Transfer Learning Model Research Articles

A New Approach for Brain Tumor Detection Using Machine Learning

Introduction: The abnormal brain cells consist of brain tumor which leads to severe organ dysfunction and potentially death. These tumors exhibit a wide range of sizes, textures, and locations. Diagnosing brain tumors process is a time-consuming process requiring the expertise of radiologists. Brain tumors are classified as glioma, meningioma, pituitary, and no tumor. As patient numbers and data volumes rise, traditional methods have become costly and inefficient. Methods: Researchers have developed algorithms for detecting and classifying brain tumors and prioritizing accuracy and efficiency. Deep learning (DL) techniques are increasingly used to create automated systems capable of precisely diagnosing or segmenting brain tumors, particularly for brain cancer classification. This approach supports the use of transfer learning models in medical imaging. This proposed model is a modification to components of Xception model by adding a lot of parameters for increasing the Xception model efficiency. Results: This proposed Xception model was applied to Masoud Nickparvar braintumor- mri-dataset, achieving an accuracy of 99.6%, sensitivity of 99.7%, and specificity of 99.7% with an F1 score of 99.9%. Discussion: The efficiency parameters of the proposed model assured that it is an effective model for diagnosing brain tumor. Comparative analysis with other models shows that the proposed framework is highly reliable for the timely detection of various brain tumors. Conclusion: The results confirm the effectiveness of our proposed model, which attains higher overall accuracy in tumor detection compared to previous models. As a result, the proposed model is considered a valuable decision-making tool for experts in diagnosing brain tumor.

Skin Cancer Detection using CNN (Convolution Neural Network) with AI Medical Assistant

Skin cancer is a very common disease, and with numerous morphological forms and characteristics that are similar between malignant and benign skin lesions, makes the diagnosis pretty difficult. Recent trends in deep learning, with CNNs being the first to enter this domain, have highly facilitated the classification and detection of skin cancer, even in its nascent stages, from images. A plethora of such research works proposed novel architectures for CNNs that may possibly improve the accuracy in the diagnosis of the lesion given as against those belonging to seborrheic keratosis, nevus, or melanoma. The best model, according to the works on these authors, was the multi-layer deep learning model known as SkinLesNet, which performed better than established models such as ResNet50 and VGG16 in all datasets. More extensive studies present a DCNN model that addresses the issue of class imbalance and is stronger than transfer learning models, such as DenseNet and MobileNet. This model achieved strong accuracy on several instances: 98.5% with HAM10000 and 97.1% with ISIC-2019 data. Optimizations in model selection methods: In addition, there existed optimizations in model selection. This was through the application of the multiple-criteria decision making approaches. For instance, the method RAPS can be taken as an example. Through the application of this method, a very significant increase in accuracy was achieved. For optimizing the use of feature selection using Inception V3 in determining diagnosis for oral cavity squamous cell carcinoma, BPSO showed an accuracy of 96.3%. Such breakthroughs underpin therefore the capability of deep learning and AI techniques in raising diagnostic accuracy while cutting costs and making a chance for early intervention of skin and other cancers.

Multi‐target cognitive electronic reconnaissance for unmanned aerial vehicles based on scene reconstruction

AbstractModel‐free deep reinforcement learning (DRL) is regarded as an effective approach for multi‐target cognitive electronic reconnaissance (MCER) missions. However, DRL networks with poor generalisation can significantly reduce mission completion rates when parameters such as reconnaissance area size, target number, and platform speed vary slightly. To address this issue, this paper introduces a novel scene reconstruction method for MCER missions and a mission group adaptive transfer deep reinforcement learning (MTDRL) algorithm. The algorithm enables quick adaptation of reconnaissance strategies for varied mission scenes by transferring strategy templates and compressing multi‐target perception states. To validate the method, the authors developed a transfer learning model for unmanned aerial vehicle (UAV) MCER. Three sets of experiments are conducted by varying the reconnaissance area size, the target number, and the platform speed. The results show that the MTDRL algorithm outperforms two commonly used DRL algorithms, with an 18% increase in mission completion rate and a 5.49 h reduction in training time. Furthermore, the mission completion rate of the MTDRL algorithm is much higher than that of a typical non‐DRL algorithm. The UAV demonstrates stable hovering and repeat reconnaissance behaviours at the radar detection boundary, ensuring flight safety during missions.

Emotion Analysis from Face and Speech: A Comprehensive Review of Current Techniques and Models

Emotion detection in facial expressions and speech plays a crucial role in enhancing interactive platforms, particularly in learning and assessment systems. This study explores advanced techniques for integrating dynamic mock test generation and interview simulation modules in an Advanced Placement Preparation Platform. The dynamic mock test uses real-time feedback to recommend questions based on a student’s performance, leveraging machine learning algorithms for adaptive learning. Additionally, the interview simulation module incorporates facial expression recognition using Convolutional Neural Networks (CNNs) and speech analysis using Recurrent Neural Networks (RNNs) to evaluate student performance. Initial results show that traditional models struggle with real-time adaptability and emotion classification accuracy, underscoring the need for specialized algorithms for complex data. To address these limitations, the system evaluates deep learning models designed for adaptive learning and emotion analysis, such as transfer learning models in emotion detection. The findings highlight the potential for using multimodal data to improve user engagement and performance evaluation in educational settings, paving the way for more immersive and intelligent learning platforms.

DIAGNOSIS OF PROSTATE CANCER WITH ENHANCED EFFICIENCY USING FINE-TUNED CNN AND TRANSFER LEARNING

Cancer is one of the high-risk diseases for humans. Prostate cases are the second most common disease in men after lung cancer, and early diagnosis is vital. Artificial intelligence technologies have begun to be used in the diagnosis of prostate cancer, and more effective and sensitive results have been obtained, preventing potential errors in human-centered methods. In this study, in order to increase the classification performance in the diagnosis of prostate cancer, transfer learning methods and fine-tuning processes, which have higher success and learning ability with less training data, unlike machine learning methods, were applied. The two-class data set consisting of prostate cancer MR images, ‘significant’ and ‘not-significant’, was classified with Alexnet, Densenet201, Googlenet, and Vgg16 models with the feature extraction approach, and 71.40%, 72.05%, 65%, and 80.13% accuracy results were obtained respectively. To increase these rates, pre-trained transfer learning models were used and accuracy results of 89.74%, 94.32%, 85.59%, and 91.05% were achieved, respectively. A 98.10% validation result was obtained using the cross-validation method in the Densenet201 model. DenseNet201 model achieved the highest accuracy result of 98.63% in transfer learning with the combination of the RMSProp optimization method. The proposed transfer learning model provided an improvement of approximately 26% compared to the feature extraction method.

An enhanced lightweight T-Net architecture based on convolutional neural network (CNN) for tomato plant leaf disease classification.

Tomatoes are a widely cultivated crop globally, and according to the Food and Agriculture Organization (FAO) statistics, tomatoes are the third after potatoes and sweet potatoes. Tomatoes are commonly used in kitchens worldwide. Despite their popularity, tomato crops face challenges from several diseases, which reduce their quality and quantity. Therefore, there is a significant problem with global agricultural productivity due to the development of diseases related to tomatoes. Fusarium wilt and bacterial blight are substantial challenges for tomato farming, affecting global economies and food security. Technological breakthroughs are necessary because existing disease detection methods are time-consuming and labor-intensive. We have proposed the T-Net model to find a rapid, accurate approach to tackle the challenge of automated detection of tomato disease. This novel deep learning model utilizes a unique combination of the layered architecture of convolutional neural networks (CNNs) and a transfer learning model based on VGG-16, Inception V3, and AlexNet to classify tomato leaf disease. Our suggested T-Net model outperforms earlier methods with an astounding 98.97% accuracy rate. We prove the effectiveness of our technique by extensive experimentation and comparison with current approaches. This study offers a dependable and understandable method for diagnosing tomato illnesses, marking a substantial development in agricultural technology. The proposed T-Net-based framework helps protect crops by providing farmers with practical knowledge for managing disease. The source code can be accessed from the given link.

Comparative exploration of deep convolutional neural networks using real-time endoscopy images

Until now various deep convolutional neural networks are designed and trained for the purpose of classifying different medical conditions related to the domain of gastroenterology. Most of the study carried out have considered publicly available datasets to train the classification networks. Nevertheless, the main motive for carrying out different works in the field gastroenterology is to administer the developed models in healthcare centers in real-world set-ups. For doing so, it is important to check the generalizing ability of the designed systems by regulating them so as to classify endoscopy images captured in a specific hospital. In this regard, the foremost work completed is the collection of the endoscopy data from the hospital and then correctly annotating the images taking the help of a senior endoscopist with experience of more than 5 years. Once the data annotation is completed, the images are segregated into the class of normal and abnormal endoscopy images. Four different models are designed in the current work based on deep learning models, transfer learning models and ensemble approaches, and trained to classify the hospital endoscopy data as normal or abnormal. The models are then tested and evaluated based on various performance measures. It is observed from the comparative analysis that the transfer learning-based ensemble model has the best generalizing ability and gives the best specificity of 100 ​%. It is believed that deep learning-based models can assist endoscopists in add-on to human prediction efficiency.

MobileNetV2-Based deep learning architecture with progressive transfer learning for accurate monkeypox detection

The recent monkeypox disease outbreak and its serious implications emphasize the urgent necessity for advanced diagnostic tools capable of rapid and accurate disease detection. In response, this study introduces a novel architecture for monkeypox disease identification, leveraging the capabilities of MobileNetV2 combined with progressive transfer learning and innovative model enhancements. The research began by analyzing six pre-trained models, with MobileNetV2 emerging as the most promising due to its balance of efficiency and performance. Using a self-assembled dataset, the model initially achieved an accuracy of 93.79%. Through strategic data augmentation, this performance increased to 97.85%. Further enhancements incorporated a Residual Dilated Spatial Pyramid Integration (ResDSPI) block, boosting feature extraction capabilities and elevating accuracy to 98.83%. The integration of an Efficient Channel Attention (ECA) module propelled the accuracy to an exceptional 99.34%. Each development stage was meticulously designed to ensure systematic improvements in the model’s ability to diagnose monkeypox effectively. The proposed model was rigorously tested on two datasets: the Monkeypox Skin Images Dataset (MSID) and the Monkeypox Skin Lesion Dataset (MSLD). In the binary classification scenario of the MSLD (monkeypox vs. others), our model achieved superior performance, with an accuracy of 95.55%, precision of 95.45%, and sensitivity of 92%. For the multi-class classification task on the MSID, which includes four classes (Chickenpox, Measles, Monkeypox, and Normal), the model achieved an accuracy of 92.95%, precision of 93.09%, and sensitivity of 89.00%. The model’s superior sensitivity and specificity underscore its potential as a critical asset in the surveillance and control of monkeypox outbreaks. These significant advancements not only enhance predictive performance but also mark a major step forward in the field of monkeypox disease image analysis, setting the stage for future applications in combating infectious diseases.

An Ensemble Transfer Learning Model for Brain Tumors Classification using Convolutional Neural Networks

An Ensemble Transfer Learning Model for Brain Tumors Classification using Convolutional Neural Networks

Dipterocarpaceae trunk texture classification using two-stage convolutional neural network-based transfer learning model

Dipterocarpaceae trunk texture classification using two-stage convolutional neural network-based transfer learning model

A Comparative Study of Deep Learning and Transfer Learning Approaches for Image Forgery Detection

The rapid growth of online activities such as commerce, education, research, and virtual conferences has led to a greater reliance on digital images as primary information sources on social media and other platforms. The extensive use, combined with the ease of modification via image-editing software, highlights the crucial need for effective image forgery detection tools. Traditional detection methods based on handcrafted features have grown less efficient, prompting the introduction of deep learning-based approaches, many of which combine transfer learning with pre-trained models to improve detection efficiency and shorten training time. This research presents a comprehensive evaluation of image forgery detection algorithms, categorizing them as classical, deep learning, and transfer learning frameworks. The study compares deep learning with transfer learning methods, assessing their strengths in feature extraction, classification, and detection accuracy. The findings indicate that, while transfer learning models are particularly effective at feature extraction using pre-trained architectures, deep learning remains superior for classification tasks. This insight intends to help academics construct high-accuracy, efficient models for detecting various forms of forgeries. Combining pre-trained models for feature extraction and deep learning for classification is the best option for real-time digital forensics, increasing detection accuracy and processing speed.

A deep learning model based on the BERT pre-trained model to predict the antiproliferative activity of anti-cancer chemical compounds

ABSTRACT Identifying new compounds with minimal side effects to enhance patients’ quality of life is the ultimate goal of drug discovery. Due to the expensive and time-consuming nature of experimental investigations and the scarcity of data in traditional QSAR studies, deep transfer learning models, such as the BERT model, have recently been suggested. This study evaluated the model’s performance in predicting the anti-proliferative activity of five cancer cell lines (HeLa, MCF7, MDA-MB231, PC3, and MDA-MB) using over 3,000 synthesized molecules from PubChem. The results indicated that the model could predict the class of designed small molecules with acceptable accuracy for most cell lines, except for PC3 and MDA-MB. The model’s performance was further tested on an in-house dataset of approximately 25 small molecules per cell line, based on IC50 values. The model accurately predicted the biological activity class for HeLa with an accuracy of 0.77 ± 0.4 and demonstrated acceptable performance for MCF7 and MDA-MB231, with accuracy between 0.56 and 0.66. However, the results were less reliable for PC3 and HepG2. In conclusion, the ChemBERTa fine-tuned model shows potential for predicting outcomes on in-house datasets.

Utilization of Deep Learning Models for the Early Diagnosis of COVID-19 through Chest Radiography

This study aims to evaluate the earlier detection of COVID-19 through deep-learning chest radiography models. The discussion starts with the impact of SARS-CoV-2 globally and the demand for early indication of the risk of handling infection. It outlines pathogenesis and clinical features of COVID-19, puts forth shortcomings of conventional diagnostics, and elaborates on its increasing role as a choice of preferred diagnostic technique in the form of chest radiography. In this study, a deep residual learning-based convolutional neural network architecture of 49 layers is proposed and its performance is compared with several advanced transfer learning models, like ResNet-50, VGG-16, VGG-19, and Inception-v3. The dataset used the chest radiographs of the patients of the COVID-19,for whom diverse enhancement techniques such as data augmentation, resolution optimization, and region of interest (ROI) selection were applied to improve the model's diagnostic performance. The methodology section describes preparation and preprocessing of the dataset followed by the proposed CNN model configuration that aims to evaluate the models concerning the performance metrics of accuracy, sensitivity, specificity, and F1-score. Comparative results show that the 49-layer CNN obtained an accuracy at the value of 97.0%, sensitivity at 98.41%, specificity at 88.45%, and an F1-score of 92.22%. This reflects that the proposed model outperformed the best-known state-of-the-art methods. The study places great emphasis on diagnostic approaches not only towards medical diagnosis but more so onto the transformative potential in redefining AI integration into healthcare systems as forces of change in addressing health matters such as the COVID-19 pandemic in global health.

Customized Weighted Ensemble of Modified Transfer Learning Models for the Detection of Sugarcane Leaf Diseases

Sugarcane is the primary crop in the global sugar industry, yet it remains highly susceptible to a wide range of diseases that significantly impact its yield and quality. An effective solution is required to address the issues caused by the manual identification of plant diseases, which is time-consuming and has low detection accuracy. This paper proposes the development of a robust Deep Ensemble Convolutional Neural Network (DECNN) model for the accurate detection of sugarcane leaf diseases. Initially, several transfer learning (TL) models, including EfficientNetB0, MobileNetV2, DenseNet121, NASNetMobile, and EfficientNetV2B0, were enhanced through the addition of specific layers. A comparative analysis was then conducted on the enlarged dataset of sugarcane leaf diseases, which was divided into six categories and 4800 images. The application of data augmentation, along with the addition of dense layers, batch normalization layers, and dropout layers, led to improved detection accuracy, precision, recall, and F1 scores for each model. Among the five enhanced transfer learning models, the modified EfficientNetB0 model demonstrated the highest detection accuracy, ranging from 97.08% to 98.54%. In conclusion, the DECNN model was developed by integrating the modified EfficientNetB0, MobileNetV2, and DenseNet121 models using a distinctive performance-based custom-weighted ensemble method, with weight optimization carried out using the Tree-structured Parzen Estimator (TPE) technique. This resulted in a model that achieved a detection accuracy of 99.17%, which outperformed the individual performance of the modified EfficientNetB0, MobileNetV2, and DenseNet121 models in detecting sugarcane leaf diseases.

Analyzing the effectiveness of transfer learning models for leukemia detection in blood cells

In recent years, transfer learning has gained significant attention as a feasible and efficient deep learning approach for a variety of medical image analysis tasks, particularly in the area of disease detection and segmentation. This study investigates the effectiveness of various transfer learning models—ResNet50, MobileNet, InceptionV3, DenseNet121, EfficientNetB4, and a proposed model—for blood cell classification. In recent years, transfer learning has gained significant attention as a feasible and efficient deep learning approach for a variety of medical image analysis tasks, particularly in the area of disease detection and segmentation. By leveraging pre-trained networks, we aimed to enhance the accuracy and efficiency of detecting leukemia in blood smear images. Each model was fine-tuned on a comprehensive dataset consisting of normal and leukemic blood cells. The proposed model demonstrated exceptional performance, achieving an accuracy of 99.43%, significantly surpassing the other architectures evaluated. This improvement highlights the potential of transfer learning in medical imaging, particularly in automating and streamlining the diagnostic process for hematological disorders. Future work will focus on optimizing these models further and exploring their applicability in real-world clinical settings.

A lightweight deep learning method to identify different types of cervical cancer

Cervical cancer is the second most common cancer in women’s bodies after breast cancer. Cervical cancer develops from dysplasia or cervical intraepithelial neoplasm (CIN), the early stage of the disease, and is characterized by the aberrant growth of cells in the cervix lining. It is primarily caused by Human Papillomavirus (HPV) infection, which spreads through sexual activity. This study focuses on detecting cervical cancer types efficiently using a novel lightweight deep learning model named CCanNet, which combines squeeze block, residual blocks, and skip layer connections. SipakMed, which is not only popular but also publicly available dataset, was used in this study. We conducted a comparative analysis between several transfer learning and transformer models such as VGG19, VGG16, MobileNetV2, AlexNet, ConvNeXT, DeiT_tiny, MobileViT, and Swin Transformer with the proposed CCanNet. Our proposed model outperformed other state-of-the-art models, with 98.53% accuracy and the lowest number of parameters, which is 1,274,663. In addition, accuracy, precision, recall, and the F1 score were used to evaluate the performance of the models. Finally, explainable AI (XAI) was applied to analyze the performance of CCanNet and ensure the results were trustworthy.

Classification of Chinese hickory with different aging times using two-dimensional correlation spectral (2DCOS) images combined with transfer learning

The Chinese hickory (Carya cathayensis Sarg.), highly valued in eastern China for its culinary appeal and nutritional value, is prone to rancidity during storage due to its high content of polyunsaturated fatty acids. This research introduces an innovative method for categorizing Chinese hickory nuts according to their aging durations by integrating two-dimensional correlation spectroscopy (2DCOS) with transfer learning, using hyperspectral imaging technology (HSI). The study captured hyperspectral images of 1,200 hickory kernels at various aging stages, extracting spectral data within the 900–––1670 nm range. Using this spectral data, synchronous, asynchronous, and integrative 2DCOS images were generated. The analytical framework comprised five spectral preprocessing methods:first derivative (FD), second derivative (SD), multiplicative scatter correction (MSC), standard normal variate (SNV), and detrend (DET); four wavelength selection techniques:successive projections algorithm (SPA), iteratively retaining information variables (IRIV), variable combination population analysis (VCPA), and iteratively variable subset optimization (IVSO); five transfer learning models:DarkNet19, DarkNet53, ResNet50, InceptionV3 and Xception; along with two traditional approaches: support vector machine (SVM) and linear discriminant analysis (LDA). This thorough strategy was designed to accurately classify the nuts on the basis of their aging time. The findings revealed that synchronous 2DCOS significantly outperformed asynchronous and integrative forms in classifying aged hickory kernels. The FD-IVSO-2DCOS-DarkNet19 model, employing 292 (62.80 %) wavelength variables, reached an impressive accuracy of 98.75 % in the test set, surpassing the SD-SPA-2DCOS-DarkNet53 model, which used only 40 (8.6 %) wavelength variables and achieved an accuracy of 97.92 %. These models excelled over other 2DCOS approaches that utilized the full wavelength range and significantly outdid the conventional one-dimensional spectral LDA and SVM models, regardless of wavelength scope. The study underscores that synchronous 2DCOS images, when combined with suitable transfer learning techniques, provide robust discriminative power, suggesting their potential utility in the classification aged hickory nuts.

Oral Cancer Classification with CNN Based State-of-the-art Transfer Learning Methods

The importance of oral and dental health closely affects other vital organs. In this study, CNN-based transfer learning models are built on histopathologic and intraoral images with benign and malignant lesions. Histopathologic and intraoral images from two different sources have benign or malignant classes of lesions in the mouth. EfficientNetB7, ResNet50, VGG16, and VGG19, Xception, ConvNextBase, and MobileNetV2 were used as transfer learning methods. Model training was performed with 80%-20% train test separation and 20% validation separation on the train set. Acc, Prec, Rec, and F1 metrics were used to evaluate the model. In histopathologocial images, Resnet50 was ahead with 81.25% Acc and 85.25% F1. In intraoral images, ConvNextBase with 0.80% Acc, and 0.80% F1 was found to be more accurate.

Scaling Down for Efficiency: Medium-Sized Transformer Models for Protein Sequence Transfer Learning.

Protein language models such as the transformer-based Evolutionary Scale Modeling 2 (ESM2) can offer deep insights into evolutionary and structural properties of proteins. While larger models, such as ESM2 15B, promise to capture more complex patterns in sequence space, they also present practical challenges due to their high dimensionality and high computational cost. We systematically evaluated the performance of all ESM2 models across many biological datasets to determine the impact of model size on transfer learning. Surprisingly, larger models do not always outperform smaller ones, especially when data is limited. Medium sized models, such as ESM2 650M, exhibited consistent performance, falling only slightly behind the 15B parameter model despite being over 20 times smaller. Additionally, we compared various methods of embedding compression to identify the most effective approach, and we found that mean embeddings consistently outperformed other compression methods. Our results show that ESM2 650M with mean embeddings offers an optimal balance between performance and efficiency, making it a practical and scalable choice for transfer learning in a variety of biological applications.

Screening for severe coronary stenosis in patients with apparently normal electrocardiograms based on deep learning

BackgroundPatients with severe coronary arterystenosis may present with apparently normal electrocardiograms (ECGs), making it difficult to detect adverse health conditions during routine screenings or physical examinations. Consequently, these patients might miss the optimal window for treatment.MethodsWe aimed to develop an effective model to distinguish severe coronary stenosis from no or mild coronary stenosis in patients with apparently normal ECGs. A total of 392 patients, including 138 with severe stenosis, were selected for the study. Deep learning (DL) models were trained from scratch and using pre-trained parameters via transfer learning. These models were evaluated based on ECG data alone and in combination with clinical information, including age, sex, hypertension, diabetes, dyslipidemia and smoking status.ResultsWe found that DL models trained from scratch using ECG data alone achieved a specificity of 74.6% but exhibited low sensitivity (54.5%), comparable to the performance of logistic regression using clinical data. Adding clinical information to the ECG DL model trained from scratch improved sensitivity (90.9%) but reduced specificity (42.3%). The best performance was achieved by combining clinical information with the ECG transfer learning model, resulting in an area under the receiver operating characteristic curve (AUC) of 0.847, with 84.8% sensitivity and 70.4% specificity.ConclusionsThe findings demonstrate the effectiveness of DL models in identifying severe coronary stenosis in patients with apparently normal ECGs and validate an efficient approach utilizing existing ECG models. By employing transfer learning techniques, we can extract “deep features” that summarize the inherent information of ECGs with relatively low computational expense.