What is fungalAi?
fungalAi is a platform technology that can detect and help diagnose fungal infections from data that is readily available in hospitals. Natural language processing screens CT reports for fungal pneumonia taking into account the variation in radiologist language; image analysis detects suspicious features on Chest CT scans and the expert system integrates radiology with other clinical data to improve fungal prediction.
How does it work?
fungalAi uses artificial neural networks to analyse free-text CT reports and recognise fungal infections on CT images. The expert system uses hand crafted rules and machine learning classifiers to integrate different types of data to improve fungal prediction and reduce false positive alerts. More advanced approaches will be possible as more electronic data becomes available.
Why use this approach?
Chest imaging is the cornerstone of diagnosis because the majority of patients with the difficult to manage mold infections like Aspergillosis, develop a culture negative fungal pneumonia. Surveillance is not occurring because fungal infections are regarded as rare and isolation of fungi in the laboratory is uncommon making monitoring really challenging. Given the importance of chest imaging to diagnosis, it makes sense to target the chest scan when searching for fungal infections. However chest imaging is reported by radiologists...So, how do you sift through a deluge of reports to locate a rare disease? Our sensitive analytics can help.
fungalAi natural language processing using our recent neural text classifier has a sensitivity of 95% and specificity of 93%. Our deep learning based image analysis of chest CT scans has an AUC of 99.3. Our expert system combining radiology, antifungal drug prescriptions and microbiology maintained a very high sensitivity of 98% and increased specificity to 75% thereby halving false notifications.
fungalAi has helped us understand our local epidemiology, patient outcomes and gaps in antifungal stewardship practice.
What about you?