KMITL

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

Browse

Search Results

Now showing 1 - 10 of 16

Some of the metrics are blocked by your
consent settings
Cross-Sensor and Cross-Population Generalization of Deep Learning Models for Digital Mammography: A Controlled Four-Country Benchmark of Five Backbone Architectures with Statistical Significance Testing
(2026-06-01)
Gabbualoy, Somprasonk
;
Phasukkit, Pattarapong
;
Tungjitkusolmun, Supan
Background/Objectives: Deep learning models for digital mammography sensor data are increasingly deployed across hospitals using different X-ray detector technologies and patient populations. Whether models trained on one sensor platform and population maintain accuracy when transferred to another has not been tested for the latest generation of mammography-specific foundation models under one controlled protocol. Methods: We fine-tuned five backbone architectures (ResNet-50, DINOv2-B14, Rad-DINO, Mammo-CLIP B5, and Mammo-FM) on CBIS-DDSM (film-digitized, USA, n = 714 validation) with three seeds, ablated a density-aware focal loss across three auxiliary weights, and evaluated transfer to three external sensor cohorts: CMMD (full-field digital, China, n = 1032), DMID (mixed digital, India, n = 509), and MIAS (film-digitized, UK, n = 322). Significance used paired DeLong z-tests with Benjamini–Hochberg FDR correction; temperature scaling tested post hoc recalibration at all transfer targets. Results: Within this single-source three-seed evaluation, ResNet-50 outperformed all four foundation models on CBIS-DDSM (AUC 0.867 vs. 0.847, 0.846, 0.813, and 0.703; all gaps p_adj < 0.05). The density-aware focal loss degraded both AUC and calibration at every weight tested. At transfer, every model lost 0.165 to 0.320 AUC points relative to in-distribution performance, with sensitivity at 95% specificity collapsing from 0.31 to 0.47 in-distribution to 0.11 to 0.22 across the three external targets. A per-seed Stouffer meta-analysis confirms that Mammo-CLIP B5 and Mammo-FM significantly outperformed ResNet-50 on DMID and Mammo-CLIP on CMMD, after BH-FDR; MIAS comparisons remained directional only. In the extremely dense subgroup (BI-RADS D4), Mammo-FM reached AUC 0.870 versus ResNet-50 at 0.842, a directional observation whose 95% CIs overlap heavily at the n = 140 sample size and which we do not interpret as a statistically supported advantage. Conclusions: In this single training-source, three-seed protocol, mammography-specific pretraining did not deliver the in-distribution AUC premium reported in the originating papers, and no architecture reached a level at which transfer deployment without local validation would be defensible. We frame these as observations specific to the present protocol rather than as broader conclusions about foundation models for mammography classification. The findings argue for sensor-stratified and population-stratified external validation and for local recalibration as practical prerequisites before clinical use. Code and weights are released under MIT license.
Some of the metrics are blocked by your
consent settings
Smart Diagnostics: Hierarchical Deep Learning of Acoustic Emission Signals for Early Crack Detection in Zirconia Dental Structures
(2026-05-01)
Tuntiwong, Kuson
;
Wangman, Rangsinee
;
Kanchanatawewat, Kanchana
;
Anucul, Boonjira
;
Sritart, Hiranya
Monolithic zirconia restorations are frequently affected by the unnoticed growth of subcritical cracks, a failure process that is not captured by traditional imaging methods like radiographs and ultrasounds in sophisticated dental architectures. To address this evaluative inadequacy, this research introduces a hierarchical deep learning framework for microcrack detection and spatial localization. We promote a hierarchical deep learning system that integrates Acoustic Emission (AE) detection alongside signal processing. Raw AE signals utilized during dynamic loading are enhanced via Kalman filtering and Continuous Wavelet Transform (CWT) to construct high-fidelity time–frequency scalograms. The diagnostic pipeline operates in two stages: first, a hybrid CNN–BiGRU network with temporal attention fulfills zirconia component-level classification; second, a ResNet-18 backbone integrated with Bidirectional LSTM and Multi-Head Attention precisely localizes defects across five anatomical crown regions. This hierarchical design effectively captures the non-stationary, transient nature of fracture-induced stress waves. The framework achieved an F1-score of 99.00% and an AUC of 0.994, significantly outperforming conventional convolutional networks. By enabling predictive maintenance through early, non-invasive damage localization, this study demonstrates a promising laboratory framework for AE-based crack detection in zirconia dental structures and prosthetics and toward enhanced clinical reliability in digital dentistry.
Some of the metrics are blocked by your
consent settings
SCTFlow: 3D MRI-to-sCT with Conditional Rectified Flow
(2026-01-01)
Rajborirug, Pharuj
;
Tungjitkusolmun, Supan
;
Gulyanon, Sarun
MRI and CT images are both crucial for radiotherapy planning, since MRI provides superior soft-tissue contrast for tumor delineation, while CT provides Hounsfield units (HU) required for dose calculation. MR-only radiotherapy offers important advantages, including reduced registration errors, elimination of additional radiation exposure, and streamlined clinical workflows. Generating synthetic CT (sCT) from MRI remains challenging due to the need for realistic HU reconstruction and the high computational demands of processing large 3D image volumes.We propose sCTFlow, a Conditional Rectified Flow (CRF) framework for 3D MRI-to-sCT translation. Unlike diffusion probabilistic models (DDPMs), sCTFlow learns a deterministic velocity field mapping noise to data, ensuring stability and requiring fewer sampling steps. Our architecture, a 3D Attention U-Net, conditions on MRI and organ segmentation via feature-wise linear modulation to predict velocity fields, which are subsequently converted into HU estimates.We evaluated our approach on the SynthRAD2023 dataset. sCTFlow achieves an MAE of 81.18 +- 19.48 HU, PSNR of 26.93, and SSIM of 0.830. We also investigated our method and found that the model captures HU distributions rather than relying on simple intensity transformations, indicating its capacity to model underlying CT characteristics. These findings demonstrate that sCTFlow has potential for reliable and clinically applicable MR-only radiotherapy workflows.
Some of the metrics are blocked by your
consent settings
PSD-EEGRepNet: A CNN Architecture with Multi-Branch RepBlocks for Power Spectral Density-Based Motor Imagery EEG Classification in BCI
(2025-01-01)
Thangthong, Kantham
;
Asadi, Fawad
;
Tungjitkusolmun, Supan
Motor Imagery (MI) based Brain-Computer Interfaces (BCIs) utilizing electroencephalography (EEG) offer significant potential, yet progress can be hindered by the computational demands of deep learning classifiers. This study introduces and evaluates a novel lightweight, multi-branch Convolutional Neural Network (CNN), inspired by efficient design principles, specifically for classifying MI tasks from Power Spectral Density (PSD) EEG features. Our objective was to achieve strong classification performance coupled with favorable development phase computational characteristics. Evaluated on 10 subjects from a public PhysioNet dataset using 5-fold cross-validation and two data overlap conditions (80%, 90%), the proposed model (~6.32M parameters, ~20 MFLOPs) demonstrated high mean classification accuracies (80.38% for 80% overlap, 85.69% for 90% overlap) and efficient training times (avg. 8.8s and 20.4s per fold, respectively). While performance scaled positively with data augmentation, inter-subject variability was noted. We conclude that the proposed architecture effectively balances high accuracy with significant offline computational efficiency offering a valuable tool for BCI research and a promising foundation for developing practical MI-BCI systems.
Some of the metrics are blocked by your
consent settings
Common Spatial Pattern Variants for Feature Extraction in Multi-Class Motor Imagery BCI
(2025-01-01)
Asadi, Fawad
;
Thangthong, Kantham
;
Tungjitkusolmun, Supan
Brain-Computer Interfaces (BCIs) based on Motor Imagery (MI) offer intuitive control but face challenges due to noisy electroencephalography (EEG) signals that often contain artifacts, necessitating effective feature extraction. Common Spatial Pattern (CSP) is a standard technique, yet its limitations in multi-class MI scenarios motivate various extensions. This paper investigates the classification performance, computational cost, and the crucial performance versus cost trade-off inherent in practical BCI applications of five prominent CSP variants, including standard CSP, Regularized CSP (RCSP), Filter Bank CSP (FBCSP), Sparse CSP, and Kernel CSP, when applied to multi-class MI classification. Using the PhysioNet MI dataset, we employed a consistent preprocessing pipeline (18 sensorimotor channels, filtering, epoching) and evaluated each variant by feeding extracted features into a Support Vector Machine (SVM) classifier with an RBF kernel. The evaluation revealed that standard CSP and RCSP achieved the highest peak average accuracy (~97.8%) with near-efficient computational cost. FBCSP yielded high accuracy (~96.7%) but incurred the highest computational cost. Sparse CSP was computationally efficient but achieved lower peak accuracy (~95.3%), while Kernel CSP showed the lowest peak accuracy (~92.8%) at moderate cost. The findings highlight the performance profiles and efficiency of these CSP methods for this multi-class MI task, indicating that standard CSP and RCSP offer a particularly effective combination of high accuracy and manageable computational load under the evaluated conditions.
Some of the metrics are blocked by your
consent settings
Automated Crack Detection in Monolithic Zirconia Crowns Using Acoustic Emission and Deep Learning Techniques
(2024-09-01)
Tuntiwong, Kuson
;
Tungjitkusolmun, Supan
;
Phasukkit, Pattarapong
Monolithic zirconia (MZ) crowns are widely utilized in dental restorations, particularly for substantial tooth structure loss. Inspection, tactile, and radiographic examinations can be time-consuming and error-prone, which may delay diagnosis. Consequently, an objective, automatic, and reliable process is required for identifying dental crown defects. This study aimed to explore the potential of transforming acoustic emission (AE) signals to continuous wavelet transform (CWT), combined with Conventional Neural Network (CNN) to assist in crack detection. A new CNN image segmentation model, based on multi-class semantic segmentation using Inception-ResNet-v2, was developed. Real-time detection of AE signals under loads, which induce cracking, provided significant insights into crack formation in MZ crowns. Pencil lead breaking (PLB) was used to simulate crack propagation. The CWT and CNN models were used to automate the crack classification process. The Inception-ResNet-v2 architecture with transfer learning categorized the cracks in MZ crowns into five groups: labial, palatal, incisal, left, and right. After 2000 epochs, with a learning rate of 0.0001, the model achieved an accuracy of 99.4667%, demonstrating that deep learning significantly improved the localization of cracks in MZ crowns. This development can potentially aid dentists in clinical decision-making by facilitating the early detection and prevention of crack failures.
Some of the metrics are blocked by your
consent settings
Advanced White Blood Cells Detection and Analysis with VGG16 Transfer Learning
(2024-01-01)
Eamkong, Anoma
;
Pintavirooj, Chuchart
;
Tungjitkusolmun, Supan
White blood cells (WBCs) provide a significant role in the immune system, and precise classification and quantification are critical for detecting a variety of diseases. This work employs the VGG16 model combined with transfer learning to improve WBC classification and counting in blood smear pictures. We utilized a dataset of 1,651 images, including samples from normal bone marrow and chronic myeloid leukemia (CML) cases, to train and evaluate the model. The workflow incorporates data augmentation, Otsu thresholding, and morphological operators to improve segmentation accuracy. The VGG16-based model achieved a high accuracy of 92.37%, with validation accuracy reaching 96.41 %. Performance metrics were evaluated using accuracy, precision, recall, and F -measure, highlighting that eosinophils provided the highest accuracy, while neutrophils demonstrated the best precision. The model effectively distinguishes between normal and elevated WBC counts, as evidenced by the results from normal and CML blood smears. Despite these promising results, further refinement is suggested. Future work may focus on specific WBC types, such as monocytes or neutrophils, to improve precision and balance memory efficiency. This approach has the potential to advance diagnostic accuracy and efficiency in medical image analysis.
Some of the metrics are blocked by your
consent settings
Crack Localization Detection in Monolithic Zirconia Dental Crowns via 1D-Convolutional Neural Networks Algorithm-Based Acoustic Emission Analysis
(2024-01-01)
Tuntiwong, Kuson
;
Wangman, Rangsinee
;
Sritart, Hiranya
;
Kanchanatawewat, Kanchana
;
Tungjitkusolmun, Supan
The feasibility of utilizing the acoustic emission (AE) technique for the detection and classification of cracks within monolithic dental crown is assessed in this study, owing to its non-destructive nature which enables passive monitoring of structures. The AE signals captured are subjected to analysis to extract pertinent information regarding the source and location of the cracks. A novel approach is proposed, employing deep learning 1D convolutional neural networks (1D-CNNs) for the recognition and classification of the recorded cracked signals. The AE signals, obtained through a handmade AE data acquisition unit, are converted into .csv format and subjected to denoising using Bayesian methods to eliminate background noise. The signals are collected through the breakage of pencil lead (Vallen systeme) Hsu-Nielsen-Source 0.5 (ASTM E976) applied to each surface of the dental crown. Subsequently, the data signals are divided into training and testing groups following an 85 / 15 split. The performance of the deep learning 1D-CNNs is evaluated based on Precision, Recall and total accuracy metrics. The applicated of automated deep learning in this study demonstrated significantly high overall accuracy (98.67%). The integration of handmade data acquisition with 1D-CNN crack detection proves to be an effective method for early screening. The novel method harnesses acoustic emission signals in 1D-CNNs, thereby enhancing the accuracy of clinical dental restorative crack identification and determining the onset time.
Some of the metrics are blocked by your
consent settings
Continuous Wavelet Transform based SqueezeNet for Damage Classification of Monolithic Zirconia Dental Crowns Using Acoustic Emission Analysis
(2024-01-01)
Tuntiwong, Kuson
;
Phasukkit, Pattarapong
;
Tungjitkusolmun, Supan
This work presents a novel approach utilizing continuous wavelet transform and deep learning for pattern recognition and classification of generated cracked signals. The research employs a smart sensing data acquisition system in a non-destructive manner, utilizing dual acoustic emission (AE) sensor to determine the specific cracked position at the crown surface. An AE data acquisition module is used to transform the AE signals captured by the AE sensor into.csv data, which is then subjected to denoising utilizing band stop and Bayesian filters to eliminate background noise. Subsequently, the denoised data are processed and classified using an algorithmic model called SqueezeNet, which facilitates the accurate localization of cracks in the Monolithic crown. For the purpose of training and validating the algorithmic model, the study uses AE signals produced by pencil lead breaking (PLB) at the incisal, labial, palatal, left, and right sides of the crown. In order to make it easier to identify the sources of AE signals resulting from PLB-induced AE events, wavelet transform (WT) is used to assess the location of the crown in respect to AE signals. AE signals are divided into two categories and the algorithm's accuracy in classifying them is assessed during the training and testing stages. With a total accuracy of 9 6. 5%, the deep learning-powered AE approach successfully detects flaws on the dental crown surfaces. The integration of a dual AE sensor with a SqueezeNet algorithm based on AE signals sets this study apart, offering an automated and efficient solution for early cracked detection in dental crowns for prototype clinical dental restorative crack identification.
Some of the metrics are blocked by your
consent settings
Deep Learning-Based Acoustic Emission Scheme for Nondestructive Localization of Cracks in Monolith Zirconia Dental crown under Load
(2023-01-01)
Tuntiwong, Kuson
;
Phasukkit, Pattarapong
;
Tungjitkusolmun, Supan
This study introduces an innovation non-destructive approach employing dual-sensor acoustic emission (AE) techniques to detect and accurately locate cracks within monolith dental crowns, even when subjected to various load directions. Throughout the process, the AE sensor recorded AE signals that were subsequently converted into.csv data using an AE data acquisition module. Afterward, these digital records underwent denoising to eliminate background and contact-related disturbances. Subsequently, the noise-filtered data underwent processing and classification via a deep learning algorithmic model, enabling accurate localization of cracks within the dental crown. The study utilized AE signals generated by pencil lead breakage (PLB) at the labial, left, palatal, and right surfaces of dental crown for both training and testing of the algorithmic model. Initially, wavelet transform (WT) was employed to analyze and examine the dental crown's location in relation to AE signals. The approach was introduced to precisely identify the sources of AE signals originating from PLB-induced AE events. During the algorithm's training and testing phases, the AE signals were categorized into two groups, and the accuracy of their classification was evaluated. The deep learning-powered AE strategy successfully identified cracks within the dental crown. The total accuracy was 90.625%. This study's innovation resides in its utilization of a dual AE sensor combined with a deep learning algorithm based on AE signals. This combination effectively detects and precisely locates cracks within dental crowns. In contrast to current AE crack-localization methods that depend on human interpretation in radiographic examination, this approach offers an automated and more efficient solution.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results