Skin lesion computational diagnosis of dermoscopic images: Ensemble models based on input feature manipulation

Skin patch based makeup finish assessment technique by deep neural network

BACKGROUND

Skin color and texture play a significant role in influencing impressions. To understand the influence of skin appearance and to develop better makeup products, objective evaluation methods for makeup finish have been explored. This study aims to apply machine learning technology, specifically deep neural network (DNN), to accurately analyze and evaluate delicate and complex cosmetic skin textures.

METHODS

"Skin patch datasets" were extracted from facial images and used to train a DNN model. The advantages of using skin patches include retaining fine texture, eliminating false correlations from non-skin features, and enabling visualization of the inferred results for the entire face. The DNN was trained in two ways: a classification task to classify skin attributes and a regression task to predict the visual assessment of experts. The trained DNNs were applied for the evaluation of actual makeup conditions.

RESULTS

In the classification task training, skin patch-based classifiers for age range, presence or absence of base makeup, formulation type (powder/liquid) of the applied base makeup, and immediate/while after makeup application were developed. The trained DNNs on regression task showed high prediction accuracy for the experts' visual assessment. Application of DNN to the evaluation of actual makeup conditions clearly showed appropriate evaluation results in line with the appearance of the makeup finish.

CONCLUSION

The proposed method of using DNNs trained on skin patches effectively evaluates makeup finish. This approach has potential applications in visual science research and cosmetics development. Further studies can explore the analysis of different skin conditions and the development of personalized cosmetics.

SBXception: A Shallower and Broader Xception Architecture for Efficient Classification of Skin Lesions

Skin cancer is a major public health concern around the world. Skin cancer identification is critical for effective treatment and improved results. Deep learning models have shown considerable promise in assisting dermatologists in skin cancer diagnosis. This study proposes SBXception: a shallower and broader variant of the Xception network. It uses Xception as the base model for skin cancer classification and increases its performance by reducing the depth and expanding the breadth of the architecture. We used the HAM10000 dataset, which contains 10,015 dermatoscopic images of skin lesions classified into seven categories, for training and testing the proposed model. Using the HAM10000 dataset, we fine-tuned the new model and reached an accuracy of 96.97% on a holdout test set. SBXception also achieved significant performance enhancement with 54.27% fewer training parameters and reduced training time compared to the base model. Our findings show that reducing and expanding the Xception model architecture can greatly improve its performance in skin cancer categorization.

Integrated Design of Optimized Weighted Deep Feature Fusion Strategies for Skin Lesion Image Classification

This study mainly focuses on pre-processing the HAM10000 and BCN20000 skin lesion datasets to select important features that will drive for proper skin cancer classification. In this work, three feature fusion strategies have been proposed by utilizing three pre-trained Convolutional Neural Network (CNN) models, namely VGG16, EfficientNet B0, and ResNet50 to select the important features based on the weights of the features and are coined as Adaptive Weighted Feature Set (AWFS). Then, two other strategies, Model-based Optimized Weighted Feature Set (MOWFS) and Feature-based Optimized Weighted Feature Set (FOWFS), are proposed by optimally and adaptively choosing the weights using a meta-heuristic artificial jellyfish (AJS) algorithm. The MOWFS-AJS is a model-specific approach whereas the FOWFS-AJS is a feature-specific approach for optimizing the weights chosen for obtaining optimal feature sets. The performances of those three proposed feature selection strategies are evaluated using Decision Tree (DT), Naïve Bayesian (NB), Multi-Layer Perceptron (MLP), and Support Vector Machine (SVM) classifiers and the performance are measured through accuracy, precision, sensitivity, and F1-score. Additionally, the area under the receiver operating characteristics curves (AUC-ROC) is plotted and it is observed that FOWFS-AJS shows the best accuracy performance based on the SVM with 94.05% and 94.90%, respectively, for HAM 10000 and BCN 20000 datasets. Finally, the experimental results are also analyzed using a non-parametric Friedman statistical test and the computational times are recorded; the results show that, out of those three proposed feature selection strategies, the FOWFS-AJS performs very well because its quick converging nature is inculcated with the help of AJS.

On the Automatic Detection and Classification of Skin Cancer Using Deep Transfer Learning

Skin cancer (melanoma and non-melanoma) is one of the most common cancer types and leads to hundreds of thousands of yearly deaths worldwide. It manifests itself through abnormal growth of skin cells. Early diagnosis drastically increases the chances of recovery. Moreover, it may render surgical, radiographic, or chemical therapies unnecessary or lessen their overall usage. Thus, healthcare costs can be reduced. The process of diagnosing skin cancer starts with dermoscopy, which inspects the general shape, size, and color characteristics of skin lesions, and suspected lesions undergo further sampling and lab tests for confirmation. Image-based diagnosis has undergone great advances recently due to the rise of deep learning artificial intelligence. The work in this paper examines the applicability of raw deep transfer learning in classifying images of skin lesions into seven possible categories. Using the HAM1000 dataset of dermoscopy images, a system that accepts these images as input without explicit feature extraction or preprocessing was developed using 13 deep transfer learning models. Extensive evaluation revealed the advantages and shortcomings of such a method. Although some cancer types were correctly classified with high accuracy, the imbalance of the dataset, the small number of images in some categories, and the large number of classes reduced the best overall accuracy to 82.9%.

Machine Learning and Deep Learning Methods for Skin Lesion Classification and Diagnosis: A Systematic Review

Computer-aided systems for skin lesion diagnosis is a growing area of research. Recently, researchers have shown an increasing interest in developing computer-aided diagnosis systems. This paper aims to review, synthesize and evaluate the quality of evidence for the diagnostic accuracy of computer-aided systems. This study discusses the papers published in the last five years in ScienceDirect, IEEE, and SpringerLink databases. It includes 53 articles using traditional machine learning methods and 49 articles using deep learning methods. The studies are compared based on their contributions, the methods used and the achieved results. The work identified the main challenges of evaluating skin lesion segmentation and classification methods such as small datasets, ad hoc image selection and racial bias.

Predicting the clinical management of skin lesions using deep learning

Automated machine learning approaches to skin lesion diagnosis from images are approaching dermatologist-level performance. However, current machine learning approaches that suggest management decisions rely on predicting the underlying skin condition to infer a management decision without considering the variability of management decisions that may exist within a single condition. We present the first work to explore image-based prediction of clinical management decisions directly without explicitly predicting the diagnosis. In particular, we use clinical and dermoscopic images of skin lesions along with patient metadata from the Interactive Atlas of Dermoscopy dataset (1011 cases; 20 disease labels; 3 management decisions) and demonstrate that predicting management labels directly is more accurate than predicting the diagnosis and then inferring the management decision ([Formula: see text] and [Formula: see text] improvement in overall accuracy and AUROC respectively), statistically significant at [Formula: see text]. Directly predicting management decisions also considerably reduces the over-excision rate as compared to management decisions inferred from diagnosis predictions (24.56% fewer cases wrongly predicted to be excised). Furthermore, we show that training a model to also simultaneously predict the seven-point criteria and the diagnosis of skin lesions yields an even higher accuracy (improvements of [Formula: see text] and [Formula: see text] in overall accuracy and AUROC respectively) of management predictions. Finally, we demonstrate our model's generalizability by evaluating on the publicly available MClass-D dataset and show that our model agrees with the clinical management recommendations of 157 dermatologists as much as they agree amongst each other.

Melanoma Detection Using Spatial and Spectral Analysis on Superpixel Graphs

Melanoma is the most fatal type of skin cancer. Detection of melanoma from dermoscopic images in an early stage is critical for improving survival rates. Numerous image processing methods have been devised to discriminate between melanoma and benign skin lesions. Previous studies show that the detection performance depends significantly on the skin lesion image representations and features. In this work, we propose a melanoma detection approach that combines graph-theoretic representations with conventional dermoscopic image features to enhance the detection performance. Instead of using individual pixels of skin lesion images as nodes for complex graph representations, superpixels are generated from the skin lesion images and are then used as graph nodes in a superpixel graph. An edge of such a graph connects two adjacent superpixels where the edge weight is a function of the distance between feature descriptors of these superpixels. A graph signal can be defined by assigning to each graph node the output of some single-valued function of the associated superpixel descriptor. Features are extracted from weighted and unweighted graph models in the vertex domain at both local and global scales and in the spectral domain using the graph Fourier transform (GFT). Other features based on color, geometry and texture are extracted from the skin lesion images. Several conventional and ensemble classifiers have been trained and tested on different combinations from those features using two datasets of dermoscopic images from the International Skin Imaging Collaboration (ISIC) archive. The proposed system achieved an AUC of [Formula: see text], an accuracy of [Formula: see text], a specificity of [Formula: see text] and a sensitivity of [Formula: see text].

Detection of skin cancer with adaptive fuzzy classifier using improved whale optimization

Skin cancer is considered as a well-known type of cancer globally, and its occurrence has been found to be raised in current days. Researchers state that the disease requires early prediction so that the identification of precise signs will make it simple for the dermatologists and clinicians. This disorder has been established to be unpredictable. Hence, this paper intends to develop an efficient skin cancer detection scheme, which classifies the nature of cancer, whether it is normal, benign or malignant. Accordingly, the skin image which is given as input is segmented using k-means clustering model and the features are extracted from segmented image using Local Vector Pattern (LVP). Moreover, the extracted features are subjected to fuzzy classifier for recognizing the cancer. In addition, the limits of membership functions are optimally selected by improved Whale Optimization Algorithm (WOA). Thus, the proposed scheme is termed as Improved Selection of Encircling and Spiral updating position of WO-based Fuzzy Classifier (ISESW-FC). From the optimized output, the type of skin cancer image can be determined, whether it is normal, benign or malignant. The performance of proposed model is compared over other conventional methods, and its efficiency is proved by means of Type I and Type II measures.

Kernel sparse representation based model for skin lesions segmentation and classification

BACKGROUND AND OBJECTIVES

Melanoma is a dangerous kind of skin disease with a high death rate, and its prevalence has increased rapidly in recent years. Diagnosis of melanoma in a primary phase can be helpful for its cure. Due to costs for dermatology, we need an automatic system to diagnose melanoma through lesion images.

METHODS

Here, we propose a sparse representation based method for segmentation and classification of lesion images. The main idea of our framework is based on a kernel sparse representation, which produces discriminative sparse codes to represent features in a high-dimensional feature space. Our novel formulation for discriminative kernel sparse coding jointly learns a kernel-based dictionary and a linear classifier. We also present an adaptive K-SVD algorithm for kernel dictionary and classifier learning.

RESULTS

We test our approach for both segmentation and classification tasks. The evaluation results on both dermoscopic and digital datasets demonstrate our approach to be competitive as compared to the available state-of-the-art methods, with the advantage of not needing any pre-processing.

CONCLUSIONS

Our method is insensitive to noise and image conditions and can be used effectively for challenging skin lesions. Our approach is so extensive to be adapted to various medical image segmentations.

Computational texture features of dermoscopic images and their link to the descriptive terminology: A survey

Computer-extracted texture features are relevant to diagnose cutaneous lesions such as melanomas. Our goal is to set a relationship between a well-established descriptive terminology, which describes the attributes of dermoscopic structures based on their aspect rather than their underlying causes, and the computational methods to extract texture-based features. By tackling this problem, we can ascertain what indicators used by dermatologists are reflected in the extracted texture features. We first review the state-of-the-art models for texture extraction in dermoscopic images. By comparing the methods' performance and goals, we conclude that (I) a single color space does not seem to give performances as good as using several ones, thus the latter is reasonable (II) the optimal number of extracted features seems to vary depending on the method's goal, and extracting a large number of features can lead to a loss of models robustness (III) methods such as GLCM, Sobel or Law energy filters are mainly used to capture local properties to detect specific dermoscopic structures (IV) methods that extract local and global features, like Gabor wavelets or SPT, tend to be used to analyze the presence of certain patterns of dermoscopic structures, e.g. globular, reticular, etc.

Integration of morphological preprocessing and fractal based feature extraction with recursive feature elimination for skin lesion types classification

BACKGROUND AND OBJECTIVE

Skin cancer is the commonest form of cancer in the worldwide population. Non-invasive and non-contact imaging modalities are being used for the screening of melanoma and other cutaneous malignancies to endorse early detection and prevention of the disease. Traditionally it has been a problem for medical personnel to differentiate melanoma, dysplastic nevi and basal cell carcinoma (BCC) diseases from one another due to the confusing appearance and similarity in the characteristics of the pigmented lesions. The paper reports an integrated method developed for identifying these skin diseases from the dermoscopic images.

METHODS

The proposed integrated computer-aided method has been employed for the identification of each of these diseases using recursive feature elimination (RFE) based layered structured multiclass image classification technique. Prior to the classification, different quantitative features have been extracted by analyzing the shape, the border irregularity, the texture and the color of the skin lesions, using different image processing tools. Primarily, a combination of gray level co-occurrence matrix (GLCM) and a proposed fractal-based regional texture analysis (FRTA) algorithm has been used for the quantification of textural information. The performance of the framework has been evaluated using a layered structure classification model using support vector machine (SVM) classifier with radial basis function (RBF).

RESULTS

The performance of the morphological skin lesion segmentation algorithm has been evaluated by estimating the pixel level sensitivity (Sen) of 0.9172, 0.9788 specificity (Spec), 0.9521 accuracy (ACU), along with the image similarity measuring indices as Jaccard similarity index (JSI) of 0.8562 and Dice similarity coefficient (DSC) of 0.9142 with respect to the corresponding ground truth (GT) images. The quantitative features extracted from the proposed feature extraction algorithms have been employed for the proposed multi-class skin disease identification. The proposed layered structure identifies all the three classes of skin diseases with a highly acceptable classification accuracy of 98.99%, 97.54% and 99.65% for melanoma, dysplastic nevi and BCC respectively.

CONCLUSION

To overcome the difficulties of proper diagnosis of diseases based on visual evaluation, the proposed integrated system plays an important role by quantifying the effective features and identifying the diseases with higher degree of accuracy. This combined approach of quantitative and qualitative analysis not only increases the diagnostic accuracy, but also provides some important information not obtainable from qualitative assessment alone.

Melanoma Detection in Dermoscopic Images using Color Features

Color is considered to be a major characteristic feature that is used for distinguishing benign and malignant melanocytic lesions. Most of malignant melanomas are characterized by the presence of six suspicious colors inspired from the ABCD dermoscopic rule. The presence of these suspicious colors histopathologically indicates the presence of melanin in the deeper layers of the epidermis and dermis. The objective of the proposed work is to evaluate the role of color features, a set of fifteen color features have been extracted from the region of interest to determine the role of color in malignancy detection. Further, a set of ensemble classifiers with dynamic selection techniques are used for classification of the extracted features, yielding an average accuracy of 87.5% for classifying benign and malignant lesions.

Melanoma recognition in dermoscopy images using lesion's peripheral region information

BACKGROUND AND OBJECTIVES

Melanoma is one of the most dangerous forms of skin cancer, but it has a high survival rate if diagnosed on time. The first diagnostic approach in melanoma recognition is to visually assess the lesion through dermoscopic images. Computer-aided diagnosis systems for melanoma recognition has attracted a lot of attention in the last decade and proved to be helpful in that area. Methods for skin lesions analysis usually involves three main steps: lesion segmentation, feature extraction, and features classification. Extracting highly discriminative features from the lesion has a great impact on the recognition task. In this paper, we are seeking a lesion recognition system that incorporates these highly discriminative features.

METHODS

For segmentation step, we use contour propagation model with a novel two-component speed function. In the feature extraction step, a new set of features based on peripheral information of the lesion are introduced. For this end, the peripheral area of the lesion is mapped to log-polar space using the Daugman's transformation and then a set of texture features are extracted from it. Newly introduced features do not need further segmentation of dermoscopic structures and are robust against lesion's scale, orientation, location, and shape variation. We also design the other global texture features to describe only the information from the lesion area. In the classification step, we evaluated two different schemes to prove the distinction power of the new features, one comprises linear SVM to recognize melanoma vs. nevus and the other scheme uses RUSBoost classifier to recognize melanoma vs. nevus and atypical-nevus. Sequential feature selection algorithm has been utilized in each classification scheme to rank features based on their distinction power.

RESULTS

Cross-validation experiments on the well-known PH² dataset resulted in an average of 97% for sensitivity and 100% for specificity on melanoma vs. nevus recognition task using only four features. Also, in the second classification scheme, we achieved high sensitivity and specificity values of 95% for melanoma vs. nevus and atypical nevus recognition experiments.

CONCLUSION

High values for evaluation metrics show that the proposed melanoma recognition system is superior to the other state-of-the-art algorithms, which proves the high distinction power of the newly introduced features.