A Transfer Learning Approach for Clinical Detection Support of Monkeypox Skin Lesions

Abstract

Monkeypox (MPX) is a disease caused by monkeypox virus (MPXV). It is a contagious disease and has associated symptoms of skin lesions, rashes, fever, and respiratory distress lymph swelling along with numerous neurological distresses. This can be a deadly disease, and the latest outbreak of it has shown its spread to Europe, Australia, the United States, and Africa. Typically, diagnosis of MPX is performed through PCR, by taking a sample of the skin lesion. This procedure is risky for medical staff, as during sample collection, transmission and testing, they can be exposed to MPXV, and this infectious disease can be transferred to medical staff. In the current era, cutting-edge technologies such as IoT and artificial intelligence (AI) have made the diagnostics process smart and secure. IoT devices such as wearables and sensors permit seamless data collection while AI techniques utilize the data in disease diagnosis. Keeping in view the importance of these cutting-edge technologies, this paper presents a non-invasive, non-contact, computer-vision-based method for diagnosis of MPX by analyzing skin lesion images that are more smart and secure compared to traditional methods of diagnosis. The proposed methodology employs deep learning techniques to classify skin lesions as MPXV positive or not. Two datasets, the Kaggle Monkeypox Skin Lesion Dataset (MSLD) and the Monkeypox Skin Image Dataset (MSID), are used for evaluating the proposed methodology. The results on multiple deep learning models were evaluated using sensitivity, specificity and balanced accuracy. The proposed method has yielded highly promising results, demonstrating its potential for wide-scale deployment in detecting monkeypox. This smart and cost-effective solution can be effectively utilized in underprivileged areas where laboratory infrastructure may be lacking.

Mpox-AISM: AI-mediated super monitoring for mpox and like-mpox

“…The early diagnosis of mpox patients holds critical importance, as it not only markedly enhances the effectiveness of treatment and mitigates the long-term impact on the patient’s health but also significantly curtails the speed of disease transmission. 29 Building upon this foundation, we specifically evaluated the potential of mpox-AISM in diagnosing earlier-stage mpox rash. This endeavor aims to enhance the confidence of users affected by earlier-stage mpox in mpox-AISM.…”

Deep hybrid model for Mpox disease diagnosis from skin lesion images

“…These studies achieved promising results on the MPox Skin Lesion Dataset (MSLD) 24 and the MPox Skin Image Dataset (MSID). 25 Furthermore, novel approaches integrating AI-based optimization techniques have been proposed for MPox diagnosis. These approaches involve fine-tuning customized CNN layers using AI-Biruni Earth Radius (BER) 3 optimizationbased stochastic fractal search (BERSFS) 22 and employing optimization algorithms like sine cosine (SC) and particle swarm (PS) for feature selection with multilayer perceptron (MLP) classifiers.…”

Leveraging deep neural networks to uncover unprecedented levels of precision in the diagnosis of hair and scalp disorders

“…Here, deep learning is immersed to determine whether or not moles on the skin are malignant 37 . To determine which deep learning model was the most sensitive, accurate, and exact, many models were pitted against one another.…”