Technology Bundle ID: NCI-E-118-2015

Eye Tracking Application in Computer Aided Diagnosis and Image Processing in Radiology

Request More Info
Licensing Contact:
Co-Inventors: 
Bradford Wood (NIH-CC), Haydar Celik (NIH-CC), Ulas Bagci (University of Central Florida)
Application: 
Software / Apps
Development Status: 

Pre-clinical (in vivo)

Medical imaging is an important resource for early diagnostic, detection, and effective treatment of cancers. However, the screening and review processes for radiologists have been shown to overlook a certain percentage of potentially cancerous image features. Such review errors may result in misdiagnosis and failure to identify tumors. These errors result from human fallibility, fatigue, and from the complexity of visual search required. Screening for early detection of cancers in visual images requires physicians to scan vast amounts of complex visual information thoroughly and interpret it correctly. Small abnormalities must be identified as potentially cancerous while avoiding false positive identifications.  Fatigue and bias during this review can cause physicians to favor certain locations for review and overlook other regions that may contain important abnormalities. 
The Centers for Medicare and Medicaid Services have recently approved screening computed tomography (CT) scans for patients ages 55-75 with a 15-year smoking history, which is expected to drastically increase the number of radiologic image reviews. Computer aided diagnosis (CAD) tools may become more important for radiologists to reduce diagnostic errors when screening medical images. 
Scientists at the National Institutes of Health – Clinical Center (NIH-CC) have developed a technology that is a CAD and eye-tracking system suitable for real-world radiology reading room settings. The system consists of an eye-tracking interface and novel algorithms to unify eye-tracking data and a CAD system. The system coordinates eye tracking and processes gaze patterns simultaneously with a deep learning algorithm in a multi-task learning platform to segment and assist the diagnosis of suspicious image features. Testing of this system in a lung cancer screening experiment with multiple radiologists shows improved accuracy in reducing false positives. The system is also generalizable to more complex applications such as prostate cancer screening with multi-parametric magnetic resonance imaging. This CAD system is able to improve radiologist diagnostic decisions during screening/diagnosis performance by determining where they have looked and tracking where they have a history of under-looking.

Applications:
  • Cancer screening
  • Computer aided diagnosis (CAD)
  • Software for coordinated human/computer vision tumor detection
Advantages:
  • Significantly improve detection of tumors

Patents

Application 62/466,516
Filed on 2017-03-03

Publications

Khosravan N, et al. A Collaborative Computer Aided Diagnosis (C-CAD) System with Eye-Tracking, Sparse Attentional Model, and Deep Learning. 
article

Updated

Aug 16, 2018

Data Source: 
nci