2015 Winter Project Week:Bolus Arrival Time Estimation in PK Modelling

From NAMIC Wiki
Revision as of 17:47, 10 July 2017 by Grundlett (talk | contribs) (Text replacement - "http://www.slicer.org/slicerWiki/index.php/" to "https://www.slicer.org/wiki/")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Home < 2015 Winter Project Week:Bolus Arrival Time Estimation in PK Modelling

Key Investigators

  • Alireza Mehrtash, BWH
  • Andrey Fedorov, BWH
  • Jim Miller, GE Global Research
  • Sandeep Gupta, GE Global Research

Project Description

PharmacoKinetics (PK) analysis of DCE-MRI images requires estimation of bolus arrival time (BAT). Inaccuracies in estimation of BAT can affect the results of derived PK metrics. The goal of this project is to improve the current implementation of BAT estimation in Slicer's PK Modelling extension [1]. The current implemented method in Slicer measures the time point before the peak gradient arrival and set it as BAT.

Objective

  • Study current implementation of BAT detection in PK Modelling software in available clinical data for prostate and breast cancers.
  • Investigate possible solutions to improve the robustness of BAT detection in PK Modelling.

Approach, Plan

  • Add option for manual assignment of BAT separately for AIF (Arterial Input Function) and TRF (Tissue Residual Function).
  • Compare different bat detection methods: LL ([1]), PL ([2]) and slicer's methods (PG) in detection of AIF BAT and TRF BAT in whole prostate gland [3].

Progress

  • PG method was used to detect BAT in QIN Prostate and OHSU Breast DCE-MRI data sets.
  • On prostate data sets we extracted the AIF from the left femoral artery and then we used LL, PL and PG methods to detect BAT. Due to the sharp rise in AIF curve all three methods successfully detect BAT in AIF signal (%100 agreement on 9 cases).
  • By studying BAT maps for the whole prostate gland we observed that for tissue residual function (TRF) concentration curves in areas with slower uptake curves, there is higher chance that PG method trapping in local minima. However some of the curves in which PG fails are voxels which are not good candidates for Tofts modelling (This could include necrotic ROIs). For future improvement of PG method Jim suggested to inpaint the false BATs by using the neighborhood information.
  • Since PG works correctly on AIF detection, we are going to use manual assignment for TRF only and use automatic bat detection for AIF.

References