The use of artificial intelligence to improve mycetoma management

Hyam Omar Ali,Lamis Yahia Mohamed Elkheir,Ahmed Hassan Fahal

DOI: https://doi.org/10.1371/journal.pntd.0011914

2024-02-09

PLoS Neglected Tropical Diseases

Abstract:Mycetoma is a serious, neglected tropical disease affecting the underprivileged populations in resource-constrained communities of the rural tropical and subtropical regions. It predominately affects individuals of low visibility and voice and continues to hinder and burden the poor, remote communities [1]. Furthermore, in endemic areas, the medical and health facilities are meagre and adequate treatment is lacking or inaccessible [2]. All these lead to a progressive disease with massive morbidities and serious medical, health, and socioeconomic consequences [3,4]. Mycetoma is a chronic disabling subcutaneous granulomatous inflammatory disease of fungal and bacterial origin [5,6]. The inflammatory granuloma progressively spreads to affect the skin, deep tissues, and bones, leading to massive tissue damage, destruction, and serious morbidities and can be fatal [6–8]. A disease characteristic is the triad of subcutaneous mass, multiple sinuses discharging purulent and sero-purulent discharge frequently containing grains [9,10]. The mycetoma development and the associated risk factors are still unclear. Likewise, the disease susceptibility, resistance, entry route, and incubation period are unclear [11–13]. Presently, the diagnosis of mycetoma is difficult and tedious [14]. The available diagnostic tests are invasive, time-consuming, of low sensitivity and specificity, and there is no point of care [15–17]. Furthermore, the available treatment of mycetoma is suboptimal, characterised by a low cure rate and high recurrence and follow-up rates, and the disease remains with patients for a while, if not for life [18,19]. Artificial intelligence (AI) continues evolving swiftly, with ongoing research and development in various domains. It can potentially bring about significant advancements in science, technology, and society. AI has made substantial inroads into the medical and healthcare arenas, revolutionising healthcare delivery, disease diagnosis, and management. Nowadays, there are many applications for AI in medical and health practice, to mention but a few: its use in medical imaging, disease diagnosis, drug discovery, personalised medicine, data predictive analytics, electronic health records, telemedicine and virtual health assistance, robotics surgery, drug dosage, treatment recommendations, health monitoring, mental health, drug adverse event detection and healthcare operations, and resource management [20,21]. Machine learning (ML) is a subset of AI that creates algorithms and models that enable computers to learn and make predictions or decisions based on data to improve performance without explicit programming. ML is essential for interpreting large datasets, automating tasks, and enhancing decision-making processes, applied across diverse domains like image and speech recognition, natural language processing, recommendation systems, autonomous vehicles, healthcare, and finance. The learning process involves adjusting model parameters based on input data through training on labelled examples. ML encompasses 3 main types: supervised learning, where models are trained on labelled datasets; unsupervised learning, which explores patterns in unllabeled data; and reinforcement learning, where agents learn by making decisions in an environment to maximise cumulative rewards [22,23]. Deep learning (DL) is a specialised field within ML that utilises artificial neural networks, particularly deep neural networks with multiple layers, to undertake intricate tasks. Mimic the human brain's structure, these networks learn hierarchical representations of data, enabling them to capture complex patterns and features. Its key characteristic includes the reliance on neural networks, where connections between nodes are adjusted during the learning process. DL automatically extracts relevant features from raw data and reduces the need for manual feature engineering. Its applications are in AI span image and speech recognition, natural language processing, autonomous vehicles, healthcare, and game playing. DL can autonomously learn from vast datasets, making it well-suited for challenging tasks where traditional methods may struggle, though it often requires substantial computational resources for training [24,25]. Advances in hardware and software have facilitated its widespread adoption in various AI applications. These applications are valuable due to their potential to enhance the accuracy and efficiency of medical diagnostics, reduce human error, and improve patient treatment outcomes. As the progress in mycetoma management, research and development has been sluggish and inadequate over the years, we thought the utility of AI might enhance and accelerate this progress. With this background, we believe the utility of AI can improve mycetoma patient management and research and development endeavours in several ways. Here are some aspects of this utility:

tropical medicine,parasitology