Lung cancer segmentation Using the Att-U-Net Model on PET-CT Images.

Background: One of the main tools for investigating pulmonary disorders is computed tomography. Starting with a CT, analyses can be qualitative (e.g., direct interpretation of 2D slices, virtual bronchoscopy) or quantitative (e.g., fibrosis score). Qualitative analyses can be performed without segmentation, but quantitative analyses require lung segmentation. Methods: We present the concepts for a class of lung segmentation methods that use region-growing algorithms, the implementation and testing details, and the results obtained in our software platform. Accurate segmentation of lung regions from medical images is a crucial step in computer-aided diagnosis (CAD) systems for pulmonary diseases such as chronic obstructive pulmonary disease (COPD), pneumonia, and lung cancer. Manual segmentation is time-consuming and subjective, while fully automated methods may fail under challenging imaging conditions. Results: This article presents a semi-automated lung segmentation approach, based on region-growing methods, that balances automation with user control. Conclusions: The proposed technique effectively delineates lung boundaries in computed tomography (CT), minimizing computational complexity and manual effort.

ABSTRACT COVID‐19 is a rapidly spreading infectious disease that has posed significant challenges to global healthcare systems. Chest computed tomography (CT) plays a crucial role in detecting pulmonary abnormalities associated with COVID‐19, especially when RT‐PCR tests yield false‐negative results. However, the limited availability of annotated CT datasets, class imbalance, and inter‐patient variability significantly affect the reliability of deep learning‐based diagnostic systems. To address these challenges, this paper proposes an augmented artificial intelligence framework that integrates deep convolutional generative adversarial networks (DCGAN), U‐Net‐based lung segmentation and deep convolutional neural networks (DCNN) for accurate COVID‐19 diagnosis using CT images. A DCGAN model is employed to synthesise additional CT images, where synthetic samples are generated class‐wise and added only to the training set after patient‐wise data separation, thereby expanding the dataset from 672 real CT images (345 COVID‐19 and 327 other pneumonia cases) to 2672 images while reducing class imbalance. The synthesised and real CT images are processed through a U‐Net architecture to segment lung regions and enhance the focus on pathological features. Subsequently, classification is performed using multiple DCNN architectures, including AlexNet, VGG‐16, VGG‐19, ResNet50 and DenseNet. A structured four‐phase experimental evaluation is conducted to independently analyse the impact of augmentation and segmentation, ensuring transparent performance comparison. Strict patient‐wise data splitting is enforced prior to augmentation to prevent data leakage and ensure unbiased generalisation. Experimental results demonstrate that the combined use of synthesised data and lung segmentation significantly improves diagnostic performance, with AlexNet achieving the highest accuracy of 97.6%, F1‐score of 0.97 and specificity of 0.98. The proposed framework provides a reliable and computationally efficient solution to support radiologists in rapid COVID‐19 screening and clinical decision‐making.

Lung cancer is currently recognized as one of the most dangerous cancers, with high mortality rate. In order to deal with lung cancer, an important task is to detect lung nodules early to improve patient survival rates, and computed tomography (CT) scans are crucial data for this. In this research, we propose a deep learning-based method for detecting lung nodules in the CT images with the goal of increasing the likelihood of nodule appearance in the input data of the network, making it easier for the model to focus on relevant areas while reducing noise from areas unrelated to the result. Specifically, we propose a simple lung region segmentation process and optimize the hyperparameters of the faster region-based convolutional neural networks (faster R-CNN) model based on the analysis of nodule characteristics in CT image data. In our experiments, to evaluate the effectiveness of our proposals, we conducted tests on the standard LUNA16 dataset with different backbone configurations for the model, namely ResNet50, ResNet50v2, and MobileNet. The best results achieved were 0.86 mAP50 and 0.91 Recall for the Resnet50, and 0.84 mAP50 and 0.94 Recall for the ResNet50v2. These impressive outcomes underscore the success of our method and establish a robust basis for future studies to further integrate AI into healthcare solutions.

IntroductionDisease-modifying treatments such as monoclonal antibodies can be highly effective in chronic inflammatory diseases such as COPD, but often fail in clinical trials due to heterogeneity of inflammation and imperfect tools to stratify patients to select optimal therapeutic approaches. Molecular imaging provides the potential to transform precision medicine in this field.MethodsWe developed and tested a novel molecular imaging platform using therapeutic monoclonal antibodies labelled with SPECT-CT detectable markers to quantify in vivo tumour necrosis factor (TNF) involved in chronic lung inflammation in humans. We undertook a proof-of-concept clinical study involving participants with COPD and healthy controls. Participants underwent SPECT-CT imaging at 6- and 24 h following injection of 99mTc-anti-TNF. Segmentation of lung regions and 99mTc-anti-TNF activity quantification was undertaken using novel semi-automated and AI-driven approaches.ResultsA significant increase in normalised activity, representing increased TNF inflammatory activity, was seen between the two time-points in the COPD group (mean±sd: 64.88±31.04%, p=0.029) and not in healthy controls (35.38±34.33%, p=0.110). However, analysis at a single time-point revealed higher normalised activity in the healthy group. We demonstrated that pulmonary blood vessel density and degree of emphysema were strongly correlated with this activity signal and identified as confounding factors, highlighting the need to address differences in target-organ characteristics in COPD. Experimental methods to adjust for these factors were developed for organ-specific signal quantification.ConclusionsWe report novel analysis techniques for molecular imaging of the human lung, presenting a platform which provides new insights into complex inflammatory disease and future precision medicine approaches.

This research introduces a unique dataset targeting Silicosis, a significant global occupational lung disease, and a member of the Pneumoconiosis family. Addressing the challenges in healthcare data collection and the need for expert annotation, this dataset aims to aid AI algorithms in medical applications. The comprehensive dataset includes not only Silicosis cases but also related conditions, such as tuberculosis and silicotuberculosis, alongside healthy lung images, addressing the diagnostic complexity due to symptom overlap. As the first public dataset of its kind, it offers detailed annotations for lung and disease region segmentation, as well as disease prediction, provided by multiple radiologists. Baseline experiments and findings demonstrate that current AI models have limited predictive accuracy for these disease classes, emphasizing the critical need for dedicated research. It is our assertion that the proposed Silicodata can be a key dataset in designing automated Silicosis detection tools and addressing challenges associated with small sample sizes in medical AI research.

Accurate segmentation of lung regions from CT scan images is critical for diagnosing and monitoring respiratory diseases. This study introduces a novel hybrid architecture Adaptive Attention U-NetAA, which combines the strengths of U-Net3 + and Transformer based attention mechanisms models for high-precision lung segmentation. The U-Net3 + module effectively segments the lung region by leveraging its deep convolutional network with nested skip connections, ensuring rich multi-scale feature extraction. A key innovation is introducing an adaptive attention mechanism within the Transformer module, which dynamically adjusts the focus on critical regions in the image based on local and global contextual relationships. This model's adaptive attention mechanism addresses variations in lung morphology, image artifacts, and low-contrast regions, leading to improved segmentation accuracy. The combined convolutional and attention-based architecture enhances robustness and precision. Experimental results on benchmark CT datasets demonstrate that the proposed model achieves an IoU of 0.984, a Dice coefficient of 0.989, a MIoU of 0.972, and an HD95 of 1.22 mm, surpassing state-of-the-art methods. These results establish U-NetAA as a superior tool for clinical lung segmentation, with enhanced accuracy, sensitivity, and generalization capability.

Lung disease including pneumonia, tuberculosis, COPD, and COVID-19remain major public health challenges. Early, accurate diagnosis from chest X-rays or CT scans is crucial, yet manual interpretation is time-consuming and susceptible to human error. This study presents a comparative deep learning framework leveraging three architecturesVGG16, ResNet18, and Vision Transformer (ViT) to detect and classify lung disease from medical imaging. We curated a dataset of 3,475 chest X-ray images labelled into three classes: normal, lung opacity, and pneumonia. Data preprocessing included resizing, normalization, lung-region segmentation, and augmentation techniques such as histogram equalization and flipping. The dataset was split into 70% training, 15% validation, and 15% testing sets. Keywords- lung disease detection, datasets, Machine Learning, pulmonary nodule, COVID-19, Chest CT, deep learning, machine learning, convolutional neural network (CNN)

Introduction:Electrical Impedance Tomography (EIT) is widely used for bedside ventilation monitoring but is limited in reconstructing cardiac-related signals due to the dominance of lung impedance changes. This study aims to reconstruct heart-related impedance imaging from lung EIT using a novel semi-Siamese U-Net architecture.Methods:A deep learning model was developed with a shared encoder and two decoders designed to segment lung and heart regions independently. The model was trained and validated on FEM-based EIT simulations and tested on real human EIT data. A weighted binary cross-entropy loss was applied to emphasize cardiac-related learning.Results:The model achieved a Dice coefficient >0.99 and MAE <0.1% on simulation data. It successfully separated lung and heart regions on human EIT frames without additional fine-tuning, demonstrating strong generalization capacity.Discussion:These findings reveal that the semi-Siamese U-Net can overcome signal dominance and improve cardiac-related EIT reconstruction. However, promising results are currently limited to qualitative evaluation of real data and simulation-based training.Conclusion:The proposed method offers a potential pathway for simultaneous lung-heart monitoring in ICU settings. Future work will focus on clinical validation and real-time implementation.

Abstract - Lung cancer remains a major global health concern due to late-stage diagnosis and limited access to fast, reliable screening tools. This research presents an integrated deep learning framework for automated lung cancer detection using CT scan images. The proposed system combines U-Net for precise segmentation of lung regions with a CNN-SVM hybrid model for accurate binary classification. Preprocessing steps such as resizing, normalization, and augmentation enhance model generalization, while the Adam optimizer and EarlyStopping techniques improve training efficiency. Evaluation results demonstrate high accuracy and stable training behavior. To support practical deployment, a Streamlit-based web application was developed, enabling real-time prediction from uploaded CT images. The framework offers a scalable and accessible solution for improving early lung cancer detection and aiding clinical decision-making. . Key Words: U-Net, CNN-SVM, Lung Cancer Detection, CT-Scan, Segmentation, Deep Learning

In this work, an efficient Hybrid Lung Nodules Detection Network (HLNDNet) is developed for early lung cancer diagnosis using the Hybrid Deep Leaning Approach (HDLA). The effect of hybridizing deep learning architectures, such as Visual Geometric Group (VGG)-16, AlexNet, and GoogleNet, has not been thoroughly examined, even though these designs have been extensively researched for various image processing applications. Hence, this work designs an HDLA to detect and classify lung nodules using Computed Tomography (CT) images. The important modules include preprocessing, lung region segmentation, and lung nodule detection by a hybrid approach. Wiener filter and morphological operations are employed in preprocessing modules to filter noise and enhance lung structures. In the second module, the lungs are segmented, and the lung’s external components are removed using the blob detection technique. The proposed semantic image segmentation is employed independently using VGG-16, AlexNet, and GoogleNet in the last module. It labels each pixel in the lung region as either normal or abnormal (lung nodule). In the last module, the outputs of the architectures are hybridized using a Weighted Voting Approach (WVA) to reduce the false positives. The HLNDNet’s performance is evaluated using CT images of 130 patients from the database of Lung Image Database Consortium (LIDC) with three performance metrics: Intersection over Union (IoU), Class Accuracy (CA), and Pixel Accuracy (PA). Results show that the proposed HLNDNet attains an impressive PA of 98.39%, mean CA of 90.42%, and mean IoU of 78.68% for detecting lung nodules, as the hybrid approach helps improve the performance metrics over their counterparts.

Abstract Tuberculosis (TB) stands as the foremost global cause of mortality, a highly contagious lung ailment trailed closely by malaria and HIV/AIDS. To facilitate precise lung X-ray image analysis, which is crucial for diagnoses like lung tuberculosis, lung X-ray image segmentation takes precedence. The formidable U-net architecture, renowned in deep learning for image segmentation, is prominent in this endeavor. This architectural marvel comprises a contracting pathway, adept at extracting high-level information, and a symmetrically expanding pathway, adept at restoring vital details. Setting itself apart, this network outshines many counterparts and exhibits the capacity for comprehensive training even with a limited dataset. In this context, the primary goal of our work is to provide an automated lung segmentation method aimed at addressing the challenge of reconstructing damaged lung sections, which makes a significant contribution to our field of medical science and in the application of artificial intelligence that automatically segments an image of the lung to aid TB detection and classification. The proposed approach can be distilled into three fundamental steps: (a)Image Acquisition: This initial step involves describing the materials and techniques employed for image collection. (b)Initial Segmentation: This critical phase utilizes the power of the U-net deep convolutional network (CNN) model and employs three distinct approaches. These approaches play a pivotal role in the initial segmentation of lung regions. (c) Ensemble Modeling: Subsequently, all three models are amalgamated through ensemble modeling. This consolidation process combines the best outputs from all three right at every pixel in each of the three approaches to yield a final result. In summary, our focus is achieving accurate lung segmentation, particularly for damaged sections, using a comprehensive method that leverages semantic segmentation, transfer learning, and deep learning techniques, notably the Hybrid U-Net model. This approach also enhances lung nodule detection, making it a valuable contribution to the field. Three transfer learning methods trained on large-size image datasets are ResNet34, Inception V3, and VGG 16. The results demonstrated impressive performance metrics, including Mean IoU, Dice-Score, and F-Score. The result section shows that the hybrid U-net model with inception V3 gives better results than the rest, with 0.975 Mean IoU, 0.987 Dice Score, and 0.9963 F-Score. After the ensemble method improved, the result of (0.976 )IoU and (0.988) Dice-Score was achieved.

All over the world, pneumonia is a leading cause of sickness and death, so it’s critical to identify the condition soon and accurately to help patients. Reading chest X-ray films has long relied on the subjective judgments of radiologists, who may also need a lot of time. A method is proposed in this study to automatically find and classify pneumonia in chest X-ray images using deep learning and machine learning for the final diagnosis. The proposed strategy is designed using images of a publicly available chest X-ray pneumonia dataset with Normal or Pneumonia labels. At the beginning of preprocessing, noise must be removed, images must be normalized and more examples of data are generated to cope with class imbalance. An active contour method developed by Chan and Vese is introduced to segment lung regions and assist in obtaining accurate features. These networks automatically identify texture and shape details that illustrate important signs of pneumonia. Each machine learning classifier, Support Vector Machine (SVM), K-Nearest Neighbors (KNN) and Logistic Regression, is trained with these features and then tested on various groups of data. Results from the experiments prove that the SVM classifier is better than other models, as it delivered a test accuracy of 93.5%, a precision of 94.7%, a recall of 95.8%, and an ROC-AUC of 0.97. According to the results, the KNN method shows stronger test accuracy, while Logistic Regression still performs decently. The analysis also shows that SVM reduces the number of both false negatives and false positives, essential for making decisions in medical practice.

Chest X-Ray (CXR) imaging has developed as an important technique for identifying lung diseases, especially in low- and middle-income nations where tuberculosis and pneumonia are serious health problems. With the onset of the COVID-19 pandemic, the need for early and accurate diagnosis has become even more pressing. This research presents a hybrid segmentation and classification for the multiclass lung disease classification using CXR images. The authors use Deep Atrous Attention U-Net (DAA-UNet), specifically designed for lung segmentation, enhancing the Region of Interest (RoI) for classification. The segmented lung regions are then classified using fine-tuned transfer learning on pre-trained models (ResNet101, ChexNet, DenseNet201, and InceptionV3). This hybrid segmentation and classification method achieves an average accuracy of 96.87%, significantly outperforming other classification models, as evidenced by metrics such as precision, sensitivity, specificity, and F1-score. This method exemplifies the potential for integrating deep learning classifiers with image segmentation to improve the diagnosis of lung disease, enabling early intervention and improved patient outcomes.

Lung cancer remains a leading cause of cancer-related mortality globally, with high fatality rates often due to late-stage diagnosis. This research explores the efficacy of computer tomography (CT) imaging in the early detection of lung cancer. CT imaging, known for its high-resolution capabilities, facilitates the early identification of small nodules and abnormalities, providing detailed visualization of lung structures. This allows for the detection of minute changes that may indicate cancer. The investigation addresses the critical health issue, aiming to reduce the significant mortality and morbidity associated with lung cancer through improved early detection methods. The suggested ensemble method starts with segmentation, using residual-UNet and U-Net with DenseNet models to choose the region of interest (ROI). Following this, the Swin Transformer is employed to extract intricate features from the segmented CT images, leveraging its state-of-the-art capabilities. Principal component analysis (PCA) is implemented as a dimensionality reduction technique to optimize computational efficiency and improve feature selection. Furthermore, the DenseNet-121 and ResNet-101 models are employed to precisely identify lung nodule patterns. The investigation found that the residual U-Net model, with a dice coefficient and accuracy of 0.912 and 93.64%, respectively, is a superior method for segmenting lung nodule regions in CT images. The Swin Transformer successfully identified and extracted 215 distinct features from the segmented data obtained from the segmented lung regions. The PCA decreases the number of features extracted by the swin transformer. With an accuracy of 98.01%, an F1 score of 93.71%, and a dice coefficient of 0.938, the residual U-Net + ResNet-101 ensemble model did the best job of finding lung nodules. The outcomes demonstrate the superior performance of the proposed ensemble models compared to each other, making them the most suitable choice for lung cancer identification.

Tuberculosis (TB) continues to be a major cause of death from infectious diseases globally. TB is treatable with antibiotics, but it is often misdiagnosed or left untreated, particularly in rural and resource-constrained regions. While chest X-rays are a key tool in TB diagnosis, their effectiveness is hindered by the variability in radiological presentations and the lack of trained radiologists in high-prevalence areas. Deep learning-based imaging techniques offer a promising approach to computer-aided diagnosis for TB, enabling precise and timely detection while alleviating the burden on healthcare professionals. This study aims to enhance TB detection in chest X-ray images by developing deep learning models. We have observed upper and lower lobe consolidation, pleural effusion, calcification, cavity formation and military nodules. A proposed preprocessing technique has been also introduced in our work based on gamma correction and gradient based technique for contrast enhancement. We leverage the Res-UNet architecture for image segmentation and introduce a novel deep learning network for classification, targeting improved accuracy and precision in diagnostic performance. A Res-UNet segmentation model was trained using 704 chest X-ray images sourced from the Montgomery County and Shenzhen Hospital datasets. Following training, the model was applied to segment lung regions in 1400 chest X-ray scans, encompassing both TB cases and normal controls, obtained from the National Institute of Allergy and Infectious Diseases (NIAID) TB Portal program dataset. The segmented lung regions were subsequently classified as either TB or normal using a deep learning model. A gradient based technique was used for contrast enhancement by capturing intensity changes in image by comparing each pixel with its neighbour with pyramid reduction unique mapping and histogram matching along with gamma correction is used. This integrated approach of segmentation and classification aims to enhance the accuracy and precision of TB detection in chest X-ray images. Classification of segmented images was done using customised convolutional neural network, and visualisation was done using Grad-CAM. The Res-UNet model demonstrated excellent performance for segmentation, achieving an accuracy of 98.18%, recall of 98.40%, precision of 97.45%, F1-score of 97.97%, Dice coefficient of 96.33%, and Jaccard index of 96.05%. Similarly, the classification model exhibited outstanding results, with a classification accuracy of 99.45%, precision of 99.29%, recall of 99.29%, F1-score of 99.29%, and an AUC of 99.9%. Enhanced gradient based method showed ambe of 16.51, entropy of 6.7370, CII of 86.80, psnr of 28.71, ssim of 86.83 which are quite satisfactory. The findings demonstrate the efficiency of our system in diagnosing TB from chest X-rays, potentially surpassing clinician-level precision. This underscores its effectiveness as a diagnostic tool, particularly in resourcelimited settings with restricted access to radiological expertise. Additionally, the modified Res-UNet model demonstrated superior performance compared to the standard U-Net, highlighting its potential for achieving greater diagnostic accuracy.

Tuberculosis remains a significant health challenge worldwide, affecting a large population. Therefore, accurate diagnosis of this disease is a critical issue. With advancements in computer systems, imaging devices, and rapid progress in machine learning, tuberculosis diagnosis is being increasingly performed through image analysis. This study proposes three segmentation models based on U-Net, V-Net, and Seg-Net architectures to improve tuberculosis detection using the Shenzhen and Montgomery databases. These deep learning-based methods aim to enhance segmentation accuracy by employing advanced preprocessing techniques, attention mechanisms, and non-local blocks. Experimental results indicate that the proposed models outperform traditional approaches, particularly in terms of the Dice coefficient and accuracy values. The models have demonstrated robust performance on popular datasets. As a result, they contribute to more precise and reliable lung region segmentation, which is crucial for the accurate diagnosis of respiratory diseases like tuberculosis. In evaluations using various performance metrics, the proposed U-Net and V-Net models achieved Dice coefficient scores of 96.43% and 96.42%, respectively, proving their competitiveness and effectiveness in medical image analysis. These findings demonstrate that the Dice coefficient values of the proposed U-Net and V-Net models are more effective in tuberculosis segmentation than Seg-Net and other traditional methods.

The accurate segmentation of lung regions in computed tomography (CT) scans is critical for the automated analysis of lung diseases such as chronic obstructive pulmonary disease (COPD) and COVID-19. This paper focuses on enhancing the accuracy of U-Net segmentation models through a robust preprocessing pipeline. The pipeline includes CT image normalization, binarization to extract lung regions, and morphological operations to remove artifacts. Additionally, the proposed method applies region-of-interest (ROI) filtering to isolate lung areas effectively. The dataset preprocessing significantly improves segmentation quality by providing clean and consistent input data for the U-Net model. Experimental results demonstrate that the Intersection over Union (IoU) and Dice coefficient exceeded 0.95 on training datasets. This work highlights the importance of preprocessing as a standalone step for optimizing deep learning-based medical image analysis.

Background: Tuberculosis (TB) remains a major global health challenge, causing significant morbidity and mortality. Early and accurate detection is crucial for timely treatment and disease control. Traditional TB diagnosis relies on radiologists analyzing chest X-rays (CXRs), a process that is time-consuming and prone to variability. Advances in artificial intelligence, particularly deep learning, have facilitated the development of computer-aided diagnostic (CAD) systems capable of automating TB detection with improved efficiency and consistency. Objective: This study aimed to develop an automated TB detection system utilizing deep learning techniques to segment lung regions and classify TB-infected CXRs, enhancing diagnostic accuracy and reducing reliance on manual interpretation. Methods: A fully convolutional network (FCN) based segmentation model was implemented to isolate lung regions from CXRs, followed by post-processing techniques to refine segmentation accuracy. The classification module employed a ResNet architecture to differentiate between normal and TB-positive cases. The model was trained and validated on three datasets: the Japanese Society of Radiological Technology (JSRT), Montgomery County (MC), and a locally curated dataset. For classification, the Shenzhen dataset was used. Model performance was evaluated using accuracy, sensitivity, specificity, Dice Similarity Coefficient (DSC), and the area under the curve (AUC). Results: Segmentation accuracy was 97.1% for JSRT, 97.7% for MC, and 94.2% for the local dataset. DSC values were recorded as 95.1%, 95.4%, and 88.0%, respectively. The classification model achieved 84.4% accuracy, with sensitivity of 84.4%, specificity of 90.09%, and AUC of 95.0%. Comparative analysis demonstrated competitive performance with existing methodologies. Conclusion: The proposed deep learning-based CAD system effectively automates TB detection, improving diagnostic efficiency. The integration of advanced segmentation and classification techniques enhances accuracy, facilitating early TB screening. Future research should explore optimizing classification through hybrid deep learning models for improved clinical applicability.

The identification of COVID-19 using chest X-ray (CXR) images plays a critical role in managing the pandemic by providing a rapid, non-invasive, and accessible diagnostic tool. This study evaluates the impact of different image preprocessing techniques on the performance of deep learning models for COVID-19 classification based on COVID-19 Radiography Database, which includes 10,192 normal CXR images, 6012 lung opacity (non-COVID lung infection) images, and 1345 viral pneumonia images. Along with the images, corresponding lung masks are also included to aid in the segmentation and analysis of lung regions. Specifically, three convolutional neural network (CNN) models were developed, each using a distinct preprocessing method: Contrast Limited Adaptive Histogram Equalization (CLAHE), traditional histogram equalization, and no preprocessing. The results revealed that while the CLAHE-enhanced model achieved the highest training accuracy (93.26%) and demonstrated superior stability during training, it showed lower performance in the validation phase, with validation accuracy of 91.31%. In contrast, the model with no preprocessing, which exhibited slightly lower training accuracy (92.98%), outperformed the CLAHE model during validation, achieving the highest validation accuracy of 91.50% and the lowest validation loss. The histogram equalization model demonstrated performance similar to that of CLAHE but with slightly higher validation loss and accuracy compared to the unprocessed model. These findings suggest that while CLAHE excels in enhancing image details during training, it may lead to overfitting and reduced generalization ability. In contrast, the model without preprocessing showed the best generalization and stability, indicating that preprocessing techniques should be chosen carefully to balance feature enhancement with the need for generalization in real-world applications. This study underscores the importance of selecting appropriate image preprocessing techniques to enhance deep learning models' performance in medical image classification, particularly for COVID-19 detection. Histogram Equalization The results contribute to ongoing efforts to optimize diagnostic tools using AI and image processing.