2025· Proceedings of the 3rd International Conference on Data Analysis and Machine Learning· pp. 537-542· 0 citations· 9 references
TL;DR
Because the application of data augmentation and regularization techniques, the training and validation loss curves of MobileNetV2 con-verged effectively, which in turn suppressed the issue of overfitting, confirming the superiority of MobileNetV2 in end-to-end feature extraction for COVID-19 imaging.
Abstract
: Since the COVID-19 outbreak, Reverse Transcription Polymerase Chain Re-action (RT-PCR) has limitations of long cycles and low sensitivity, making medical imaging (CT/X-ray) critical for early auxiliary diagnosis. However, manual image interpretation by radiologists is inefficient and prone to errors due to subjective factors.To address this, this paper compares two automated solutions using the CoronaHack dataset with preprocessing: LBP-based SVM and transfer learning-enabled MobileNetV2. Experimental results indicated that the optimized LBP-SVM model attained a test accuracy of 80.65%, whereas the MobileNetV2 model achieved a test accuracy of 95.36% and a recall rate of 98.40 — with only 4 out of 250 true positive samples being missed. Because the application of data augmentation and regularization techniques, the training and validation loss curves of MobileNetV2 con-verged effectively, which in turn suppressed the issue of overfitting. This confirms the superiority of MobileNetV2 in end-to-end feature extraction for COVID-19 imaging.
Timely and accurate COVID-19 diagnosis remains a pressing clinical challenge, particularly in settings where RT-PCR testing is not available or delayed. Chest X-ray imaging offers a widely accessible alternative; yet a fundamental question persists in the literature: does the choice of image preprocessing determine diagnostic success more than the classifier itself? To answer this, a controlled factorial experiment was used to evaluate 108 pipeline configurations on a publicly available X-ray dataset of 251 training and 66 test images spanning three classes: COVID-19, Normal, and Viral Pneumonia. Each configuration combined one of four preprocessing techniques (Grayscale conversion, Histogram equalization, Adaptive histogram equalization, Gaussian filtering), one of three feature extractors (HOG, LBP, GLCM), one of three dimensionality reduction strategies (None, PCA, LDA), and one of three classifiers (SVM, KNN, DT). What emerged from this experiment was striking; accuracy ranged from 39.39% to 96.97% across configurations that used identical classifiers and feature extractors, with preprocessing selection alone accounting for the bulk of this 57.58 percentage point spread. Beyond this finding, the experiment revealed that preprocessing and feature extraction do not operate independently; Gaussian filtering actively enhanced the gradient signal that HOG depends on, while grayscale conversion preserved the intensity patterns that LBP encodes, demonstrating that alignment between preprocessing and feature descriptor is what separates high-performing pipelines from poor ones. The best-performing configuration—Gaussian filtering paired with HOG and SVM—reached 96.97% accuracy with macro-averaged precision, recall, and F1-score of 0.97, running entirely on standard CPU hardware with no deep learning infrastructure required. These findings reframe preprocessing from a routine preparatory step into the central design decision of any classical ML-based COVID-19 screening pipeline.
A. M. Madbouly, S. Mostafa, M. M. Abdelhamied· Engineering, Technology &...· 0 citations
A deep learning-based system that uses chest X-ray pictures to automatically detect tuberculosis, using transfer learning using MobileNetV2 and DenseNet architectures to classify chest Xrays as either TB-positive or Healthy, reaching notable accuracy.
Zoya Nasreen, Afshan Fatima, Ruqiya Fatima· International Journal of AI...· 0 citations
A hybrid ensemble learning approach to classify chest X-ray images into four classes—Normal, COVID-19, Pneumonia, and Tuberculosis exhibited high sensitivity in detecting Tuberculosis with considerable stability in classifying Normal, COVID-19, and Pneumonia.
Abdul Rehman Khan Tareen, Muhammad Laiq Ur Rahman Shahid, Muhammad Hamza Zafar et al.· Allied Medical Research Jour...· 0 citations
Accurate diagnosis of pulmonary lesions, pneumonia and COVID-19 in the early stages is difficult due to overlapping symptoms and major logistic limitations of advanced molecular testing techniques. In this study, a computational approach is proposed in which the DenseNet121 architecture and the Contrast Limited Adaptive Histogram Equalization (CLAHE) are combined for analysis of full frame thoracic digital radiograph. The system reduces selection bias and computational latency while evaluating uncropped thoracic fields, preserving peripheral lung textures and minimizing interference from surrounding structures such as the rib cage and heart. Evaluation against an independent blind dataset of 50 images yielded a classification accuracy of 92%. The framework showed robust screening ability for infectious diseases with a recall of 1.00 for both COVID-19 and Viral Pneumonia, with F1-scores of 0.92 and 0.97, respectively. Interpretability validation with Gradient-weighted Class Activation Mapping (Grad-CAM) confirmed the network’s focus on basal and peripheral pulmonary features, although visual inspection indicated a potential for shortcut learning due to radiopaque non-anatomical markers. This architecture is not intended to replace standard clinical diagnostic procedures; instead, it serves as a scalable triage layer to optimize clinical resource allocation, reduce specialist fatigue, and provide a screening tool medical centers and resource-limited clinics.
Keywords: Artificial Intelligence; Deep Learning; DenseNet121; CLAHE; COVID-19; Pneumonia; Respiratory Pathology; Screening Systems.
Raúl Isea· Innovative Insights in Digit...· 0 citations
The differential diagnosis of pulmonary tuberculosis (TB) and non-tuberculous mycobacterial (NTM) lung diseases is practically difficult due to the considerable overlap in their clinical and radiological findings, and incorrect TB diagnosis can lead to unnecessary antituberculosis treatment, drug toxicity, and delayed recognition of the true etiology. A significant portion of the current literature on TB–NTM differentiation relies on a limited number of two-dimensional (2D) CT slices or manually designed radiomic features. These approaches fail to adequately utilize the comprehensive information of the three-dimensional (3D) lung structure and often disregard probability reliability, which is critical for clinical decision support. This study presents an end-to-end deep learning approach for TB–NTM differentiation using NIfTI-format data from 3D computed tomography (CT) volumes, employing a two-stage training strategy based on 2.5D architecture (MobileOne–LSTM). Model performance was rigorously evaluated through stratified 5-fold cross-validation using accuracy, sensitivity, specificity, and F1 score; additionally, the clinical reliability of the probability outputs was analyzed using calibration curves and the Brier score. Experimental results show that the model achieved a mean sensitivity of 0.917 and an F1-score of 0.863 for TB, with an AUC of 0.830, substantially reducing the risk of false negatives at the optimal classification threshold. The satisfactory level of probability calibration demonstrates that the model outputs can be reliably used in clinical decision support scenarios. The presented study offers an original and clinically applicable approach that aims to overcome the limitations of existing methods by introducing a volumetric 2.5D deep learning and calibration-focused perspective to the TB–NTM discrimination problem.
Unknown authors· Gümüşhane Üniversitesi Fen B...· 0 citations
Accurate and interpretable diagnosis of coronavirus disease 2019 (COVID-19) and pneumonia from chest X-ray images is critical for timely clinical decision-making, yet many deep learning models remain difficult to interpret in medical settings. In this study, a convolutional neural network (CNN) was developed to classify chest X-ray images into healthy, pneumonia, and COVID-19 categories, and its decision-making process was analyzed using explainable artificial intelligence (XAI) techniques. Layer-wise relevance propagation (LRP) and Deep Taylor Decomposition (DTD) were applied to generate attribution heatmaps identifying image regions most influential in model predictions, and multiple LRP variants were compared under both binary and multi-class classification settings. Model performance was evaluated using accuracy, precision, recall, F1-score, and confusion matrices. Explainability was assessed through complementary qualitative and quantitative analyses: heatmaps from representative clinically characterized cases were compared with reported radiographic findings, while intersection over union (IoU) quantified spatial alignment between thresholded relevance maps and manually delineated lung-field regions of interest. The binary classifier achieved 95.4% accuracy, while the three-class model achieved 95.7% overall accuracy; COVID-19 classification yielded a precision of 0.988, recall of 0.982, and F1-score of 0.985. Qualitative analysis showed that several positive-attribution methods produced relevance patterns corresponding with radiographically described pulmonary abnormalities. Quantitative IoU analysis demonstrated substantial variation in anatomical lung-field alignment across XAI methods, with the highest observed IoU values reaching approximately 0.77 at a relevance threshold of t = 0.5. Together, these results highlight the importance of combining accurate classification with interpretable model explanations for more transparent and trustworthy medical imaging applications, particularly in rural and under-resourced healthcare settings.
