Decision support system for objective nasal airway obstruction assessment using computational fluid dynamics

Nasal airway obstruction (NAO) is a routinely encountered complaint by ear-nose-throat (ENT) physicians, affecting all age groups. NAOs yearly reduces the overall quality of life for millions of patients and, together with sinus diseases, cost the European healthcare systems billions of euros. The aetiology of NAO is determined by different conditions. If pharmacological treatment fails, as in many cases of anatomic abnormalities, surgery is often the treatment of choice. Nowadays, no proper gold standard currently exists for assessing nasal function impairment. Clinical examination is mainly used to make treatment decisions but frequently fails to pinpoint the cause of the patient’s perceived nasal obstruction. Current routine measurement techniques are found to correlate poorly with patients’ subjective feeling of nasal airway obstruction. However, these methods are used by rhinologists during decision making because no better objective measures are available. Therefore, the decision to proceed to surgery is generally based on the surgeon’s assessment, often without clear objective criteria. The reported surgical correction of nasal anatomic deformities with post-surgical symptom relief is inconsistent, and the range on reported failure rates inordinate. The introduction of a new standard for objective assessment of nasal airway obstruction could reduce the high failure rates by guiding surgical decision making. Computational fluid dynamics (CFD) models have the potential to fill this gap by providing consistent and accurate information on nasal airflow and function. CFD can also be coupled to different physical laws, such as particle deposition and heat loss. Particle deposition in the nose depends on the airflow and is essential to understand nasal cleansing of airborne particles and for targeted drug delivery. Such physics-based models would enable an accurate diagnosis for each patient individually, together with an evidence-based selection of the most effective therapy while enabling postoperative evaluation. Ideally, and in combination with other techniques, allow taking into account the evolution of the internal nasal anatomy and its effect on airflow. However, nowadays, some limitations inhibit such an approach becoming viable in the medical setting. Model-creation remains labour-intensive and time-consuming. The manual editing of X-ray tomographic cross-sections is not only tedious but also makes the model-creation prone to errors. On top of this, the numerical model design requires specific technical expertise that is not available for most physicians. In this dissertation, different machine learning and computer vision techniques were researched and combined with CFD to build proof-of-concept solutions to overcome existing limitations.

Jaar van publicatie:2020

Trefwoorden:Doctoral thesis

Toegankelijkheid:Closed