Patient submitted visual data — new ways of working


On the face of it, capturing an image or video and submitting it to a clinician to be reviewed is a simple enough task. With Isla, we’ve been digging into this for the last 6 months and are happy to share learnings on the complexities.

Capture: For many patients, submitting this information will be a new process and may be a one off meaning that the system and onboarding must be intuitive and low friction.

Storage: Medical photography can be highly sensitive so we must be able to control the storage of images end to end, manage permissions to view these images and protect privacy.

Review: With almost all data, value is unlocked through the process of turning data into information and information into insight. The same is true of visual data so we need to create an intuitive way to structure what is, in essence, unstructured data.

The concept of ‘friction’ is frequently talked about in technology; keeping clicks low, reading short and typing shorter goes a long way and rules out conventional apps where download and registration convene the last two of these. Rise, the ‘progressive web app’, essentially a website which also allows for a single click ‘install’ for engaged users wanting an application experience. At Isla, we love this approach and it has meant that we can support one off patients to capture and submit an image with only 5 clicks whilst engaged users retain the option to install the platform, login and view their entire visual record.

Next up, storage. With the growing need for patients, nurses, GPs and others to be able to seamlessly share visual data, the privacy concern of using personal mobiles for medical photographs becomes very real. At Isla, we have devised a process to capture images and videos by directly accessing the camera on the device and avoiding the phone’s local storage. All data is captured and stored temporarily in the browser, then encrypted and stored on NHS approved cloud servers. In addition, users can mark images as sensitive to add a blur to images unless being actively viewed and a complete audit history of who can, and is, accessing images is provided. The commitment that is made if using this approach is that testing must be very robust to ensure that the software performs as expected for every combination of browser and device. No small task.

The third challenge, review, is perhaps the most interesting. As clinicians have begun to review and provide feedback on images we learn about the challenges they face. So far, image quality is top of the list. We have found this is best addressed through technology combined with behavioural nudging. With issues such as blur, computer vision gives a neat solution, allowing machines to assess images in a fraction of a second and request that low quality submissions are retaken. However, identifying issues with skew or rotation, i.e that a photo has been taken from a different angle to a previous submission is a much harder problem and best met with a simple solution; show the patient the previous image whilst they capture the new one and the results are dramatically improved.

Tackling all of these challenges comes at a cost but we believe the value delivered far outweighs this. As we develop our platform we see time and again that visual data will play a key role in future care delivery as the health sector remodels into continuous patient monitoring. We will continue to break down the barriers to unlock the potential of visual data and support healthcare innovation.

See how Isla works.