Hereditary neuromuscular diseases are multifactorial pathologies that are difficult to diagnose: 60% of patients do not have an accurate diagnosis. The European project CoMPaSS-NMD proposes an innovative approach for the diagnosis of these diseases based on Artificial Intelligence.
Applications of artificial intelligence in the healthcare sector
There are diseases for which making an accurate diagnosis is not easy. And not because of a lack of preparation or application by doctors: they are so complex and incredibly similar in symptoms, despite having different causes, that confirming them is complicated. This is the case with Hereditary Neuromuscular Diseases (HNMDs), which cause progressive muscle weakness and atrophy, leading to permanent disability in 30% of cases. In Europe, 500,000 people suffer from these diseases, of which 60% do not have a precise molecular diagnosis. This is because underlying HNMDs are mutations and combinations of mutations affecting over 600 genes.
Where the doctor fails, Artificial Intelligence can sometimes succeed, thanks to its ability to simultaneously and rapidly analyse an incredibly large amount of data, including clinical and health data. Not surprisingly, despite all the necessary precautions and the reluctance of doctors who struggle to trust technology, there is a growing interest in the applications of Artificial Intelligence in the medical and healthcare field. Especially in the field of diagnostic imaging (a well-trained algorithm can “read” reports like and better than the human eye); to identify risk factors for certain diseases based on the patient’s clinical history; to develop personalized therapies based on age, sex, and lifestyles. In a recent European project in which Deep Blue participated, Artificial Intelligence was also used to anticipate signs of depression in former oncology patients.
The CoMPaSS-NMD project: algorithms help doctors in diagnosis
It is in this context that CoMPaSS-NMD (Computational Models for new Patients Stratification Strategies of Neuromuscular Disorders) comes into play, a European project funded with nearly 6 million euros under the Horizon Program for research and innovation coordinated by the University of Modena and Reggio Emilia (the consortium members are 11 universities, research centres, technical and industrial partners). “The goal is to develop Artificial Intelligence algorithms to improve the diagnosis of hereditary neuromuscular diseases“, explains Annalisa De Angelis, dissemination and communication consultant at Deep Blue, a project partner. “As these pathologies are caused by different and complex genetic interactions, they are very difficult to understand even for a specialist”.
Today, the diagnostic path of HNMDs includes a series of investigations: blood tests to search for possible genetic mutations; histopathological and magnetic resonance imaging muscle analysis; interviews to collect the patient’s clinical history (symptoms and family history of the disease).
“It is a large amount of information that, moreover, reaches the doctor in an extremely fragmented way, which increases the difficulties of a diagnosis that is already very complex: patients with the same symptoms and similar changes in muscle anatomy may have different mutations or apparently harmless genetic variants. This determines different prognoses and treatments”, adds De Angelis.
“For this reason, it is necessary to find a way to effectively use patient data to quickly understand what is happening and intervene promptly with an appropriate therapeutic plan,” says Unimore geneticist Rossella Tupler, project coordinator. “CoMPaSS-NMD wants to develop a patient classification system that leverages Artificial Intelligence to support doctors, helping them process very complex data and quickly understand the diagnostic and therapeutic direction to be taken”.
With CoMPaSS-NMD, the approach to the diagnosis of hereditary neuromuscular diseases changes radically: the data collected from the patient are inserted into Artificial Intelligence algorithms trained to recognize recurring clinical patterns, or to classify the different forms of the disease quickly and accurately. “We expect to reduce waiting times and increase the number of correct diagnoses by 30%” says De Angelis. In perspective, this will also help in the development of personalised therapies, with evident benefits not only for patients but also for the healthcare system. Today, hereditary neuromuscular diseases ‘cost’ 30,000 euros per year per patient.
The CoMPaSS-NMD Atlas, a digital archive available to doctors and researchers
The algorithm training will be conducted using thousands of clinical, genetic, histological, and magnetic resonance imaging data provided by project partner clinical centres. The verification of the correct functioning of the algorithm, instead, will be carried out on 500 new patients awaiting diagnosis. All data collected will be made available in an Atlas, a real “digital map” to navigate the world of hereditary neuromuscular diseases. “We are collecting information on existing health atlases online so as to understand how best to develop ours, also in relation to the design of the interface that must facilitate consultation and navigation as much as possible (using filters, images, and graphs) for doctors and researchers, the end-users of the product”, explains De Angelis. The Atlas will be ready in 2025, in the meantime, the consortium’s work continues. Not only on the algorithm development front.
Sensitive data and patient privacy, how to regulate the use of AI in the medical field
The collection and storage of clinical and health data raise obvious privacy issues. For this reason, the development of Artificial Intelligence tools in the medical field requires rules to protect people’s rights. “Deep Blue will analyse the ethical and legal impacts of using Artificial Intelligence and the Atlas to resolve any problems already in the design phase and ensure that the new diagnostic approach is ethical and does not expose doctors and researchers who will use it to legal consequences,” emphasises De Angelis. “This analysis will also serve for the formulation of guidelines to follow to ensure safety, transparency, and responsibility in the use of patient data”.
In the project, Deep Blue is also leader in communication and dissemination actions of the results and will be involved in engaging stakeholders from the development phase to the use of the system by end-users. “We will work to facilitate the introduction of the new system into the daily routine of doctors in order to make it a ‘protagonist’ of the diagnostic and therapeutic process”, concludes De Angelis. “We will get in touch with doctors and researchers working on hereditary neuromuscular diseases and with patient associations to make them aware of the potential of the new tool, promoting its conscious and responsible adoption in hospitals, medical practices, and research centres”.
