Fine-tuned CNN Research Articles

A Unified Pipeline for Simultaneous Brain Tumor Classification and Segmentation Using Fine-Tuned CNN and Residual UNet Architecture.

In this paper, I present a comprehensive pipeline integrating a Fine-Tuned Convolutional Neural Network (FT-CNN) and a Residual-UNet (RUNet) architecture for the automated analysis of MRI brain scans. The proposed system addresses the dual challenges of brain tumor classification and segmentation, which are crucial tasks in medical image analysis for precise diagnosis and treatment planning. Initially, the pipeline preprocesses the FigShare brain MRI image dataset, comprising 3064 images, by normalizing and resizing them to achieve uniformity and compatibility with the model. The FT-CNN model then classifies the preprocessed images into distinct tumor types: glioma, meningioma, and pituitary tumor. Following classification, the RUNet model performs pixel-level segmentation to delineate tumor regions within the MRI scans. The FT-CNN leverages the VGG19 architecture, pre-trained on large datasets and fine-tuned for specific tumor classification tasks. Features extracted from MRI images are used to train the FT-CNN, demonstrating robust performance in discriminating between tumor types. Subsequently, the RUNet model, inspired by the U-Net design and enhanced with residual blocks, effectively segments tumors by combining high-resolution spatial information from the encoding path with context-rich features from the bottleneck. My experimental results indicate that the integrated pipeline achieves high accuracy in both classification (96%) and segmentation tasks (98%), showcasing its potential for clinical applications in brain tumor diagnosis. For the classification task, the metrics involved are loss, accuracy, confusion matrix, and classification report, while for the segmentation task, the metrics used are loss, accuracy, Dice coefficient, intersection over union, and Jaccard distance. To further validate the generalizability and robustness of the integrated pipeline, I evaluated the model on two additional datasets. The first dataset consists of 7023 images for classification tasks, expanding to a four-class dataset. The second dataset contains approximately 3929 images for both classification and segmentation tasks, including a binary classification scenario. The model demonstrated robust performance, achieving 95% accuracy on the four-class task and high accuracy (96%) in the binary classification and segmentation tasks, with a Dice coefficient of 95%.

Knowledge discovery in weather forecasting: mining fuzzy image association rules with fine-tuned CNN and fuzzy HIFP algorithm

Knowledge discovery in weather forecasting: mining fuzzy image association rules with fine-tuned CNN and fuzzy HIFP algorithm

Gradient aggregation based fine-grained image retrieval: A unified viewpoint for CNN and Transformer

The gradients of CNN are traditionally utilized for optimization and visualization. In this paper, we find that a discriminative representation hides in the gradients of convolution filters. Based on this, we propose a corresponding feature extraction and aggregation method for fine-grained image retrieval (FGIR). Firstly, we propose a metric to evaluate manually-designed loss functions and design a loss function originating from Grad-CAM in the testing phase based on it to extract the gradients of the convolution filters. Secondly, we take the gradients as the new features and design a succinct approach to aggregate them into a compact vector, which is named as Convolution Filters Gradient Aggregation (CFGA) feature. CFGA features can be extracted from pre-trained and fine-tuned CNN models. Extensive experiments are conducted on FGIR to verify the effectiveness of our proposed CFGA approach, compared with five supervised state-of-the-art methods and two unsupervised methods on two standard fine-grained retrieval datasets. Moreover, we generalize the CFGA method designed for CNN to Swin Transformer, and propose the Transformer parameter gradients aggregation (TPGA) method, which proves the applicability of the core idea of CFGA/TPGA to mainstream feature extraction models. We achieve state-of-the-art FGIR performance on CUB-200-2011 dataset and CARS196 dataset.

Automated breast tumor ultrasound image segmentation with hybrid UNet and classification using fine-tuned CNN model

IntroductionBreast cancer stands as the second most deadly form of cancer among women worldwide. Early diagnosis and treatment can significantly mitigate mortality rates. PurposeThe study aims to classify breast ultrasound images into benign and malignant tumors. This approach involves segmenting the breast's region of interest (ROI) employing an optimized UNet architecture and classifying the ROIs through an optimized shallow CNN model utilizing an ablation study. MethodSeveral image processing techniques are utilized to improve image quality by removing text, artifacts, and speckle noise, and statistical analysis is done to check the enhanced image quality is satisfactory. With the processed dataset, the segmentation of breast tumor ROI is carried out, optimizing the UNet model through an ablation study where the architectural configuration and hyperparameters are altered. After obtaining the tumor ROIs from the fine-tuned UNet model (RKO-UNet), an optimized CNN model is employed to classify the tumor into benign and malignant classes. To enhance the CNN model's performance, an ablation study is conducted, coupled with the integration of an attention unit. The model's performance is further assessed by classifying breast cancer with mammogram images. ResultThe proposed classification model (RKONet-13) results in an accuracy of 98.41 %. The performance of the proposed model is further compared with five transfer learning models for both pre-segmented and post-segmented datasets. K-fold cross-validation is done to assess the proposed RKONet-13 model's performance stability. Furthermore, the performance of the proposed model is compared with previous literature, where the proposed model outperforms existing methods, demonstrating its effectiveness in breast cancer diagnosis. Lastly, the model demonstrates its robustness for breast cancer classification, delivering an exceptional performance of 96.21 % on a mammogram dataset. ConclusionThe efficacy of this study relies on image pre-processing, segmentation with hybrid attention UNet, and classification with fine-tuned robust CNN model. This comprehensive approach aims to determine an effective technique for detecting breast cancer within ultrasound images.

Automatic type Identification of Sub - Axial Vertebral Column Fractures using Deep learning Technique

Traumas like falls, sports injuries, or vehicle accidents are the main causes of Vertebral Column Fracture(VCFs) or spine fracture. The symptoms of VCF include back pain, swelling, numbness, and height decrease. VCF causes extreme pain, paralysis, difficulty in movement, etc. CT scans are inexpensive and effective in providing precise and quick VCF type detection. The determination of type VCF, however subject to inter-observer variability. To address this limitation this work introduces an automatic system for the detection of VCF type based on an ensemble of deep fine tuned models. This can assist the orthopaedicians in type identification of VCF which decreases the image interpretation time and increases the patient care time. For type identification, a total of two fine-tuned CNN architectures are used. Then, to further enhance the type identification performance, we developed an ensemble model of fine tuned VGG16 and ResNet50 deep learning models, which average the outputs of the models during final prediction. We used a dataset of 2820 CT images from eight different VCF types that were collected from Kasturba Medical college, Manipal Academy of Higher education, Manipal. The results shows that the ensemble model performs well compare to individual fine-tuned models with accuracy of 82%, precision of 0.84, recall of 0.82 and F1-score of 0.81 for VCF type identification from CT scans. This study shows that type identification of VCF in CT scans may be successfully accomplished using an ensemble deep transfer learning system with various fine tuned CNN architectures.

Video Scene Detection Using Transformer Encoding Linker Network (TELNet).

This paper introduces a transformer encoding linker network (TELNet) for automatically identifying scene boundaries in videos without prior knowledge of their structure. Videos consist of sequences of semantically related shots or chapters, and recognizing scene boundaries is crucial for various video processing tasks, including video summarization. TELNet utilizes a rolling window to scan through video shots, encoding their features extracted from a fine-tuned 3D CNN model (transformer encoder). By establishing links between video shots based on these encoded features (linker), TELNet efficiently identifies scene boundaries where consecutive shots lack links. TELNet was trained on multiple video scene detection datasets and demonstrated results comparable to other state-of-the-art models in standard settings. Notably, in cross-dataset evaluations, TELNet demonstrated significantly improved results (F-score). Furthermore, TELNet's computational complexity grows linearly with the number of shots, making it highly efficient in processing long videos.

Transfer learning with fine-tuned deep CNN model for COVID-19 diagnosis from chest X-ray images

Transfer learning with fine-tuned deep CNN model for COVID-19 diagnosis from chest X-ray images

Brain Tumor Detection and Classification Using Fine-Tuned CNN with ResNet50 and U-Net Model: A Study on TCGA-LGG and TCIA Dataset for MRI Applications.

Nowadays, brain tumors have become a leading cause of mortality worldwide. The brain cells in the tumor grow abnormally and badly affect the surrounding brain cells. These cells could be either cancerous or non-cancerous types, and their symptoms can vary depending on their location, size, and type. Due to its complex and varying structure, detecting and classifying the brain tumor accurately at the initial stages to avoid maximum death loss is challenging. This research proposes an improved fine-tuned model based on CNN with ResNet50 and U-Net to solve this problem. This model works on the publicly available dataset known as TCGA-LGG and TCIA. The dataset consists of 120 patients. The proposed CNN and fine-tuned ResNet50 model are used to detect and classify the tumor or no-tumor images. Furthermore, the U-Net model is integrated for the segmentation of the tumor regions correctly. The model performance evaluation metrics are accuracy, intersection over union, dice similarity coefficient, and similarity index. The results from fine-tuned ResNet50 model are IoU: 0.91, DSC: 0.95, SI: 0.95. In contrast, U-Net with ResNet50 outperforms all other models and correctly classified and segmented the tumor region.

Unusual Driver Behavior Detection in Videos Using Deep Learning Models.

Anomalous driving behavior detection is becoming more popular since it is vital in ensuring the safety of drivers and passengers in vehicles. Road accidents happen for various reasons, including health, mental stress, and fatigue. It is critical to monitor abnormal driving behaviors in real time to improve driving safety, raise driver awareness of their driving patterns, and minimize future road accidents. Many symptoms appear to show this condition in the driver, such as facial expressions or abnormal actions. The abnormal activity was among the most common causes of road accidents, accounting for nearly 20% of all accidents, according to international data on accident causes. To avoid serious consequences, abnormal driving behaviors must be identified and avoided. As it is difficult to monitor anyone continuously, automated detection of this condition is more effective and quicker. To increase drivers' recognition of their driving behaviors and prevent potential accidents, a precise monitoring approach that detects abnormal driving behaviors and identifies abnormal driving behaviors is required. The most common activities performed by the driver while driving is drinking, eating, smoking, and calling. These types of driver activities are considered in this work, along with normal driving. This study proposed deep learning-based detection models for recognizing abnormal driver actions. This system is trained and tested using a newly created dataset, including five classes. The main classes include Driver-smoking, Driver-eating, Driver-drinking, Driver-calling, and Driver-normal. For the analysis of results, pre-trained and fine-tuned CNN models are considered. The proposed CNN-based model and pre-trained models ResNet101, VGG-16, VGG-19, and Inception-v3 are used. The results are compared by using the performance measures. The results are obtained 89%, 93%, 93%, 94% for pre-trained models and 95% by using the proposed CNN-based model. Our analysis and results revealed that our proposed CNN base model performed well and could effectively classify the driver's abnormal behavior.

Human Activity Recognition Using Wearable Sensors Based on Image Classification

Two-Dimensional Fast Fourier Transform (2-D FFT) and Wigner-Ville Transform (WVT) are two popular transforms applied to find frequency and time-frequency representations, respectively. By using them, signals acquired from different axes or sensors are mapped to these representations, considered 2D images. Based on this opinion, three novel methods are presented in this paper. The first two methods are called basic methods. The 2D images based on the magnitude of the 2-D FFT are constructed in method 1 and a fine-tuned CNN is also applied for classifying. WVT is used for constructing 2D compressed images in method 2, and classifying is done like method 1. Fusing two basic methods is presented as the proposed method. It attains the accuracies of 93.45%, 96.47%, 99.00%, and 98.20% for UCI HAR, MOTIONSENSE, MHEALTH, and WISDM datasets, respectively. Achieved results show the superiority of the proposed method compared with the state-of-the-art approaches.

LungNet22: A Fine-Tuned Model for Multiclass Classification and Prediction of Lung Disease Using X-ray Images.

In recent years, lung disease has increased manyfold, causing millions of casualties annually. To combat the crisis, an efficient, reliable, and affordable lung disease diagnosis technique has become indispensable. In this study, a multiclass classification of lung disease from frontal chest X-ray imaging using a fine-tuned CNN model is proposed. The classification is conducted on 10 disease classes of the lungs, namely COVID-19, Effusion, Tuberculosis, Pneumonia, Lung Opacity, Mass, Nodule, Pneumothorax, and Pulmonary Fibrosis, along with the Normal class. The dataset is a collective dataset gathered from multiple sources. After pre-processing and balancing the dataset with eight augmentation techniques, a total of 80,000 X-ray images were fed to the model for classification purposes. Initially, eight pre-trained CNN models, AlexNet, GoogLeNet, InceptionV3, MobileNetV2, VGG16, ResNet 50, DenseNet121, and EfficientNetB7, were employed on the dataset. Among these, the VGG16 achieved the highest accuracy at 92.95%. To further improve the classification accuracy, LungNet22 was constructed upon the primary structure of the VGG16 model. An ablation study was used in the work to determine the different hyper-parameters. Using the Adam Optimizer, the proposed model achieved a commendable accuracy of 98.89%. To verify the performance of the model, several performance matrices, including the ROC curve and the AUC values, were computed as well.

A Deep Learning-Based Approach to Classification of Baby Sign Language Images

Baby Sign Language is used by hearing parents to hearing infants as a preverbal communication which reduce frustration of parents and accelerated learning in babies, increases parent-child bonding, and lets babies communicate vital information, such as if they are hurt or hungry is known as a Baby Sign Language . In the current research work, a study of various existing sign language has been carried out as literature and then after realizing that there is no dataset available for Baby Sign Language, we have created a static dataset for 311 baby signs, which were classified using a MobileNet V1, pretrained Convolution Neural Network [CNN].The focus of the paper is to analyze the effect of Gradient Descent based optimizers, Adam and its variants, Rmsprop optimizers on fine-tuned pretrained CNN model MobileNet V1 that has been trained using customized dataset. The optimizers are used to train and test on MobileNet for 100 epochs on the dataset created for 311 baby Signs. These 10 optimizers Adadelta, Adam, Adamax, SGD, Adagrad, RMSProp were compared based on their processing time.

Deep CNN-based damage classification of milled rice grains using a high-magnification image dataset

• Damage classification of milled rice grains is demonstrated using fine-tuned deep CNN models. • High-magnification dataset of 8048 images consisting of seven types of rice grain damages is constructed. • Five state-of-the-art CNN models (EfficientNet-B0, ResNet-50, InceptionV3, MobileNetV2, MobileNetV3) used for damage classification. • EfficientNet-B0 is the best performing CNN model with an overall classification accuracy of 98.32 % • EfficientNet-B0 successfully classifies the chalky damaged rice further into three subclasses. Surface quality evaluation of pre-processed rice grains is a key factor in determining their market acceptance, storage stability, processing quality, and the overall customer approval. On one end the conventional methods of surface quality evaluation are time-intensive, subjective, and inconsistent. On the other end, the current methods are limited to either sorting of healthy rice grains from the damaged ones, without classifying the latter, or focusing on segregating the different types of rice. A detailed classification of damage in milled rice grains has been largely unexplored due to the lack of an extensive labelled image dataset and the application of advanced CNN models thereon; that enables quick, accurate, and precise classification by excelling at end-to-end tasks, minimizing pre-processing, and eliminating the need for manual feature extraction. In this study, a machine vision system is developed to first construct a dataset of 8048 high-magnification (4.5 x) images of damaged rice refractions, that are obtained through the on-field collection. The dataset spans across seven damage classes, namely, healthy, full chalky, chalky discolored, half chalky, broken, discolored, and normal damage. Subsequently, five different state-of-the-art memory efficient Deep-CNN models, namely, EfficientNet-B0, ResNet-50, InceptionV3, MobileNetV2, and MobileNetV3 are adopted and fine-tuned to enable damage classification of milled rice grains. Experimental results show that the EfficientNet-B0 is the best performing model in terms of the accuracy, average recall, precision, and F1-score. It achieves an individual class accuracy of 98.33%, 96.51%, 95.45%, 100%, 100%, 99.26%, and 98.72% for healthy, full chalky, chalky discolored, half chalky, broken, discolored, and normal damage class respectively. The EfficientNet-B0 architecture achieves an overall classification accuracy of 98.37 % with a significantly reduced model size (47 MB) and a small prediction time of 0.122 s and can sub-classify the chalky class further into 3 different classes i.e. , full chalky, half chalky, and chalky discolored. Overall, this study demonstrates the Deep CNN architectures applied to a high-magnification image dataset enables the classification of damaged rice grains with high accuracy, which could be utilized as a tool for better and more objective quality assessment of the damaged rice grains at market and trading locations.

Tissue and Tumor Epithelium Classification using Fine-tuned Deep CNN Models

The field of Digital Pathology (DP) has become more interested in automated tissue phenotyping in recent years. Tissue phenotyping may be used to identify colorectal cancer (CRC) and distinguish various cancer types. The information needed to construct automated tissue phenotyping systems has been made available by the introduction of Whole Slide Images (WSIs). One of the typical pathological diagnosis duties for pathologists is the histopathological categorization of epithelial tumors. Artificial intelligence (AI) based computational pathology approaches would be extremely helpful in reducing the pathologists ever-increasing workloads, particularly in areas where access to pathological diagnosis services is limited. Investigating several deep learning models for categorizing the images of tumor epithelium from histology is the initial goal. The varying accuracy ratings that were achieved for the deep learning models on the same database demonstrated that additional elements like pre-processing, data augmentation, and transfer learning techniques might affect the models' capacity to attain better accuracy. The second goal of this publication is to reduce the time taken to classify the tissue and tumor Epithelium. The final goal is to examine and fine-tune the most recent models that have received little to no attention in earlier research. These models were checked by the histology Kather CRC image database's nine classifications (CRC-VAL-HE-7K, NCT-CRC-HE-100K). To identify and recommend the most cutting-edge models for each categorization, these models were contrasted with those from earlier research. The performance and the achievements of the proposed preprocessing workflow and fine-tuned Deep CNN models (Alexnet, GoogLeNet and Inceptionv3) are greater compared to the prevalent methods.

Transfer learning with fine-tuned deep CNN ResNet50 model for classifying COVID-19 from chest X-ray images

COVID-19 cases are putting pressure on healthcare systems all around the world. Due to the lack of available testing kits, it is impractical for screening every patient with a respiratory ailment using traditional methods (RT-PCR). In addition, the tests have a high turn-around time and low sensitivity. Detecting suspected COVID-19 infections from the chest X-ray might help isolate high-risk people before the RT-PCR test. Most healthcare systems already have X-ray equipment, and because most current X-ray systems have already been computerized, there is no need to transfer the samples. The use of a chest X-ray to prioritize the selection of patients for subsequent RT-PCR testing is the motivation of this work. Transfer learning (TL) with fine-tuning on deep convolutional neural network-based ResNet50 model has been proposed in this work to classify COVID-19 patients from the COVID-19 Radiography Database. Ten distinct pre-trained weights, trained on varieties of large-scale datasets using various approaches such as supervised learning, self-supervised learning, and others, have been utilized in this work. Our proposed iNat2021_Mini_SwAV_1k model, pre-trained on the iNat2021 Mini dataset using the SwAV algorithm, outperforms the other ResNet50 TL models. For COVID instances in the two-class (Covid and Normal) classification, our work achieved 99.17% validation accuracy, 99.95% train accuracy, 99.31% precision, 99.03% sensitivity, and 99.17% F1-score. Some domain-adapted (ImageNet_ChestX−ray14) and in-domain (ChexPert, ChestX-ray14) models looked promising in medical image classification by scoring significantly higher than other models.

Quantification of Dynamic Track Stiffness Using Machine Learning

Railway track stiffness is an essential factor influencing the track conditions and long-term deterioration. However, the traditional ways to measure the track stiffness are based on inverse computations using multi-body simulations and/or finite element models, which are time-consuming and at low-speed operation. To overcome these challenges, we propose a convolutional neural network framework to predict the track dynamic stiffness using the accelerations captured by accelerometers mounted on the axle box in real-time. To provide a benefit of computational cost-friendly, a dilated convolutional layer has been added which allows the framework to be applied to a compact device. In our study, a nonlinear finite element model of train-track interactions has been calibrated and used to generate unbiased, full range of data sets of axle box accelerations under various track and operational factors. Subsequently, the simulated data is formatted to three different sample sizes: 250-timesteps, 500-timesteps, and 1,000-time steps. The fine-tuned CNN model is developed based on the three datasets and provides the optimal R squared of 0.94, 0.94, and 0.97. The insights gained from this study can assist the track stiffness measurement in the field with a novel measurement method providing continuous, cost-friendly, fast, and implementable benefits. The quantification of dynamic track stiffness will help track engineers to locate problematic and defective tracks promptly on the vast railway networks such as mud pumping, loss of support, pulverized ballast, and so on.

Melanoma Classification from Dermoscopy Images Using Ensemble of Convolutional Neural Networks

Human skin is the most exposed part of the human body that needs constant protection and care from heat, light, dust, and direct exposure to other harmful radiation, such as UV rays. Skin cancer is one of the dangerous diseases found in humans. Melanoma is a form of skin cancer that begins in the cells (melanocytes) that control the pigment in human skin. Early detection and diagnosis of skin cancer, such as melanoma, is necessary to reduce the death rate due to skin cancer. In this paper, the classification of acral lentiginous melanoma, a type of melanoma with benign nevi, is being carried out. The proposed stacked ensemble method for melanoma classification uses different pre-trained models, such as Xception, Inceptionv3, InceptionResNet-V2, DenseNet121, and DenseNet201, by employing the concept of transfer learning and fine-tuning. The selection of pre-trained CNN architectures for transfer learning is based on models having the highest top-1 and top-5 accuracies on ImageNet. A novel stacked ensemble-based framework is presented to improve the generalizability and increase robustness by fusing fine-tuned pre-trained CNN models for acral lentiginous melanoma classification. The performance of the proposed method is evaluated by experimenting on a Figshare benchmark dataset. The impact of applying different augmentation techniques has also been analyzed through extensive experimentations. The results confirm that the proposed method outperforms state-of-the-art techniques and achieves an accuracy of 97.93%.

BreastNet18: A High Accuracy Fine-Tuned VGG16 Model Evaluated Using Ablation Study for Diagnosing Breast Cancer from Enhanced Mammography Images.

Identification and treatment of breast cancer at an early stage can reduce mortality. Currently, mammography is the most widely used effective imaging technique in breast cancer detection. However, an erroneous mammogram based interpretation may result in false diagnosis rate, as distinguishing cancerous masses from adjacent tissue is often complex and error-prone. Six pre-trained and fine-tuned deep CNN architectures: VGG16, VGG19, MobileNetV2, ResNet50, DenseNet201, and InceptionV3 are evaluated to determine which model yields the best performance. We propose a BreastNet18 model using VGG16 as foundational base, since VGG16 performs with the highest accuracy. An ablation study is performed on BreastNet18, to evaluate its robustness and achieve the highest possible accuracy. Various image processing techniques with suitable parameter values are employed to remove artefacts and increase the image quality. A total dataset of 1442 preprocessed mammograms was augmented using seven augmentation techniques, resulting in a dataset of 11,536 images. To investigate possible overfitting issues, a k-fold cross validation is carried out. The model was then tested on noisy mammograms to evaluate its robustness. Results were compared with previous studies. Proposed BreastNet18 model performed best with a training accuracy of 96.72%, a validating accuracy of 97.91%, and a test accuracy of 98.02%. In contrast to this, VGGNet19 yielded test accuracy of 96.24%, MobileNetV2 77.84%, ResNet50 79.98%, DenseNet201 86.92%, and InceptionV3 76.87%. Our proposed approach based on image processing, transfer learning, fine-tuning, and ablation study has demonstrated a high correct breast cancer classification while dealing with a limited number of complex medical images.

Explainability of Convolutional Neural Networks for Dermatological Diagnosis

Background Convolutional neural networks (CNNs) are regarded as state-of-the-art artificial intelligence (AI) tools for dermatological diagnosis, and they have been shown to achieve expert-level performance when trained on a representative dataset. CNN explainability is a key factor to adopting such techniques in practice and can be achieved using attention maps of the network. However, evaluation of CNN explainability has been limited to visual assessment and remains qualitative, subjective, and time consuming. Objective This study aimed to provide a framework for an objective quantitative assessment of the explainability of CNNs for dermatological diagnosis benchmarks. Methods We sourced 566 images available under the Creative Commons license from two public datasets—DermNet NZ and SD-260, with reference diagnoses of acne, actinic keratosis, psoriasis, seborrheic dermatitis, viral warts, and vitiligo. Eight dermatologists with teledermatology expertise annotated each clinical image with a diagnosis, as well as diagnosis-supporting characteristics and their localization. A total of 16 supporting visual characteristics were selected, including basic terms such as macule, nodule, papule, patch, plaque, pustule, and scale, and additional terms such as closed comedo, cyst, dermatoglyphic disruption, leukotrichia, open comedo, scar, sun damage, telangiectasia, and thrombosed capillary. The resulting dataset consisted of 525 images with three rater annotations for each. Explainability of two fine-tuned CNN models, ResNet-50 and EfficientNet-B4, was analyzed with respect to the reference explanations provided by the dermatologists. Both models were pretrained on the ImageNet natural image recognition dataset and fine-tuned using 3214 images of the six target skin conditions obtained from an internal clinical dataset. CNN explanations were obtained as activation maps of the models through gradient-weighted class-activation maps. We computed the fuzzy sensitivity and specificity of each characteristic attention map with regard to both the fuzzy gold standard characteristic attention fusion masks and the fuzzy union of all characteristics. Results On average, explainability of EfficientNet-B4 was higher than that of ResNet-50 in terms of sensitivity for 13 of 16 supporting characteristics, with mean values of 0.24 (SD 0.07) and 0.16 (SD 0.05), respectively. However, explainability was lower in terms of specificity, with mean values of 0.82 (SD 0.03) and 0.90 (SD 0.00) for EfficientNet-B4 and ResNet-50, respectively. All measures were within the range of corresponding interrater metrics. Conclusions We objectively benchmarked the explainability power of dermatological diagnosis models through the use of expert-defined supporting characteristics for diagnosis. Acknowledgments This work was supported in part by the Danish Innovation Fund under Grant 0153-00154A. Conflict of Interest None declared.

WBC-CNN: Efficient CNN-Based Models to Classify White Blood Cells Subtypes

Blood is essential to life. The number of blood cells plays a significant role in observing an individual’s health status. Having a lower or higher number of blood cells than normal may be a sign of various diseases. Thus it is important to precisely classify blood cells and count them to diagnose different health conditions. In this paper, we focused on classifying white blood cells subtypes (WBC) which are the basic parts of the immune system. Classification of WBC subtypes is very useful for diagnosing diseases, infections, and disorders. Deep learning technologies have the potential to enhance the process and results of WBC classification. This study presented two fine-tuned CNN models and four hybrid CNN-based models to classify WBC. The VGG-16 and MobileNet are the CNN architectures used for both feature extraction and classification in fine-tuned models. The same CNN architectures are used for feature extraction in hybrid models; however, the Support Vector Machines (SVM) and the Quadratic Discriminant Analysis (QDA) are the classifiers used for classification. Among all models, the fine-tuned VGG-16 performs best, its classification accuracy is 99.81%. Our hybrid models are efficient in detecting WBC as well. 98.44% is the classification accuracy of the VGG-16+SVM model, and 98.19% is the accuracy of the MobileNet+SVM.