InDetector – Automatic detection of infected driveline regions

Shravan Aras, Thienne Johnson, Christopher Gniady, Rinku Skaria, Zain I Khalpey

Research output: Contribution to journalArticlepeer-review


Although there have been significant advancements in Left Ventricular Assist Device(LVAD) technology and improvements in mortality rates, infection remains one of the major complications associated with LVAD therapy with an incidence of 25-80% cases annually. Amongst such infections driveline infections are the most common and account for 14-28% according to the RE-MATCH trial of the total of LVAD related infections. If a patient is diagnosed with LVAD infection, it is important to initiate antibiotic therapy as early as possible. If left untreated it can lead to sepsis with multi-organ failure, longer hospital stay, delay heart transplant or early mortality. To improve infection detection and monitoring we propose InDetector, a driveline infection detection system that allows at-home patients to use a smartphone to capture images of their driveline regions to check for infections. InDetector uses a Convolutional Neural Network along with image augmentation techniques for inferring infected images and achieves an overall classification accuracy of 93.75% on our validation dataset.

Original languageEnglish (US)
Pages (from-to)170-178
Number of pages9
JournalSmart Health
StatePublished - Dec 2018

ASJC Scopus subject areas

  • Medicine (miscellaneous)
  • Information Systems
  • Health Informatics
  • Computer Science Applications
  • Health Information Management


Dive into the research topics of 'InDetector – Automatic detection of infected driveline regions'. Together they form a unique fingerprint.

Cite this