Assessment of three dead detector correction methods for cone-beam computed tomography

Area of Science:

• Medical Physics
• Image Processing
• Radiotherapy Technology

Background:

• Dead detectors in electronic portal imaging devices (EPIDs) for cone-beam CT (CBCT) cause image artifacts.
• These artifacts, such as ring artifacts, degrade image quality and impact treatment accuracy.
• Megavoltage CBCT (MVCBCT) is particularly susceptible to these issues due to higher radiation doses.

Purpose of the Study:

• To evaluate three dead detector correction methods for MVCBCT.
• To compare the effectiveness of bidirectional linear interpolation (BLI), quad-directional linear interpolation (QLI), and Laplacian solution (LS) methods.
• To determine the failure points of these correction methods under various conditions.

Main Methods:

• MVCBCT projections were acquired from four linacs at 8 and 60 MU.
• Simulated dead detectors (randomly distributed, clusters, disks) were introduced with varying percentages and sizes.
• Correction algorithms (BLI, QLI, LS) were applied, and failure was defined by image artifacts or Modulation Transfer Function (MTF) drop.

Main Results:

• All methods failed at lower doses (8 MU) and with smaller defects compared to higher doses (60 MU).
• The LS method generally outperformed BLI and QLI.
• Failure modes included ring artifacts and MTF drop, with artifacts being the primary failure point.

Conclusions:

• The LS method is superior for correcting dead detectors in MVCBCT.
• Correction effectiveness diminishes with increasing imaging dose.
• Acceptable artifact levels allow for significant detector failure (up to 25% random, 4mm NRDD) before spatial resolution is compromised.