Use of the Smartphone Camera to Monitor Adherence to Inhaled Therapy

Adherence to inhaled controller medications is crucial for patients with chronic respiratory illnesses to achieve favorable clinical outcomes. Selfmanagement measures have been shown to enhance health outcomes, decrease unnecessary interventions, and improve disease control. However, compliance evaluations have difficulties in establishing a high level of trustworthiness, as patient’s self-reported high compliance rates are frequently regarded unreliable. A mobile application module to objectively verify inhalation usage using image snapshots of the inhalation counter and optic character recognition has shown to be promising, but insufficient for some inhaler models. In this paper a model specific approach was explored to enable reliable adherence measurement. To achieve this, a machine learning model was trained on an inhaler image dataset. The trained model had an average accuracy of 88% in recognizing the digits on the dose counter of an inhaler model. These results show the potential to gain additional evidence for inhaler compliance.