Automated Organ Volume Measurement Tool for Improved Clinical Decision-Making

Principal Investigator: 

Martin Prince, Professor of Radiology

Background & Unmet Need

  • Organ volume is an important biomarker for diagnosing and tracking response to therapy in several diseases including Autosomal Dominant Polycystic Kidney Disease (ADPKD)
  • Accurate measurement of kidney volume is essential for ADPKD patients since this disease progresses slowly and patients may need to wait for more than a year to assess response to therapy
  • Currently, the main method of measuring total kidney volume (TKV) for ADPKD requires manual tracing of the kidney boundaries on MRI or CT scans for each imaging slice
  • This process is not only tedious, but also subject to variability between observers, leading to random errors up to 11%
  • Unmet Need: Methods for measuring organ volume which are faster, more accurate, and more reproducible

Technology Overview

  • The Technology:
  • An automated, deep learning-based method for measuring organ volumes using averages to reduce variability
  • Organ volume is calculated using a 2D U-net model trained on pulsed sequences of MRI images
  • Patients are scanned using T2-weighted fast spin echo (axial, coronal), steady state free precession (axial, coronal), and T1-weighted 3D gradient echo (axial)
  • These automated measurements are averaged for an improved and reproducible measurement of TKV
  • PoC Data: Automated measurements were calculated significantly faster than manual measurements (2:49 mins vs 11:34 mins), but did not statistically differ from the manual calculation
  • Averaging measurements from the five models significantly reduced variability to an average of 2.5% (compared to an average of 5.9% for five manual observers)

Technology Applications

  • Automatic measurement of kidney volume to track disease progression in ADPKD
  • Predict risk of developing end-stage kidney disease and determine eligibility for tolvaptan therapy
  • Track disease progression and treatment response in other indications involving organ enlargement, such as splenomegaly and hepatomegaly

Technology Advantages

  • More reproducible and faster measurement of organ volume
  • Enables organ volume measurement more frequently and from multiple orientations
  • Lower risk of intra- and inter-observer error

Contours of kidney volume drawn manually are more reproducible using model assisted labelling.



Intellectual Property

Patents

  • Provisional Filed

Cornell Reference

  • 10591

Contact Information

Donna Rounds, Ph.D

For additional information please contact

Donna Rounds
Associate Director, Business Development and Licensing
Phone: (646) 962-7044
Email: djr296@cornell.edu