AI applications in lower limb peripheral artery disease

Lower limb peripheral arterial disease (PAD)

Peripheral arterial disease corresponds to an obstruction of one or more arteries vascularizing the lower limbs. The main arteries vascularizing the lower limbs are the iliac arteries, which divide into the femoral arteries, which in turn give rise to the popliteal arteries at the knee, which then give rise to the tibial and fibular arteries.

When the obstruction of the artery is partial, we speak of stenosis, corresponding to a narrowing of the diameter of the artery. If the obstruction is total, it is a thrombosis, corresponding to an occlusion of the artery. The arterial obstruction leads to a reduction of blood supply and oxygen to the lower limbs. The lack of vascularization causses ischemia of the tissues.

How common is this?

PAD is a common disease affecting more than 230 million people worldwide. It is a global public health problem.

How serious is this?

Although the disease can remain asymptomatic for a long time, it is not benign as it can develop into life-threatening complications or functional impairment of the lower limb. At the beginning of the disease, symptoms are not very marked and can be expressed by pain in the limb which appears during an effort, such as walking, and which disappears a few minutes after the effort has stopped. As the disease progresses, the pain can even be felt at rest; it increases when the patient is lying down and is relieved by being in a horizontal position due to the increased blood supply caused by gravity. In the most advanced stage, trophic disorders may appear such as ulcers (skin wounds), necrosis, or gangrene. This stage threatens the viability of the limb and is associated with a risk of amputation.

What is it due to?

PAD is caued by pathological arterial remodeling of the arterial wall by atheromatous plaque deposits. These plaques are partly related to the accumulation of cholesterol and cause an inflammatory remodeling in the wall leading to a diffuse arterial disease called atherosclerosis. Several cardiovascular risk factors for atherosclerosis have been identified, including smoking, high blood pressure, diabetes, dyslipidemia (increased cholesterol levels), family heredity and male gender.

How is the diagnosis made?

The diagnosis of PAD is made and confirmed thanks to imaging. Doppler ultrasound of the lower limb is a simple, non-invasive examination that can show narrowing of the arteries, and enable to measure the degree of stenosis or identify thrombosis. In combination with blood pressure measurement, it allows the calculation of the systolic pressure index (SPI). When the SPI is outside the norm, indicates damages to the arterial wall. Depending on the clinical context, other examinations such as computed tomography angiography (CTA), magnetic resonance angiography (MRA) or arteriography of the lower limb may be necessary to accurately assess the location and extent of arterial lesions.

How is it treated?

The treatment of PAD is tailored to each patient depending on the severity and stage of the disease. It relies on a multidisciplinary approach that aims to control cardiovascular risk factors, prevent and manage complications and also treat arterial obstruction by revascularizing the lower limb. The medical treatment is based on lifestyle changes (diet, physical activity, stopping smoking) to control the factors contributing to the development and aggravation of the disease. Medication can be used to maintain correct blood pressure, control cholesterol levels or treat any associated diabetes. Anti-platelet agents are also prescribed to prevent the development of clots (thrombus) in the atherosclerotic plaques that could worsen the arterial blockage.

Depending on the severity of the disease, surgical treatment can be indicated to revascularize the lower limb. Different techniques can be performed including:

Balloon dilatation of the artery (angioplasty) which may be associated with the placement of a stent.
Endarterectomy (or thromboendarterectomy) in which the surgeon removes the atheromatous plaque.
Arterial bypass surgery, which involves bypassing the blocked artery and restoring the blood supply to the downstream area by creating a new circuit.

Clinical guidelines for the management of patients with PAD are established by international groups of experts to ensure the optimal care with evidence-based medicine constantly updated to the latest knowledge.

In the field of vascular diseases, guidelines for clinical practice are established by international groups of experts: – the European Society of Vascular Surgery (ESVS) in Europe, and – the Society of Vascular Surgery (SVS) in the USA. In France, the French Society of Vascular and Endovascular Surgery (SCVE) also provides practical information to better understand vascular diseases and their treatments).

Our research

The aim of research is to better understand factors underlying the development of PAD and the outcomes after surgical treatment in order to improve patients’ care. At the Institute 3IA Côte d’Azur, our team is working on the applications of Artificial Intelligence (AI) to improve the management of patients with vascular diseases, including PAD. We are developing an innovative software to facilitate and improve imaging analysis of patient with vascular diseases.

We have been working for several years on AI applications in vascular diseases and we have witnessed the strong dynamism and the great advances made in this field. We have published an article analyzing academic publications which confirms the major expansion of this field of research, in particular over the last 5 years. Although research into the application of AI to improve the management of AI is still in its infancy, early results offer encouraging perspectives.

Contributions of Natural Language Processing (NLP)

Natural Language Processing (NLP) is a field of AI that focuses on applications for the analysis of written or spoken language. NLP offers many perspectives of applications in healthcare, especially for improving and optimizing information systems (see dedicated chapter).

Although the prevalence of PAD is very high worldwide, affecting more than 230 million people, this disease is often under-diagnosed and treated at severe and advanced stages. Several recently published studies have shown that NLP could offer new techniques to perform large-scale analysis of health data and allow professionals to better identify patients with PAD from medical records.

Contributions in the field of vision

Computer vision is a branch of AI that deals with the analysis and processing of images or videos. Medical imaging plays a central role in the management of patients with PAD, as it allows the diagnosis of the disease, the assessment of its severity, the preparation of therapeutic intervention and the follow-up of patients. One of the major challenges of AI is to automatize and improve the performances of medical imaging analysis. Various pilot studies have shown that AI could be used to enable automatic detection of arterial lesions in patients with PAD. This kind of tool could complement the expertise of health professionals and help them to better detect the disease, especially in asymptomatic patients or those with early lesions that are more difficult to diagnose. The development of automatic software could also facilitate the classification of lesions and the assessment of the severity disease, by enabling automatic and standardized measurements of the vessels. This kind of application could improve the accuracy and reproducibility of measurements, while saving time for professionals. In addition to a clinical application, this technology also offers new perspectives to better understand the mechanisms behind PAD by allowing the automatic and rapid analysis of precise anatomical data of the lower limb vasculature. Our team has developed fully automated software for the analysis of scans of patients with vascular diseases

Thanks to this new tool, we were able, for example, to study in greater detail the impact of vascular calcifications in patients who underwent revascularization for PAD. The potential applications are very diverse and could lead to a better understanding of mechanisms underlying the disease in order to improve the management of patients.

Contributions of Machine Learning

Machine Learning (ML) is a branch of AI which allows a machine or computer to learn from data and it can be used to develop predictive models. In clinical practice, the therapeutic decision is based on the assessment of the benefit/risk balance, taking into consideration the disease and its severity, as well as the patient’s general condition and the existence of other possible associated diseases. Predictive models are of great interest in clinical practice as they could help to better assess the prognosis of patients by grading the risk of disease-related complications against potential treatment-related complications. Several studies have recently shown the value of ML in assessing the risk of worsening of PAD as well as the risk of post-operative complications. This kind of application offers the perspective of developing a personalized medicine adapted to each patient.

For detailed information on the state of the art in this field, you can consult the following article we published: https://pubmed.ncbi.nlm.nih.gov/35921995/.

Remaining challenges

Although AI offers very promising perspectives for improving care, including the management of PAD, research is still in its infancy and essential steps remain before implementing these applications in daily clinical practice. First, the AI-applications need to be evaluated and validated by healthcare professionals to assess their performance, efficiency, and safety of use. AI-derived applications are developed from health data and need to be trained on large volumes of data (big data) in order to be efficient and applicable to all patients. From one center to another and from one country to another, medical data are very heterogeneous as they can be generated by different providers and stored in different formats in several registries and repositories. There is thus a real need of standardization and a need to federate actors of research to build international multidisciplinary collaborations in which AI experts and health professionals work together to develop fair, ethic, responsible and trustworthy AI. Ethics remain at the heart of the development of AI-applications which are subject to very strict regulations, in order to guarantee data protection, respect of privacy and patients’ wishes. Finally, innovation and development of AI-derived applications require adapted infrastructures as well as competent experts. It can represent a significant economic investment, with potential long-term benefits for health at individual and collective levels.