Artificial intelligence as an aid to medical diagnosis

What is artificial intelligence?

Simply put, artificial intelligence refers to software solutions capable of performing tasks that usually require human intervention. These tasks can be more or less complex, ranging from detecting objects or elements to formulating recommendations or producing content and conversations.

The use of artificial intelligence is particularly relevant in order to analyze and process large amounts of data. However, AI remains to date a tool at the service of mankind. In the healthcare sector, AI is not destined to establish diagnoses in full autonomy and can’t be held responsible for its results. The aim is rather to guide decision-making by cross-referencing diverse sources of information, quickly and self-learningly, reinforcing analyses with factual statistics, or guiding healthcare professionals on specific or subtle points, thus saving them time.

How is AI revolutionizing medical diagnosis?

AI-assisted medical imaging analysis for faster, more reliable diagnosis

Medical imaging is a field that produces large quantities of data, often heterogeneous, which are analyzed by radiologists and other professionals involved in establishing a diagnosis. For example, in 2020, over 10 million medical images (CT scans and MRIs) were carried out in public and private non-profit establishments in France[1].

In response, numerous medical imaging analysis tools have been developed. They cover a wide spectrum of specialties (e.g. oncology, angiology, cardiology, traumatology, etc.) and all image acquisition modalities (X-ray, CT scan, MRI, ultrasound). The algorithms are trained on datasets including both healthy and sick cases, comprising medical images, clinical and omic data characterizing these images. Once learning is complete, these tools can :

• Accurately detect the probability of abnormalities,
• Draw the attention of specialists to specific area of the image,
• Assist disease monitoring, for example by calculating the evolution of nodule size.

Thanks to these artificial intelligence-enabled solutions, specialists are allowed to concentrate on higher value-added cases, and save precious time. Meanwhile, non-specialists can better guide patients through the diagnostic process.

AI-assisted genetic analysis for simplified genetic interpretation

More and more frequently, healthcare professionals are also basing their medical diagnoses on genetic data. In fact, genetic analyses are a complementary source of information, and are particularly encouraged by the “Plan France Médecine Génomique 2025”, which was aiming at the processing of 235,000 genome sequences per year by 2020.

Genetic analyses are often very time-consuming due to the sheer volume of data to be processed. The algorithms developed are therefore capable of sifting through genetic datasets and providing a shortlist of variants responsible for genetic syndromes, reducing the choices and therefore the duration of interpretation.

The latest algorithms also include explanations to justify the result, including the prioritization of variants. This provides practitioners with a logical pathway, and ensures the reliability of the analysis performed by artificial intelligence algorithm.

Discover examples of healthcare AI projects carried out with our customers >

AI medical diagnosis : what are the remaining challenges?

Recent years have seen an explosion in the number of artificial intelligence solutions applied to medicine. In France, 102 healthcare AI start-ups were counted in 2019, compared with 191 in 2020! Among these, 59 are focused on diagnosis assistance.

However, even though their numbers are growing, healthcare companies implementing artificial intelligence as a medical diagnosis aid face several challenges:

• Enabling data interoperability, to ensure fluidity and ease of integration into practitioners’ workflows,
• Ensuring confidentiality and secure data storage,
• Propose viable business models, or consider reimbursement by Health Insurances,
• Demonstrate the reliability of solutions through clinical studies, and obtain CE or FDA marking,
• Quantify the benefits of artificial intelligence through operational impact studies (e.g., reducing the time and/or cost of treatment);
• And, above all, convince healthcare professionals to adopt these solutions, with the question of legal liability for diagnostics still central to the debate.

Nevertheless, evidence of the faith of the healthcare ecosystem in artificial intelligence based-solutions can already be seen. Several hospitals, both public and private, are now using decision-making tools based on artificial intelligence algorithms. These tools are integrated into the daily practice of healthcare professionals, without replacing their expertise, and letting them have the final decision. At the same time, calls for proposals are published to support innovation, and international initiatives are created, such as the Focus Group on Artificial Intelligence for Health (FG-AI4H), driven by the World Health Organization (WHO).

In conclusion, artificial intelligence solutions are already helping doctors in their practice, by providing support for medical diagnosis, making them more reliable and faster. These solutions are continually improving, notably to detect pathologies more efficiently, enabling treatment at the earliest stage, and thus improving the quality of care. However, their inventors still face a number of challenges.

While a future without doctors is neither conceivable nor envisaged at present, the question arises as to who will bear the legal responsibility for patient care and the choice of therapeutic strategies employed tomorrow if AI is at the source of all decisions.

At Alcimed, we are convinced of the future of artificial intelligence in diagnostic assistance, and our team is ready to support you in your healthcare AI projects. Don’t hesitate to contact us!

[1] Imaging equipment for public and private not-for-profit healthcare establishments – DREES (2020)

About the author, 

Line, Consultant in Alcimed’s Innovation and Public Policy team in France