelligence (AI) has become widely applied across many fields, including medical education. Content validation and its answers are based on training datasets and the optimization of each model. The accuracy of large language model (LLMs) in basic medical examinations and factors related to their accuracy have also been explored. Objective: We evaluated factors associated with the accuracy of LLMs (GPT-3.5, GPT-4, Google Bard, and Microsoft Bing) in answering multiple-choice questions from basic medical science examinations. Methods: We used questions that were closely aligned with the content and topic distribution of Thailand’s Step 1 National Medical Licensing Examination. Variables such as the difficulty index, discrimination index, and question characteristics were collected. These questions were then simultaneously input into ChatGPT (with GPT-3.5 and GPT-4), Microsoft Bing, and Google Bard, and their responses were recorded. The accuracy of these LLMs and the associated factors were analyzed using multivariable logistic regression. This analysis aimed to assess the effect of various factors on model accuracy, with results reported as odds ratios (ORs). Results: The study revealed that GPT-4 was the top-performing model, with an overall accuracy of 89.07% (95% CI 84.76%‐92.41%), significantly outperforming the others (P<.001). Microsoft Bing followed with an accuracy of 83.69% (95% CI 78.85%‐87.80%), GPT-3.5 at 67.02% (95% CI 61.20%‐72.48%), and Google Bard at 63.83% (95% CI 57.92%‐69.44%). The multivariable logistic regression analysis showed a correlation between question difficulty and model performance, with GPT-4 demonstrating the strongest association. Interestingly, no significant correlation was found between model accuracy and question length, negative wording, clinical scenarios, or the discrimination index for most models, except for Google Bard, which showed varying correlations. Conclusions: The GPT-4 and Microsoft Bing models demonstrated equal and superior accuracy compared to GPT-3.5 and Google Bard in the domain of basic medical science. The accuracy of these models was significantly influenced by the item’s difficulty index, indicating that the LLMs are more accurate when answering easier questions. This suggests that the more accurate models, such as GPT-4 and Bing, can be valuable tools for understanding and learning basic medical science concepts.

15 May 2024
Verified
Conference Paper
Open access
ISIsaman Sangbumrung
Isaman Sangbumrung
Sarunya KanjanawattanaSarunya Kanjanawattana
Sarunya Kanjanawattana
GBGun Bhakdisongkhram
Gun Bhakdisongkhram
Electroencephalography (EEG) is a crucial tool for evaluation and planning rehabilitation of patients with movement disorders. However, accurate interpretation of complex EEG data requires advanced analytical skills and extensive experience to avoid potential misdirection in treatment procedures. Artificial intelligence (AI) offers a promising solution to this challenge. This study aimed to determine the optimal number of electrodes and electrode patterns for classifying motor imagery (MI) using EEGNet. A conceptual framework was developed, and significant electrode positions on the EEG headset were identified. Through careful analysis and elimination of outlier subjects, various six-, eight-, and ten-electrode patterns were evaluated. The ten-electrode pattern, specifically 10E-P4, achieved the highest accuracy (0.782) among all configurations. This pattern, which covers a substantial area of the central brain region, allows for improved detection of brainwave signals associated with MI as opposed to those related to actual body movements. Consequently, the 10E-P4 pattern significantly enhances the accuracy and effectiveness of MI classification, offering valuable insights for researchers and developers in the field of brain-computer interfaces and related applications. Future research may focus on further optimization and the development of more advanced algorithms to improve classification performance using this optimal electrode arrangement.

18 June 2026 in Computers
Verified
Journal Article
Open access
PDF file
Sarunya KanjanawattanaSarunya Kanjanawattana
Sarunya Kanjanawattana
ISIsaman Sangbumrung
Isaman Sangbumrung
DWDulyawat Wiriyaphong
Dulyawat Wiriyaphong
GBGun Bhakdisongkhram
Gun Bhakdisongkhram
An electroencephalogram (EEG) signals provide vital insights for stroke rehabilitation, yet analyzing these complex, high-dimensional data to detect gait anomalies remains challenging. Artificial intelligence offers a promising solution to precisely identify abnormal movements, assisting physicians in optimizing personalized treatments. This exploratory pilot study aims to evaluate multi-class deep learning frameworks for classifying eight distinct normal and abnormal motor activities in stroke patients using EEG data. EEG signals from eight stroke patients were utilized to train and evaluate a customized Convolutional Neural Network (CNN), DeepConvNet, and EEGNet. Furthermore, channel reduction configurations (32, 22, and 15 channels) were investigated to determine optimal clinical setups. In the Leave-One-Out Cross-Validation (LOOCV) evaluation involving seven patients, EEGNet attained the highest descriptive average F1-score of 0.810. Moreover, when assessed independently on an unseen patient, it achieved an F1-score of 0.915, indicating its potential in accommodating individual differences within this limited cohort. Moreover, EEGNet exhibited a low false positive rate of 0.175, minimizing false alarms. While the 32-channel setup yielded the highest consistency, reduced configurations served as hypothesis-generating for specific tasks. In conclusion, EEGNet demonstrated superior average performance in differentiating complicated gait patterns in this exploratory pilot study, underscoring its promise for real-time, non-invasive monitoring in stroke neurorehabilitation.

15 May 2026 in IEEE Sensors Journal
Verified
Journal Article
Closed access
WPWipamas Polpakdee
Wipamas Polpakdee
PAPhairot Autthasan
Phairot Autthasan
RCRattanaphon Chaisaen
Rattanaphon Chaisaen
PTPhattanun Thabarsa
Phattanun Thabarsa
SKSupavit Kongwudhikunakorn
Supavit Kongwudhikunakorn
WSWanumaidah Saengmolee
Wanumaidah Saengmolee