KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Search Results

Now showing 1 - 10 of 33

Some of the metrics are blocked by your
consent settings
Comparative performance of deep learning models and non-dermatologists in diagnosing psoriasis, dermatophytosis, and eczema
(2026-12-01)
Yodrabum, Nutcha
;
Wongpraparut, Chanisada
;
Titijaroonroj, Taravichet
;
Chularojanamontri, Leena
;
Bunyaratavej, Sumanas
Accurately differentiating scaly erythematous rashes among psoriasis, eczema, and dermatophytosis remains a clinical challenge, particularly for non-dermatologists. This study aimed to develop and evaluate deep learning models using macroscopic clinical images to classify these conditions and compare their performance with that of non-specialists. A total of 2940 images were sourced from public datasets, the Siriraj Dermatology databank, and newly collected images from Thai participants. Among sixteen evaluated models, the Swin demonstrated the best performance and interpretability. Gradient-weighted Class Activation Mapping (Grad-CAM) visualizations confirmed that the model focused on clinically relevant lesion features. Most importantly, in a pilot comparison, the Swin outperformed non-specialists in diagnostic accuracy. However, given the limited sample size of 30 images and 30 evaluators, these results should be interpreted as exploratory. Future studies with larger datasets and diverse clinician cohorts are warranted to confirm these findings and to support clinical integration.
Some of the metrics are blocked by your
consent settings
Enhanced forecasting of friction and cohesion of augmented unsaturated soil with nanostructured quarry fines (NQF) addition
(2026-12-01)
Kamchoom, Viroon
;
Van, Duc Bui
;
Hosseini, Shahab
;
Alimoradijazi, Mohammadreza
;
Amini-Khoshalan, Hasel
Shear strength parameters such as friction angle and cohesion are fundamental to solving geotechnical engineering problems related to slope stability, foundation design, and earthwork construction. This study presents the prediction of friction angle (Fi) and cohesion (Nc) of an unsaturated lateritic soil using three intelligent learning techniques: Support Vector Machine (SVM), Radial Basis Function (RBF), and Multilayer Perceptron (MLP), with Linear Multivariate Regression (LMR) adopted as a baseline model to evaluate agreement between input and output variables. The motivation for employing machine learning approaches stems from the limitations of complex laboratory testing and the need for reliable predictive tools that can support design and field applications. The investigated soil, classified as A-7-6 and poorly graded, exhibited coefficients of uniformity and curvature of 2.05 and 0.84, respectively. It was characterized by high plasticity and significant clay content, with a clay fraction of 23.02%, clay activity of 2, friction angle of 15°, maximum dry density of 1.84 g/cm3 at an optimum moisture content of 16.2%, and was tested under cyclic direct shear conditions. Multiple datasets were generated from varying treatment conditions and soil descriptors, forming the basis for model development. Eleven input parameters were used to predict Fi and Nc, and model performance was evaluated using Variance Accounted For (VAF), Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and coefficient of determination (R2). The results indicate that RBF and MLP outperformed SVM and LMR in both training and testing phases for predicting cohesion and friction angle, demonstrating superior generalization capability. Sensitivity analysis using the Cosine Domain Method revealed that unsaturated unit weight had the greatest influence on friction angle prediction, while clay content was the most influential parameter for cohesion. Among all models, MLP achieved the highest accuracy and overall predictive performance. Based on this optimal model, a Graphical User Interface was developed to enable users to input soil parameters and obtain rapid predictions, providing a practical tool for researchers and practitioners in geotechnical engineering.
Some of the metrics are blocked by your
consent settings
Utilizing deep learning from mobile phone photos for early detection of horizontal strabismus: a screening approach
(2026-12-01)
Chawuthai, Rathachai
;
Sermswan, Anawat
;
Boonnithititikul, Chatree
;
Hokierti, Kiatthida
;
Sermsripong, Wasawat
To develop and validate an artificial intelligence pipeline for binary screening of horizontal strabismus versus orthotropia using smartphone-acquired facial images and geometric landmark analysis. This two-stage system combines Real-Time Detection Transformer (RT-DETR) to localize nine ocular landmarks per eye across three gaze directions (left, center, right), and supervised machine learning classifiers. A feature set of five biometric ratios was derived from coordinates including the canthi, limbi, and corneal light reflexes. The model was trained on facial images from 150 participants (96 with strabismus and 54 controls). To address class imbalance and improve generalizability, Synthetic Minority Oversampling Technique (SMOTE) and 4-fold cross-validation were applied. RT-DETR achieved an intersection over union of 0.62 and a mean center-point error of 6.52 pixels in landmark localization. The Random Forest classifier achieved an accuracy of 0.95, sensitivity of 0.96, specificity of 0.94, positive predictive value of 0.97, and negative predictive value of 0.92. This study demonstrates the feasibility of combining transformer-based landmark detection with geometric ratios for strabismus screening. The framework shows high performance under controlled conditions. While the use of biometric ratios allows for feature-level inspection, further research is required to establish full clinical interpretability and performance in uncontrolled environments.
Some of the metrics are blocked by your
consent settings
Systematic evaluation of one-dimensional-to-two-dimensional near-infrared spectroscopy transformations with deep learning for quantifying coconut sap adulteration
(2026-10-01)
Lapcharoensuk, Ravipat
;
Sitorus, Agustami
Near-infrared (NIR) spectroscopy have limitations when combined with deep learning (DL) algorithms because they rely on low-dimensional datasets. Therefore, we investigated the potential of transforming one-dimensional (1D) NIR spectra into two-dimensional (2D) spectrograms using synchronous and asynchronous techniques and the continuous wavelet transform (CWT) and their effectiveness by integrating with DL for detecting adulteration in coconut sap. NIR spectra (12,500–4000 cm<sup>−1</sup>) were collected from binary mixtures (0%–100%; w /w). The performance of all DL (convolutional neural networks-CNN, AlexNet and ResNet) models was compared with that of partial least squares (PLS). The models were ranked in the mentioned order based on their performances: 2D-CWT ' 2D-asynchronous ' 2D-synchronous ' 1D/2D-PLS. The important features of the best model can be explained and visualized using gradient-weighted-class-activation-mapping. The findings highlight that the 1D-to-2D NIR data transformation combined with DL is a highly robust approach because it addresses the feature representation gap in NIR data and effectively captures the spatial–spectral correlations.
Some of the metrics are blocked by your
consent settings
Hybrid-AI-sep: A multi-agent computer-aided tool for separation process problems solving and learning
(2026-06-01)
Prasopsanti, Kris
;
Yadbantung, Rungroj
;
Phanusupawimol, Thunyaras
;
Areerat, Surat
;
Mansouri, Seyed Soheil
AbstractThis paper presents a multi-agent based and artificial intelligence augmented computer-aided software tool, Hybrid-AI-sep, with problem solving and education modules for the important topic of design of separation operations. Three types of agents are employed by Hybrid-AI-sep, a library of knowledge-tools containing theory, concepts, and glossary terms related to separation operations; a library of database tools consisting of different types of measured data; and a library of computational tools that generate problem specific data that may be needed for decision making and explanation for the problem solution and educational modules. The paper presents software architecture together with examples of the three types of agents. Illustrative examples highlighting various features of Hybrid-AI-sep are given in the main manuscript and the supplementary material.
Some of the metrics are blocked by your
consent settings
Deep learning-based comparative evaluation of EEG, HRV, and EDA biomarkers for personal thermal comfort prediction
(2026-03-01)
Sahoh, Bukhoree
;
Wongsontham, Fatimah
;
Tipsavak, Apaporn
;
Chaithong, Paweena
;
Kliangkhlao, Mallika
Personal thermal comfort prediction significantly impacts health, well-being, and productivity, yet existing systems typically employ multiple physiological biomarkers without clear evidence for optimal single-modality solutions. This study presents a deep learning (DL)-based metrologically grounded framework to evaluate and compare three biomarkers—electroencephalography (EEG), heart rate variability (HRV), and electrodermal activity (EDA)—as traceable sensors under precisely controlled thermal conditions. We developed specialized signal preprocessing and feature engineering pipelines for each modality: 1) EEG spectral decomposition via Welch's power spectral density estimation across frequency bands (delta, theta, alpha, beta, and gamma); 2) HRV analysis through Fast Fourier Transform spectral estimation focusing on low-frequency to high-frequency component ratios; and 3) EDA feature extraction utilizing optimized filtering techniques to isolate skin conductance level and response characteristics. Three biomarker-specific DL architectures undergo Bayesian hyperparameter optimization, enabling equitable comparison of each modality's predictive performance. Results demonstrate that the HRV-based model achieves superior performance (F-measure = 0.95) while requiring minimal computational resources (0.78 MB memory footprint, 0.89 relative cost compared to the EEG baseline), establishing it as the optimal single-modality solution. These findings provide a paradigm for reproducible physiological measurement systems, advancing thermal comfort prediction by combining clinical-grade reliability with practical implementation benefits for clinical applications and next-generation indoor environmental control systems.
Some of the metrics are blocked by your
consent settings
Exploring the potential of AI in tourism industry: combination TAM and innovation-decision process
(2026-01-01)
Nookhao, Saowakhon
;
Homsud, Noppanon
The rapid advancement of artificial intelligence (AI) has introduced transformative opportunities into the global tourism industry. However, the adoption of AI-driven services in tourism remains relatively limited, particularly in emerging markets such as Thailand. This study explores the key factors influencing the acceptance and adoption of AI-powered tourism services by integrating the Technology Acceptance Model (TAM) with the Innovation-Decision Process (IDP). A conceptual framework consisting of eight constructs–trust in AI, perceived ease of use, perceived usefulness, perceived value, perceived enjoyment, attitude towards use AI, intention to use, and actual use–was developed and empirically validated. Data were collected from 462 Thai tourists using a structured questionnaire and analyzed using partial least squares structural equation modeling (PLS-SEM). The results revealed that trust in AI and perceived ease of use are significant antecedents of perceived usefulness and attitude towards use AI, whereas perceived value and enjoyment positively influence attitude towards use AI. Attitude towards use AI and perceived usefulness subsequently affect intention to use, which in turn drives actual use behavior. By aligning TAM constructs with the corresponding stages of the IDP, this study contributes a novel framework for understanding the diffusion of AI innovations in tourism. These findings offer theoretical insights and practical implications for stakeholders aiming to promote sustainable and technology-enhanced tourism experiences in the next normal era.
Some of the metrics are blocked by your
consent settings
Developing a Conceptual Framework UTAUT-TTF-TOE Model to Examine AI’s Influence on HR Performance in Higher Education
(2026-01-01)
Luomou
;
Chaveesuk, Singha
;
Chaiyasoonthorn, Wornchanok
;
Kamales, Nayika
The swift advancement of artificial intelligence (AI) is exerting a profound influence on the operational paradigm of higher education, yet the degree of its implementation exhibits substantial disparities among institutions. Presently, there is a dearth of systematic research on the moderating mechanisms of AI adoption intensity on organizational performance. This research integrates the Unified Theory of Acceptance and Use of Technology (UTAUT), task-technology fit (TTF), and the technology-organization-environment (TOE) framework to formulate a comprehensive model for exploring the mechanisms through which AI impacts human resource performance in universities. The research centers on analyzing how technological, organizational, and environmental factors interact to affect AI adoption, specifically investigating the moderating function of adoption intensity in this relationship. This research endeavors to uncover the specific pathways and boundary conditions for AI to improve university management efficiency.
Some of the metrics are blocked by your
consent settings
Investigation of physiological disorder classification in mangosteen fruit using visible and shortwave near-infrared spectroscopy combined with machine learning
(2025-12-01)
Ruttanadech, Nuttapong
;
Momin, Abdul
;
Phetpan, Kittisak
;
Chaichanyut, Montree
;
Thongphut, Chitwadee
Accurate classification of physiological disorders in mangosteen fruit is crucial for ensuring production quality, safety, sustainability, and economic viability. This study investigates the application of visible and shortwave near-infrared (Vis/SWNIR) reflectance spectroscopy, combined with machine learning algorithms, to classify three primary disorders: normal fruit (NF), translucent flesh disorder (TFD), and TFD with yellow gummy latex (TFD & YGL). The study specifically examines the effects of light intensity, spectral pretreatments, and machine learning models on classification performance. Spectral data were collected using two light intensities (50 % and 100 % of a 150 W light source) and processed with three pretreatments: standard normal variate (SNV), second derivative Savitzky-Golay (SGD2), and a combination of SNV and SGD2. Random forest (RF), support vector machine (SVM), and multi-layer perceptron (MLP) algorithms were used for classification. The SGD2 method improved differentiation, especially for the TFD & YGL class, in the 700–725 nm wavelength range, which is associated with xanthone content in the fruit's pericarp. Higher light intensity (100 %) significantly improved classification accuracy, achieving an overall accuracy of 0.71 and an average F1 score of 0.61 with the RF model. Despite these improvements, the model struggled to distinguish the TFD class from NF due to their similar spectral profiles. Overall, the Vis/SWNIR spectroscopy and machine learning combination shows strong potential for the non-destructive classification of mangosteen fruit disorders. Both light intensity and spectral pretreatments play critical roles in enhancing performance. Future studies should focus on improving spectral sensitivity to better capture internal fruit characteristics.
Some of the metrics are blocked by your
consent settings
Forecasting maize yield from growth parameters using machine learning in a biochar-inorganic fertilizer amended soil under drip irrigation
(2025-12-01)
Faloye, Oluwaseun Temitope
;
Ajayi, Ayodele Ebenezer
;
Kamchoom, Viroon
;
Sinsamutpadung, Natdanai
;
Adeyeri, Oluwafemi
The combined application of biochar and inorganic fertilizers has demonstrated significant potential to enhance crop productivity under both rainfed and irrigated conditions. However, predictive modeling approaches utilizing machine learning (ML) to simulate field outcomes under diverse agronomic scenarios remain understudied. This study addresses two critical objectives: (i) to evaluate the efficacy of ML models—Support Vector Machine (SVM), Artificial Neural Network (ANN), and Boosted Trees (BT)—in predicting maize grain yieldin biochar-inorganic fertizer amended soil under drip irrigation; and (ii) to identify the growth stage(s) and ML models that deliver the most accurate predictions. A three-year factorial field experiment was conducted during dry seasons, testing five biochar rates (0, 3, 6, 10 and 20 t/ha), two fertilizer levels (0 and 300 kg/ha), and deficit irrigation treatments (60%, 80%, and 100% of full irrigation). Growth parameters were measured at vegetative (35 days after planting – DAP), flowering stage (49 DAP), and maturity stage (77 DAP), with grain yield recorded at harvest (90 DAP). The measured growth parameters at the different DAP were used for the grain yield forecast. 70, 15 and 15% of the dataset were used for model training, validation and testing, respectively. Field results revealed progressive increases in growth parameters from vegetative to maturity stages, with treatment efficacy following the order: control < biochar-only < fertilizer-only < combined biochar-fertilizer. ML predictions mirrored this hierarchy, with ANN achieving superior accuracy (R² = 0.73–0.85, RMSE = 0.43–0.76, NRMSE = 0.095–0.17 at maturity) compared to SVM and BT. Predictive performance was weakest at the vegetative stage (35 DAP) but improved during flowering (49 DAP) and maturity (77 DAP), underscoring the importance of later growth data for reliable yield forecasting. This study demonstrates that ML models, particularly ANN, can effectively predict maize yield using accessible growth metrics, offering a cost- and labor-efficient complement to traditional field research. By enabling rapid scenario analysis, such models empower stakeholders to optimize resource allocation and inform crop management decisions under varying irrigation and soil amendment strategies.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results