Foundations of Image Analysis

• Image registration:

  • process of aligning two images to each other
  • use: image guided radiotherapy, adaptive treatment planning
• Image fusion: outcome

  • combined display of two images after they have been registered to each other
• Types of image registration:

  • rigid registration
  • affine registration
  • deformable registration
• Rigid registration:

  • best fit registration where images are not modified in order to match each other
  • shapes are rigidly maintained
  • only translation and rotation of image as a whole
• Affine registration:

  • modifications change the distance between points in an image
  • but do not change the relationship between the lines
  • translation, rotation, scaling, sheering, plane reflection
• Deformable registration:

  • more complex registration where one or more objects are reshaped in order to more accurately align to each other
  • can correct for organ motion and other positional variations by deforming one image to match another
• Image guided radiation therapy (IGRT):

  • use of imaging before radiation treatment to improve the precision of radiation delivery
  • in - room imaging guiding radiation delivery
  • not used to diagnose or plan radiation treatment
• Pros of IGRT:

  • improves accuracy of treatment by allowing RTs to verify target location
  • allows for identification of and correction for set up discrepancies or errors
  • allows for the assessment of anatomical or tumour changes so that timely re - planning can occur
  • can result in decreased margin size due to improved "confidence of set up"
• Image matching:

  • process of aligning the treatment image to the DRR to localize treatment location
  • once images are aligned, shift information is used to move the treatment couch - put patient in correct location relative to the machine isocenter
  • need to clearly see key structures to perform match well
• Image filters:

  • used to emphasize certain objects or structures on planar images which are poorly visible on standard view setting
  • can only apply one image filter at a time
• Primary match structure:

  • structures which best represent the treatment target
  • in close proximity to the target
  • tumour or target itself
  • if not seen: then next best structure is chosen
• Secondary match structure:

  • structures which help with the image match but they alone don't accurately represent the target
  • useful for match confirmation
  • do not use to directly match (don't match off this structure)
• Manual match:

  • RT can perform the matching of the treatment image to the DRR
• Auto match:

  • computer performs a match
  • attempts best match for all anatomy in the image
• Region of interest (ROI):

  • defines the area to be searched during an auto match
  • limiting ROI: focuses auto match algorithm and speeds up auto match
  • VOI = volume of interest for CBCT auto matching
• 3 P's of image guidance best practice:

  • patient: diagnosis + staging
  • plan: technique, treatment parameters, treatment field, treatment targets)
  • protocol: department protocols
• Image guidance best practice:

  • review all patient and treatment related documentation
  • review prior images where applicable
  • determine imaging parameters
  • acquire quality images
  • assess anatomy
  • assess treatment positioning
  • perform image match (reference image) or image assessment (BEV)
  • initiate corrective action
  • adjust set up
  • perform shifts
• Patient rotations:

  • pitch: rotation about the lateral axis
  • roll: rotation about longitudinal axis
  • yaw: rotation about the anterior / posterior axis
• Translational shifts:

  • superior / inferior
  • left / right
  • anterior / posterior
• Online set up correction:

  • set up correction done in real time with patient on the treatment couch
  • reduces both systemic and random errors through real time image matching and application of bed shift corrections within a treatment fraction
• Shift action level:

  • couch shift threshold for an image match at which bed shifts should be applied
  • commonly = one half the PTV margin
• Image repeat action level:

  • a couch shift threshold for an image match at which bed shifts should be applied and subsequently verified with a repeat image
  • typically: equal to the PTV margin
• Offline set up correction:

  • set up correction done with patient off the treatment couch (in between treatment fractions)
  • reduces systematic error - evaluates multiple fractions of set up errors and applying a correction to a subsequent fraction
  • random set up error is not accounted for
• No action level protocol (NAL):

  • offline set up correction strategy
  • patient set up errors are averaged over a predetermined number of fractions and always corrected for
  • mean shift is calculated and applied to the patient by remarking the isocenter
• Systematic set up error:

  • set up deviations that occur in the same direction and magnitude through a treatment course
  • will be unchanged over the course of treatment for a patient
  • mean value of set up error
  • result: shift of dose from the treatment plan
• Random set up error:

  • set up deviations that can vary in direction and magnitude during treatment
  • standard deviation of set up errors
  • result: blurring of the dose in relation to the treatment plan
• Residual set up error:

  • remaining error after performing a correction
  • ex. anatomical movements
• If you need a BEV, then MV image is good - can see the field and MLC
• If we just need to position the isocenter using bony anatomy as landmarks - then KV or KV / KV pair is sufficient
• If we need to localize and organ or tumour for isocenter placement (ex. prostate) - need an image with more / better soft tissue information = CBCT