Automated Melanoma Detection Research Articles

Automated melanoma detection. An algorithm inspired from human intelligence characterizing disordered pattern of melanocytic lesions improving a convolutional neural network

Automated melanoma detection. An algorithm inspired from human intelligence characterizing disordered pattern of melanocytic lesions improving a convolutional neural network

Performance Improvement of Melanoma Detection Using a Multi-Network System Based on Decision Fusion

The incidence of melanoma cases continues to rise, underscoring the critical need for early detection and treatment. Recent studies highlight the significance of deep learning in melanoma detection, leading to improved accuracy. The field of computer-assisted detection is extensively explored along all lines, especially in the medical industry, as the benefit in this field is to save hu-man lives. In this domain, this direction must be maximally exploited and introduced into routine controls to improve patient prognosis, disease prevention, reduce treatment costs, improve population management, and improve patient empowerment. All these new aspects were taken into consideration to implement an EHR system with an automated melanoma detection system. The first step, as presented in this paper, is to build a system based on the fusion of decisions from multiple neural networks, such as DarkNet-53, DenseNet-201, GoogLeNet, Inception-V3, InceptionResNet-V2, ResNet-50, ResNet-101, and compare this classifier with four other applications: Google Teachable Machine, Microsoft Azure Machine Learning, Google Vertex AI, and SalesForce Einstein Vision based on the F1 score for further integration into an EHR platform. We trained all models on two databases, ISIC 2020 and DermIS, to also test their adaptability to a wide range of images. Comparisons with state-of-the-art research and existing applications confirm the promising performance of the proposed system.

3D Whole-body skin imaging for automated melanoma detection.

Existing artificial intelligence for melanoma detection has relied on analysing images of lesions of clinical interest, which may lead to missed melanomas. Tools analysing the entire skin surface are lacking. To determine if melanoma can be distinguished from other skin lesions using data from automated analysis of 3D-images. Single-centre, retrospective, observational convenience sample of patients diagnosed with melanoma at a tertiary care cancer hospital. Eligible participants were those with a whole-body 3D-image captured within 90 days prior to the diagnostic skin biopsy. 3D-images were obtained as standard of care using VECTRA WB360 Whole Body 3-dimensional Imaging System (Canfield Scientific). Automated data from image processing (i.e. lesion size, colour, border) for all eligible participants were exported from VECTRA DermaGraphix research software for analysis. The main outcome was the area under the receiver operating characteristic curve (AUC). A total of 35 patients contributed 23,538 automatically identified skin lesions >2 mm in largest diameter (102-3021 lesions per participant). All were White patients and 23 (66%) were males. The median (range) age was 64 years (26-89). There were 49 lesions of melanoma and 22,489 lesions that were not melanoma. The AUC for the prediction model was 0.94 (95% CI: 0.92-0.96). Considering all lesions in a patient-level analysis, 14 (28%) melanoma lesions had the highest predicted score or were in the 99th percentile among all lesions for an individual patient. In this proof-of-concept pilot study, we demonstrated that automated analysis of whole-body 3D-images using simple image processing techniques can discriminate melanoma from other skin lesions with high accuracy. Further studies with larger, higher quality, and more representative 3D-imaging datasets would be needed to improve and validate these results.

Incorporating clinical knowledge with constrained classifier chain into a multimodal deep network for melanoma detection

In recent years, vast developments in Computer-Aided Diagnosis (CAD) for skin diseases have generated much interest from clinicians and other eventual end-users of this technology. Introducing clinical domain knowledge to these machine learning strategies can help dispel the black box nature of these tools, strengthening clinician trust. Clinical domain knowledge also provides new information channels which can improve CAD diagnostic performance. In this paper, we propose a novel framework for malignant melanoma (MM) detection by fusing clinical images and dermoscopic images. The proposed method combines a multi-labeled deep feature extractor and clinically constrained classifier chain (CC). This allows the 7-point checklist, a clinician diagnostic algorithm, to be included in the decision level while maintaining the clinical importance of the major and minor criteria in the checklist. Our proposed framework achieved an average accuracy of 81.3% for detecting all criteria and melanoma when testing on a publicly available 7-point checklist dataset. This is the highest reported results, outperforming state-of-the-art methods in the literature by 6.4% or more. Analyses also show that the proposed system surpasses the single modality system of using either clinical images or dermoscopic images alone and the systems without adopting the approach of multi-label and clinically constrained classifier chain. Our carefully designed system demonstrates a substantial improvement over melanoma detection. By keeping the familiar major and minor criteria of the 7-point checklist and their corresponding weights, the proposed system may be more accepted by physicians as a human-interpretable CAD tool for automated melanoma detection.

Automatic segmentation of melanoma using superpixel region growing technique

Melanoma is the most life threatening type of cancer which contributes to the highest mortality rate. Early detection of melanoma facilitates better prognosis and increases survival rates. High infiltration of melanoma and advanced digital imaging technologies have exhilarated concern among the public, calling for initial screenings. However, melanoma screening is considered as a non trivial problem even by expert medical practitioners, in spite of several diagnostic algorithms. There is an enthralling need for automated melanoma detection systems due to the surge in the melanoma population and lack of trained dermatologists. Computational models for Melanoma detection are based on learning from the Region of Interest (RoI). Nevertheless, identification of RoI itself poses several challenges due to the diverse structural and chromatic features on the surface of the skin. This paper proposes a superpixel region growing based approach for segmentation of the melanoma region for further analysis. It is based on the Gaussian Mixture Model superpixels which segment the candidate image into accurate homogenous regions. The superiority of the system is demonstrated with performance metrics and comparisons on a standard dataset. This system is an impending solution to perform melanoma screenings with ease.

Deep Convolutional Neural Network for Melanoma Detection using Dermoscopy Images.

Developing a fast and accurate classifier is an important part of a computer-aided diagnosis system for skin cancer. Melanoma is the most dangerous form of skin cancer which has a high mortality rate. Early detection and prognosis of melanoma can improve survival rates. In this paper, we propose a deep convolutional neural network for automated melanoma detection that is scalable to accommodate a variety of hardware and software constraints. Dermoscopic skin images collected from open sources were used for training the network. The trained network was then tested on a dataset of 2150 malignant or benign images. Overall, the classifier achieved high average values for accuracy, sensitivity, and specificity of 82.95%, 82.99%, and 83.89% respectively. It outperfomed other exisitng networks using the same dataset.

Augmenting data with GANs to segment melanoma skin lesions

This paper presents a novel strategy that employs Generative Adversarial Networks (GANs) to augment data in the skin lesion segmentation task, which is a fundamental first step in the automated melanoma detection process. The proposed framework generates both skin lesion images and their segmentation masks, making the data augmentation process extremely straightforward. In order to thoroughly analyze how the quality and diversity of synthetic images impact the efficiency of the method, we remodel two different well known GANs: a Deep Convolutional GAN (DCGAN) and a Laplacian GAN (LAPGAN). Experimental results reveal that, by introducing such kind of synthetic data into the training process, the overall accuracy of a state-of-the-art Convolutional/Deconvolutional Neural Network for melanoma skin lesion segmentation is increased.

A survey on automated melanoma detection

Skin cancer is defined as the rapid growth of skin cells due to DNA damage that cannot be repaired. Melanoma is one of the deadliest types of skin cancer, which originates from melanocytes. While other skin cancer types have limited spreading capabilities, the danger of melanoma comes from its ability to spread (metastasize) rapidly. Fortunately, it can be detected by visual inspection of the skin surface, and it is 100% curable if identified in the early stages. However, detection by “subjective visual inspection” creates an important problem, due to investigators’ different levels of experiences and education. Dermoscopy (dermatoscopy) has significantly increased the diagnostic accuracy of melanoma since late 90’s. In addition, several systems have been proposed in order to assist investigators or to perform an automatic melanoma detection. This survey focuses on the algorithms for automated melanoma detection in dermoscopic images through an extensive analysis of the stages in methodologies proposed in the literature, and by examining related concepts and describing possible future directions through open problems in this domain of research.

Melanoma Detection using Statistical Texture Distinctiveness Segmentation

is the most dangerous form of skin cancer. It must be detected in the initial stage to increase the survival rates. In medical field, Melanoma detection is usually done by clinical analysis and biopsy tests. These methods are time consuming, expensive and have many side effects. Thus, an automated melanoma detection system is better to assess a patient's risk of melanoma in the initial phase with high accuracy. Existing automated melanoma detection systems make use of thresholding, statistical region merging and Otsu's method for segmentation. These segmentation methods do not include texture analysis, so the accuracy is less. Accuracy of segmentation and melanoma detection can be improved by examining the textural features of skin lesion. Computer aided melanoma detection system using image processing techniques is proposed for accurate and early detection of melanoma. This system has different stages which include preprocessing for image enhancement, segmentation of skin lesion using textural features to improve accuracy, feature extraction and classification. The input image is preprocessed using contrast stretching for image enhancement. The enhanced image is segmented using Texture Distinctiveness Lesion Segmentation (TDLS) algorithm to extract the lesion area from the background skin. Feature extraction is done using graylevel cooccurrence matrix. The system is trained with the extracted features using a good classifier to classify the lesion as malignant or benign melanoma. Accuracy of the proposed system is computed and compared with other segmentation and classification algorithms.

Automated melanoma detection with a novel multispectral imaging system: results of a prospective study

The aim of this research was to evaluate the performance of a new spectroscopic system in the diagnosis of melanoma. This study involves a consecutive series of 1278 patients with 1391 cutaneous pigmented lesions including 184 melanomas. In an attempt to approach the ‘real world’ of lesion population, a further set of 1022 not excised clinically reassuring lesions was also considered for analysis. Each lesion was imaged in vivo by a multispectral imaging system. The system operates at wavelengths between 483 and 950 nm by acquiring 15 images at equally spaced wavelength intervals. From the images, different lesion descriptors were extracted related to the colour distribution and morphology of the lesions. Data reduction techniques were applied before setting up a neural network classifier designed to perform automated diagnosis. The data set was randomly divided into three sets: train (696 lesions, including 90 melanomas) and verify (348 lesions, including 53 melanomas) for the instruction of a proper neural network, and an independent test set (347 lesions, including 41 melanomas). The neural network was able to discriminate between melanomas and non-melanoma lesions with a sensitivity of 80.4% and a specificity of 75.6% in the 1391 histologized cases data set. No major variations were found in classification scores when train, verify and test subsets were separately evaluated. Following receiver operating characteristic (ROC) analysis, the resulting area under the curve was 0.85. No significant differences were found among areas under train, verify and test set curves, supporting the good network ability to generalize for new cases. In addition, specificity and area under ROC curve increased up to 90% and 0.90, respectively, when the additional set of 1022 lesions without histology was added to the test set. Our data show that performance of an automated system is greatly population dependent, suggesting caution in the comparison with results reported in the literature. In our opinion, scientific reports should provide, at least, the median values of thickness and dimension of melanomas, as well as the number of small (⩽6 mm) melanomas.

Optical imaging and automated melanoma detection: questions and answers.

Early detection and prompt excision of cutaneous melanoma is of paramount importance to improve patient survival, and the clinician should be aware of the clinical features that suggest the presence of a malignant lesion. The clinical diagnosis is mainly based on observation of the colour and shape of a given skin lesion. Unfortunately, evaluation of a pigmented lesion is to a large extent subjective and is closely related to the experience of the clinician. To overcome this problem, optical imaging techniques using different instrumentation (i.e. colour video camera, epiluminescence microscopy, reflectance spectrophotometry) and computer image analysis have been proposed in an attempt to provide quantitative measurements in an objective and reproducible fashion. The different procedures employed to perform the diagnosis automatically all have a common denominator: mimicking the eye and the brain of the clinician by image processing and computerized analysis programs, respectively. Sensitivity and specificity data reported in the literature suggest that the computer-based diagnosis of melanoma does not greatly differ from the diagnostic capability of an expert clinician, and is independent of the optical acquisition method employed to analyse the lesions. Most of the computer-processed morphometric variables useful in automated diagnosis are not recognizable nor can be objectively evaluated by the human eye, except that of lesion dimension. However, several questions should be answered before assessing the actual usefulness, including the potential and limitations, of computer-based diagnostic procedures. The purpose of this study was to briefly review the different kinds of instrumentation being used to diagnose melanoma, and to raise questions and whenever possible provide answers in an attempt to establish whether there will be a future for these computerized systems.