Sonopill is coming to a close in less than a year, in that time we have tested various sensor modalities during in-vivo trials. This work has not only been challenging, rewarding, and sometimes frustrating, but it points toward a trend occurring throughout medicine, due to the increasing importance of technology in this field. As sensors get smaller and less expensive, we can expect to see them more widely used in medicine, opening up new diagnostic possibilities, bringing medicine into the data-age and transforming it into a more analytical and evidence-based discipline. This leads to several exciting opportunities for both engineers and clinicians.

The rich data-sets produced by multi-modal devices will highlight unexpected pattern and suggest new questions about the role of different factors in the causes of various diseases and enable quantitative analysis of the effectiveness of therapies. However, despite the advantages of multi-modal capsule endoscopy, there are physical and economic limits as to how many sensors are integrated within the pill-sized package. Engineers and physicians will need to work together on preliminary studies to determine what are the optimal number and type of sensors required to realise this vision while also providing solutions to packaging, power, processing and integration challenges.

These preliminary studies using multi-modal capsule endoscopy will produce a stream of data from each patient. Once this point is reached, it is common sense that a more efficient solution for analysing this data is required. From conversations we have had with gastroenterologists over the course of the Sonopill project, we have noticed that a recurring bugbear of clinicians is the amount of time taken to review the footage produced by video capsule endoscopes. This is time that could be better spent elsewhere. It can be safely assumed that this issue will be even more significant for a capsule endoscope containing both white light imaging as well as some other modality, such as micro-ultrasound imaging. Computer Aided Diagnostics (CADx) would be required to deal with the more detailed and extensive data sets produced over the lifetime of the patient to flag up any changes compared to baseline readings and potentially help generate more accurate diagnoses.

The market for these analytical systems for making sense of these data sets is estimated to be on the order of $300 to $450 billion a year according to McKinsey & Company. The race to develop these systems and bring them to market is further exemplified by the recent authorisation by the FDA by the first Artificial Intelligent ophthalmologic diagnostic system by a company called IDX. This system is capable of providing a screening decision without the need of interpretation of the image or results from a clinician. However, whether this will be an exception or the rule will require an ongoing conversation between ethicists, engineers, patients and clinicians.

Analysis of data-sets from different patients will have many benefits once the anonymity of the patient is preserved. Multiple data-sets from different patients will enable continuous improvement of machine learning algorithms for diagnosis. This “black box” detection would flag unexpected biomarkers for gastrointestinal disease. The identification of new measurands will, in turn, enhance the design of medical devices, spur research into new medical sensor technologies and suggest where capsule endoscopy should expand beyond the physiome into the different environments such as the microbiome, metabolome and more.

Dr Gerard Cummins is a technical coordinator on the Sonopill project and a postdoctoral researcher at Heriot-Watt University.

First published: 17 April 2018

<< Blog