EXPRESS: Estimation of the Global Equivalence Ratio of a Swirl Combustor from a Small Number of Absorption Spectra Using Machine Learning.

A new optical diagnostic method that predicts the global fuel-air equivalence ratio of a swirl combustor using absorption spectra from only three optical paths is proposed here. Under normal operation, the global equivalence ratio and total flow rate determine the temperature and concentration fields of the combustor, which subsequently determine the absorption spectra of any combustion species. Therefore, spectra, as the fingerprint for a produced combustion field, were employed to predict the global equivalence ratio, one of the key operational parameters, in this study. Specifically, absorption spectra of water vapor at wavenumbers around 7444.36, 7185.6, and 6805.6 cm-1 measured at three different downstream locations of the combustor were used to predict the global equivalence ratio. As it is difficult to find analytical relationships between the spectra and produced combustion fields, a predictive model was a data-driven acquisition. The absorption spectra as an input were first feature-extracted through stacked convolutional autoencoders (CAEs) and then a dense neural network (DNN) was used for regression prediction between the feature scores and the global equivalence ratio. The model could predict the equivalence ratio with an absolute error of ±0.025 with a probability of 96%, and a gradient-weighted regression activation mapping analysis revealed that the model leverages not only the peak intensities but also the variations in the shape of absorption lines for its predictions.

Deep learning ensemble approach with explainable AI for lung and colon cancer classification using advanced hyperparameter tuning.

Lung and colon cancers are leading contributors to cancer-related fatalities globally, distinguished by unique histopathological traits discernible through medical imaging. Effective classification of these cancers is critical for accurate diagnosis and treatment. This study addresses critical challenges in the diagnostic imaging of lung and colon cancers, which are among the leading causes of cancer-related deaths worldwide. Recognizing the limitations of existing diagnostic methods, which often suffer from overfitting and poor generalizability, our research introduces a novel deep learning framework that synergistically combines the Xception and MobileNet architectures. This innovative ensemble model aims to enhance feature extraction, improve model robustness, and reduce overfitting.Our methodology involves training the hybrid model on a comprehensive dataset of histopathological images, followed by validation against a balanced test set. The results demonstrate an impressive classification accuracy of 99.44%, with perfect precision and recall in identifying certain cancerous and non-cancerous tissues, marking a significant improvement over traditional approach.The practical implications of these findings are profound. By integrating Gradient-weighted Class Activation Mapping (Grad-CAM), the model offers enhanced interpretability, allowing clinicians to visualize the diagnostic reasoning process. This transparency is vital for clinical acceptance and enables more personalized, accurate treatment planning. Our study not only pushes the boundaries of medical imaging technology but also sets the stage for future research aimed at expanding these techniques to other types of cancer diagnostics.

CSXAI: a lightweight 2D CNN-SVM model for detection and classification of various crop diseases with explainable AI visualization.

Plant diseases significantly impact crop productivity and quality, posing a serious threat to global agriculture. The process of identifying and categorizing these diseases is often time-consuming and prone to errors. This research addresses this issue by employing a convolutional neural network and support vector machine (CNN-SVM) hybrid model to classify diseases in four economically important crops: strawberries, peaches, cherries, and soybeans. The objective is to categorize 10 classes of diseases, with six diseased classes and four healthy classes, for these crops using the deep learning-based CNN-SVM model. Several pre-trained models, including VGG16, VGG19, DenseNet, Inception, MobileNetV2, MobileNet, Xception, and ShuffleNet, were also trained, achieving accuracy ranges from 53.82% to 98.8%. The proposed model, however, achieved an average accuracy of 99.09%. While the proposed model's accuracy is comparable to that of the VGG16 pre-trained model, its significantly lower number of trainable parameters makes it more efficient and distinctive. This research demonstrates the potential of the CNN-SVM model in enhancing the accuracy and efficiency of plant disease classification. The CNN-SVM model was selected over VGG16 and other models due to its superior performance metrics. The proposed model achieved a 99% F1-score, a 99.98% Area Under the Curve (AUC), and a 99% precision value, demonstrating its efficacy. Additionally, class activation maps were generated using the Gradient Weighted Class Activation Mapping (Grad-CAM) technique to provide a visual explanation of the detected diseases. A heatmap was created to highlight the regions requiring classification, further validating the model's accuracy and interpretability.

Near-field detection and peak frequency metric for substrate and activation mapping of ventricular tachycardias in two- and three-dimensional circuits.

AIMS: Successful ventricular arrhythmia (VA) ablation requires identification of functionally critical sites during contact mapping. Estimation of the peak frequency (PF) component of the electrogram (EGM) may improve correct near-field (NF) annotation to identify circuit segments on the mapped surface. In turn, assessment of NF and far-field (FF) EGMs may delineate the three-dimensional path of a ventricular tachycardia (VT) circuit. METHODS AND RESULTS: A proprietary NF detection algorithm was applied retrospectively to scar-related re-entry VT maps and compared with manually reviewed maps employing first deflection (FDcorr) for VT activation maps and last deflection (LD) for substrate maps. Ventricular tachycardia isthmus location and characteristics mapped with FDcorr vs. NF were compared. Omnipolar low-voltage areas, late activating areas, and deceleration zones (DZ) in LD vs. NF substrate maps were compared. On substrate maps, PF estimation was compared between isthmus and bystander sites. Activation mapping with entrainment and/or VT termination with radiofrequency (RF) ablation confirmed critical sites. Eighteen patients with high-density VT activation and substrate maps (55.6% ischaemic) were included. Near-field detection correctly located critical parts of the circuit in 77.7% of the cases compared with manually reviewed VT maps as reference. In substrate maps, NF detection identified deceleration zones in 88.8% of cases, which overlapped with FDcorr VT isthmus in 72.2% compared with 83.3% overlap of DZ assessed by LD. Applied to substrate maps, PF as a stand-alone feature did not differentiate VT isthmus sites from low-voltage bystander sites. Omnipolar voltage was significantly higher at isthmus sites with longer EGM durations compared with low-voltage bystander sites. CONCLUSION: The NF algorithm may enable rapid high-density activation mapping of VT circuits in the NF of the mapped surface. Integrated assessment and combined analysis of NF and FF EGM-components could support characterization of three-dimensional VT circuits with intramural segments. For scar-related substrate mapping, PF as a stand-alone EGM feature did not enable the differentiation of functionally critical sites of the dominant VT from low-voltage bystander sites in this cohort.

Cluster-CAM: Cluster-weighted visual interpretation of CNNs' decision in image classification.

Despite the tremendous success of convolutional neural networks (CNNs) in computer vision, the mechanism of CNNs still lacks clear interpretation. Currently, class activation mapping (CAM), a famous visualization technique to interpret CNN's decision, has drawn increasing attention. Gradient-based CAMs are efficient, while the performance is heavily affected by gradient vanishing and exploding. In contrast, gradient-free CAMs can avoid computing gradients to produce more understandable results. However, they are quite time-consuming because hundreds of forward interference per image are required. In this paper, we proposed Cluster-CAM, an effective and efficient gradient-free CNN interpretation algorithm. Cluster-CAM can significantly reduce the times of forward propagation by splitting the feature maps into clusters. Furthermore, we propose an artful strategy to forge a cognition-base map and cognition-scissors from clustered feature maps. The final salience heatmap will be produced by merging the above cognition maps. Qualitative results conspicuously show that Cluster-CAM can produce heatmaps where the highlighted regions match the human's cognition more precisely than existing CAMs. The quantitative evaluation further demonstrates the superiority of Cluster-CAM in both effectiveness and efficiency.

The value of ripple mapping in the age of coherent mapping in scar-related atrial tachycardia.

BACKGROUND: An accurate display of scar-related atrial tachycardia (ATs) is a key determinant of ablation success. The efficacy of ripple mapping (RM) in identifying the mechanism and critical isthmus of scar-related ATs during coherent mapping is unknown. METHODS: A total of 97 patients with complex ATs who underwent radiofrequency catheter ablation at our center between October 2018 and September 2022 were included. ATs was mapped using a multielectrode mapping catheter on the CARTO3v7 CONFIDENCE module. Coherent and RM were used to identify the reentrant circuit. RESULTS: The mechanisms of 128 ATs were analyzed retrospectively (84 anatomic-reentrant ATs and 44 non-anatomic reentrant ATs). The median AT cycle length was 264 ± 25ms. The correct diagnosis was achieved in 83 ATs (68%) using only coherent mapping. Through coherent mapping plus RM, 114 ATs (84.2%) were correctly diagnosed (68% vs. 89%, p = .019). In non-anatomical reentrant ATs, 81% of the diagnostic rate was achieved by reviewing both coherent and ripple mapping compared to reviewing coherent mapping alone (81% vs. 52%, p = .03). Reviewing coherent mapping and ripple mapping showed a higher diagnostic rate in patients who underwent cardiac surgery than those with Coherent mapping alone (64% vs. 88%, p = .04). CONCLUSION: Coherent mapping combined with RM was superior to coherent mapping alone in identifying the mechanism of scar-related ATs post-cardiac surgery and non-anatomic reentrant ATs.

A novel endocardial activation-derived predictor for focal paraseptal atrial tachycardias: Evidence from a multicenter cohort study.

BACKGROUND: The electrocardiogram-based algorithm for predicting paraseptal atrial tachycardia (PSAT) is limited by the significant overlaps in P-wave morphology originating from various paraseptal sites. OBJECTIVES: The goals of this study were to investigate the endocardial activation characteristics of PSAT and to seek an endocardial activation-derived predictor for the ablation site. METHODS: Forty-four patients [11 men (25%); mean age 62.6 ± 14.7 years] with PSAT ablation in 4 tertiary medical centers were assigned to 3 groups according to the ablation site: right atrial (RA) para-Hisian region (group 1, n = 10), noncoronary cusp (NCC) (group 2, n = 13), and left atrial (LA) paraseptal area (group 3, n = 21). Multiple-chamber activation mapping was performed guided by a 3-dimensional navigation system. The discrepancies in the earliest activation time between 2 of 3 chambers (ΔRA-LA, ΔRA-NCC, and ΔLA-NCC) were calculated in each group and used for pairwise comparisons. RESULTS: There was a significant difference in ΔRA-LA, ΔRA-NCC, and ΔLA-NCC among the 3 groups. ΔRA-LA was the only parameter that could consistently predict the ablation site of PSAT with good accuracy (area under the curve 1.000, sensitivity 100% and specificity 100%, and cutoff value 7 ms for predicting right para-Hisian or NCC ablation; area under the curve 0.974, sensitivity 92.3% and specificity 95.2%, and cutoff value -4 ms for predicting NCC or left paraseptal ablation). Based on 2 cutoff values, a 2-step algorithm was developed to predict the ablation site of PSAT with a positive predictive value of 95.4% and a negative predictive value of 97.0%. CONCLUSION: ΔRA-LA is a useful endocardial activation-derived parameter for predicting the successful ablation site of PSAT.

Identifying significant structural factors associated with knee pain severity in patients with osteoarthritis using machine learning.

Our main objective was to use machine learning methods to identify significant structural factors associated with pain severity in knee osteoarthritis patients. Additionally, we assessed the potential of various classes of imaging data using machine learning techniques to gauge knee pain severity. The data of semi-quantitative assessments of knee radiographs, semi-quantitative assessments of knee magnetic resonance imaging (MRI), and MRI images from 567 individuals in the Osteoarthritis Initiative (OAI) were utilized to train a series of machine learning models. Models were constructed using five machine learning methods: random forests (RF), support vector machines (SVM), logistic regression (LR), decision tree (DT), and Bayesian (Bayes). Employing tenfold cross-validation, we selected the best-performing models based on the area under the curve (AUC). The study results indicate no significant difference in performance among models using different imaging data. Subsequently, we employed a convolutional neural network (CNN) to extract features from magnetic resonance imaging (MRI), and class activation mapping (CAM) was utilized to generate saliency maps, highlighting regions associated with knee pain severity. A radiologist reviewed the images, identifying specific lesions colocalized with the CAM. The review of 421 knees revealed that effusion/synovitis (30.9%) and cartilage loss (30.6%) were the most frequent abnormalities associated with pain severity. Our study suggests cartilage loss and synovitis/effusion lesions as significant structural factors affecting pain severity in patients with knee osteoarthritis. Furthermore, our study highlights the potential of machine learning for assessing knee pain severity using radiographs.

CT-Net: an interpretable CNN-Transformer fusion network for fNIRS classification.

Functional near-infrared spectroscopy (fNIRS), an optical neuroimaging technique, has been widely used in the field of brain activity recognition and brain-computer interface. Existing works have proposed deep learning-based algorithms for the fNIRS classification problem. In this paper, a novel approach based on convolutional neural network and Transformer, named CT-Net, is established to guide the deep modeling for the classification of mental arithmetic (MA) tasks. We explore the effect of data representations, and design a temporal-level combination of two raw chromophore signals to improve the data utilization and enrich the feature learning of the model. We evaluate our model on two open-access datasets and achieve the classification accuracy of 98.05% and 77.61%, respectively. Moreover, we explain our model by the gradient-weighted class activation mapping, which presents a high consistent between the contributing value of features learned by the model and the mapping of brain activity in the MA task. The results suggest the feasibility and interpretability of CT-Net for decoding MA tasks.

Optimizing vitiligo diagnosis with ResNet and Swin transformer deep learning models: a study on performance and interpretability.

Vitiligo is a hypopigmented skin disease characterized by the loss of melanin. The progressive nature and widespread incidence of vitiligo necessitate timely and accurate detection. Usually, a single diagnostic test often falls short of providing definitive confirmation of the condition, necessitating the assessment by dermatologists who specialize in vitiligo. However, the current scarcity of such specialized medical professionals presents a significant challenge. To mitigate this issue and enhance diagnostic accuracy, it is essential to build deep learning models that can support and expedite the detection process. This study endeavors to establish a deep learning framework to enhance the diagnostic accuracy of vitiligo. To this end, a comparative analysis of five models including ResNet (ResNet34, ResNet50, and ResNet101 models) and Swin Transformer series (Swin Transformer Base, and Swin Transformer Large models), were conducted under the uniform condition to identify the model with superior classification capabilities. Moreover, the study sought to augment the interpretability of these models by selecting one that not only provides accurate diagnostic outcomes but also offers visual cues highlighting the regions pertinent to vitiligo. The empirical findings reveal that the Swin Transformer Large model achieved the best performance in classification, whose AUC, accuracy, sensitivity, and specificity are 0.94, 93.82%, 94.02%, and 93.5%, respectively. In terms of interpretability, the highlighted regions in the class activation map correspond to the lesion regions of the vitiligo images, which shows that it effectively indicates the specific category regions associated with the decision-making of dermatological diagnosis. Additionally, the visualization of feature maps generated in the middle layer of the deep learning model provides insights into the internal mechanisms of the model, which is valuable for improving the interpretability of the model, tuning performance, and enhancing clinical applicability. The outcomes of this study underscore the significant potential of deep learning models to revolutionize medical diagnosis by improving diagnostic accuracy and operational efficiency. The research highlights the necessity for ongoing exploration in this domain to fully leverage the capabilities of deep learning technologies in medical diagnostics.

Ventricular dyssynchrony imaging, echocardiographic and clinical outcomes of left bundle branch pacing and biventricular pacing.

BACKGROUND: Left bundle branch pacing (LBBP) is a novel physiological pacing technique which may serve as an alternative to cardiac resynchronization therapy (CRT) by biventricular pacing (BVP). This study assessed ventricular activation patterns and echocardiographic and clinical outcomes of LBBP and compared this to BVP. METHODS: Fifty consecutive patients underwent LBBP or BVP for CRT. Ventricular activation mapping was obtained by ultra-high-frequency ECG (UHF-ECG). Functional and echocardiographic outcomes and hospitalization for heart failure and all-cause mortality after one year from implantation were evaluated. RESULTS: LBBP resulted in greater resynchronization vs BVP (QRS width: 170 ± 16 ms to 128 ± 20 ms vs 174 ± 15 to 144 ± 17 ms, p = 0.002 (LBBP vs BVP); e-DYS 81 ± 17 ms to 0 ± 32 ms vs 77 ± 18 to 16 ± 29 ms, p = 0.016 (LBBP vs BVP)). Improvement in LVEF (from 28 ± 8 to 42 ± 10 percent vs 28 ± 9 to 36 ± 12 percent, LBBP vs BVP, p = 0.078) was similar. Improvement in NYHA function class (from 2.4 to 1.5 and from 2.3 to 1.5 (LBBP vs BVP)), hospitalization for heart failure and all-cause mortality were comparable in both groups. CONCLUSIONS: Ventricular dyssynchrony imaging is an appropriate way to gain a better insight into activation patterns of LBBP and BVP. LBBP resulted in greater resynchronization (e-DYS and QRS duration) with comparable improvement in LVEF, NYHA functional class, hospitalization for heart failure and all-cause mortality at one year of follow up.

3D CNN-based Deep Learning Model-based Explanatory Prognostication in Patients with Multiple Myeloma using Whole-body MRI.

Although magnetic resonance imaging (MRI) data of patients with multiple myeloma (MM) are used to predict prognosis, few reports have applied artificial intelligence (AI) techniques for this purpose. We aimed to analyze whole-body diffusion-weighted MRI data using three-dimensional (3D) convolutional neural networks (CNNs) and Gradient-weighted Class Activation Mapping (Grad-CAM), an explainable AI, to predict prognosis and explore the factors involved in prediction. We retrospectively analyzed the MRI data of a total of 142 patients with MM obtained from two medical centers. We defined the occurrence of progressive disease after MRI evaluation within 12 months as a poor prognosis and constructed a 3D CNN-based deep learning model to predict prognosis. Images from 111 cases were used as the training and internal validation data; images from 31 cases were used as the external validation data. Internal validation of the AI model with stratified 5-fold cross-validation resulted in a significant difference in progression-free survival (PFS) between good and poor prognostic cases (2-year PFS, 91.2% versus [vs.] 61.1%, P = 0.0002). The AI model clearly stratified good and poor prognostic cases in the external validation cohort (2-year PFS, 92.9% vs. 55.6%, P = 0.004), with an area under the receiver operating characteristic curve of 0.804. According to Grad-CAM, the MRI signals of the spleen and bones of the vertebrae and pelvis contributed to prognosis prediction. This study is the first to show that image analysis of whole-body MRI using a 3D CNN without any other clinical data is effective in predicting the prognosis of patients with MM.

Classification of Hippocampal Ripples: Convolutional Neural Network Learns Episode-Specific Changes.

The hippocampus is known to play an important role in memory by processing spatiotemporal information of episodic experiences. By recording synchronized multiple-unit firing events (ripple firings with 300 Hz-10 kHz) of hippocampal CA1 neurons in freely moving rats, we previously found an episode-dependent diversity in the waveform of ripple firings. In the present study, we hypothesized that changes in the diversity would depend on the type of episode experienced. If this hypothesis holds, we can identify the ripple waveforms associated with each episode. Thus, we first attempted to classify the ripple firings measured from rats into five categories: those experiencing any of the four episodes and those before experiencing any of the four episodes. In this paper, we construct a convolutional neural network (CNN) to classify the current stocks of ripple firings into these five categories and demonstrate that the CNN can successfully classify the ripple firings. We subsequently indicate partial ripple waveforms that the CNN focuses on for classification by applying gradient-weighted class activation mapping (Grad-CAM) to the CNN. The method of t-distributed stochastic neighbor embedding (t-SNE) maps ripple waveforms into a two-dimensional feature space. Analyzing the distribution of partial waveforms extracted by Grad-CAM in a t-SNE feature space suggests that the partial waveforms may be representative of each category.

Advancements in gene therapy approaches for atrial fibrillation: Targeted delivery, mechanistic insights and future prospects.

Atrial fibrillation (AF) remains a complex and challenging arrhythmia to treat, necessitating innovative therapeutic strategies. This review explores the evolving landscape of gene therapy for AF, focusing on targeted delivery methods, mechanistic insights, and future prospects. Direct myocardial injection, reversible electroporation, and gene painting techniques are discussed as effective means of delivering therapeutic genes, emphasizing their potential to modulate both structural and electrical aspects of the AF substrate. The importance of identifying precise targets for gene therapy, particularly in the context of AF-associated genetic, structural, and electrical abnormalities, is highlighted. Current studies employing animal models, such as mice and large animals, provide valuable insights into the efficacy and limitations of gene therapy approaches. The significance of imaging methods for detecting atrial fibrosis and guiding targeted gene delivery is underscored. Activation mapping techniques offer a nuanced understanding of AF-specific mechanisms, enabling tailored gene therapy interventions. Future prospects include the integration of advanced imaging, activation mapping, and percutaneous catheter-based techniques to refine transendocardial gene delivery, with potential applications in both ventricular and atrial contexts. As gene therapy for AF progresses, bridging the translational gap between preclinical models and clinical applications is imperative for the successful implementation of these promising approaches.

Successful high-resolution three-dimensional electroanatomical mapping and radiofrequency catheter ablation of a posteroseptal accessory pathway in a dog using CARTO 3.

A one-year-and-seven-month-old, 28 kg, male castrated crossbreed dog was presented for supraventricular tachycardia causing recurrent episodes of anorexia and lethargy. Sotalol (2.2 mg/kg q12 h) reduced the frequency of symptomatic episodes but did not provide full relief. Three-dimensional electroanatomical mapping was performed at the Ghent University Small Animal Teaching hospital using the CARTO 3. Right atrial activation mapping identified the earliest atrial activation right posteroseptal, near the tricuspid annulus. Fast retrograde ventriculoatrial conduction during tachycardia and extrastimulus testing confirmed the presence of a concealed right posteroseptal accessory pathway. Six radiofrequency catheter ablation applications were delivered, and tachycardia remained uninducible. The dog recovered well from the procedure. Sotalol was stopped three weeks later, and no more clinical signs were noted by the owner. Repeated 24-hour electrocardiography monitoring on day one and at one, three, and 12 months after the procedure showed no recurrence of tachycardia.

Using Vision Transformer for high robustness and generalization in predicting EGFR mutation status in lung adenocarcinoma.

BACKGROUND: Lung adenocarcinoma is a common cause of cancer-related deaths worldwide, and accurate EGFR genotyping is crucial for optimal treatment outcomes. Conventional methods for identifying the EGFR genotype have several limitations. Therefore, we proposed a deep learning model using non-invasive CT images to predict EGFR mutation status with robustness and generalizability. METHODS: A total of 525 patients were enrolled at the local hospital to serve as the internal data set for model training and validation. In addition, a cohort of 30 patients from the publicly available Cancer Imaging Archive Data Set was selected for external testing. All patients underwent plain chest CT, and their EGFR mutation status labels were categorized as either mutant or wild type. The CT images were analyzed using a self-attention-based ViT-B/16 model to predict the EGFR mutation status, and the model's performance was evaluated. To produce an attention map indicating the suspicious locations of EGFR mutations, Grad-CAM was utilized. RESULTS: The ViT deep learning model achieved impressive results, with an accuracy of 0.848, an AUC of 0.868, a sensitivity of 0.924, and a specificity of 0.718 on the validation cohort. Furthermore, in the external test cohort, the model achieved comparable performances, with an accuracy of 0.833, an AUC of 0.885, a sensitivity of 0.900, and a specificity of 0.800. CONCLUSIONS: The ViT model demonstrates a high level of accuracy in predicting the EGFR mutation status of lung adenocarcinoma patients. Moreover, with the aid of attention maps, the model can assist clinicians in making informed clinical decisions.

Fusion of transfer learning models with LSTM for detection of breast cancer using ultrasound images.

Breast Cancer (BC) is one of the top reasons for fatality in women worldwide. As a result, timely identification is critical for successful therapy and excellent survival rates. Transfer Learning (TL) approaches have recently shown promise in aiding in the early recognition of BC. In this work, three TL models, MobileNetV2, ResNet50, and VGG16, were combined with LSTM to extract the features from Ultrasound Images (USIs). Furthermore, the Synthetic Minority Over-sampling Technique (SMOTE) with Tomek (SMOTETomek) was employed to balance the extracted features. The proposed method with VGG16 achieved an F1 score of 99.0 %, Matthews Correlation Coefficient (MCC) and Kappa Coefficient of 98.9 % with an Area Under Curve (AUC) of 1.0. The K-fold method was applied for cross-validation and achieved an average F1 score of 96 %. Moreover, the Gradient-weighted Class Activation Mapping (Grad-CAM) method was applied for visualization, and the Local Interpretable Model-agnostic Explanations (LIME) method was applied for interpretability. The Normal Approximation Interval (NAI) and bootstrapping methods were used to calculate Confidence Intervals (CIs). The proposed method achieved a Lower CI (LCI), Upper CI (UCI), and Mean CI (MCI) of 96.50 %, 99.75 %, and 98.13 %, respectively, with the NAI, while 95 % LCI of 93.81 %, an UCI of 96.00 %, and a bootstrap mean of 94.90 % with the bootstrap method. Furthermore, the performance of the six state-of-the-art (SOTA) TL models, such as Xception, NASNetMobile, InceptionResNetV2, MobileNetV2, ResNet50, and VGG16, were compared with the proposed method.