Clarity and Precision: Using Artificial Intelligence to measure and improve image quality

The Isla platform is an innovative visual-record and patient monitoring system which allows clinicians to request and view patient-submitted images, video and forms. By the end of 2021, Isla had received over 90,000 submissions to its patient records. A large percentage of these are images submitted remotely by patients or their carers. Clinical teams are then able to assess their medical needs and define a care pathway. However, poor image quality, through blurred or incorrectly taken images, can lead to submissions being clinically unusable. This is expected to occur in 1-2% of submitted images in UK research studies, according to the Cardiac SSI network. In 2020, Isla’s co-founder James Jurkiewicz and clinical colleagues at Royal Brompton & Harefield hospitals (RB&HH) authored a study on the potential use of AI in improving image quality. The paper, titled “Using artificial intelligence to improve wound image quality: a feasibility study” focused on the post-surgical wound care pathway at RB&HH. The authors noted that the presence of blur in a wound image can prevent a clinician from identifying a suture knot or assessing the placement of surgical clips. These features, and a number of others, are crucial in identifying whether a post-surgical wound is likely to become infected.

In the feasibility study, the authors investigated the use of AI-based techniques for blur detection on images taken at discharge. They noted the need for high quality training data in any additional AI-models built on images, where a blur detection and blur reduction algorithm would be the first step in data cleaning and processing. Blur detection in an image can also be used to prompt a user to retake a blurry photo, leading to more efficient use of clinical time during review

The AI team at Isla has recently built on this work, implementing a novel and improved algorithm for blur detection and image enhancement. The model, which has been developed and will soon be available within Isla, offers blur-detection capabilities with improved accuracy. Additionally, users will have the capability to enhance the image quality, using image sharpening techniques which are tuned to detect foreground and background. The tool will be available not only for post-surgical wound pathways, but for any image submitted to Isla. Techniques of this nature offer increased accuracy and clinical efficiency, as well as consistency in care and patient management across the UK. 

The feasibility study is available for healthcare professionals at https://www.wounds-uk.com/journals/issue/628/article-details/using-artificial-intelligence-improve-wound-image-quality-feasibility-study