Aristotle in his Metaphysics has proclaimed that value increases when combining diverse elements together to create synergies. Somewhat misquoted, nonetheless true and profound,"The Whole is Greater than the Sum of the Parts" is the perfect phrase to describe what happens when pathologists and AI work together for the benefit of patients.
Expert pathologists are one of the most important links in the chain of delivering high quality healthcare. Their contribution to timely, accurate and specific diagnosis is critical to the multi-disciplinary effort that is modern medicine. It is, however, well documented that pathology services are under pressure, with requests rising faster than available resources. Digital pathology is increasingly adopted to help bridge the resource gap. Excellent tools, like Inspirata's Dynamyx, help to deliver efficiencies in reviewing cases, linking with patient data, recalling archives, reporting, collaborating and controlling quality within and across labs.
Computer analysis of digitised pathology images promises the next leap forward towards higher efficiency and accuracy. Although traditional image processing methods proved helpful, state-of-the art artificial intelligent technology is better suited to the complexity of the problem at hand.
DeepMed IO Ltd., an Inspirata AI partner, has been working to deliver a series of pathology solutions with an initial focus in oncology, starting from the DeepPath lymph node metastatic detection system. Using cutting-edge deep learning methods, DeepMed has trained neural networks to identify and annotate metastatic regions, including micrometastases and isolated tumour cells, within a few minutes after the image is generated by whole slide image scanners. The annotation layer is stored as image metadata and can be called upon by histopathologists within Dynamyx. Rather than having to examine every nook and cranny of the high magnification image, the pathologist can "call" the suspect regions to appear in sequence at the click of a button and deliver their expert diagnosis within mere seconds.
Having the experts as decision makers allows DeepMed to bias the algorithms towards sensitivity: histopathologists can spot false positives at the "blink of an eye", therefore discarding a few suggestions per slide is worth the extra time for minimising false negatives. Initial validation in real world conditions shows that the combination of expert and software delivers a 4% increase in diagnostic accuracy in half the time. Aristotle was, once again, right!
This is a guest post by Dr Emmanuel Raptakis, CMO of DeepMed IO Ltd, a UK-based company and Inspirata approved partner that offers AI solutions for several feature detection and tumor grading applications.
Disclaimer: DeepPath is currently for research use only and not suitable for primary diagnosis.
Sign up to receive updates about new blog posts.