Principal component analysis as a method of respiratory motion detection on a solid state CZT dedicated cardiac camera
2016
1969 Objectives Patient and respiratory motion can introduce artefacts into myocardial perfusion imaging (MPI)1,2. Previous work by our group divided the MPI acquisition into a series of 30s images and developed a method using image registration to detect and correct for patient motion on the CZT solid state gamma camera3. Images ≤ 1s are required to detect respiratory motion and due to the level of noise on these images, this technique was unsuccessful for respiratory motion detection. By extension, the current study investigates incorporation of principal component analysis (PCA), to facilitate a data driven respiratory motion detection technique for the CZT camera.
Methods A data driven technique exploiting PCA was developed. The overall motion of the heart was determined from the PCA components (method 1) and from a combination of PCA with image registration (method 2). Overall motion was divided into respiratory and patient components. Respiratory motion validation was performed using a dynamic phantom and by comparing the respiratory signal to that measured using an external device on 7 rest MPI patients. Patient motion was compared to measurements using our previous technique3. The data was separated into discrete bins according to features of the data driven signal and the signal from the external device. The binned images were registered to measure the amplitude of motion and to produce a motion corrected image.
Results On phantom studies, the data driven technique recovered the respiratory signal and motion artefacts were removed following motion correction. The magnitude of respiratory motion on patient studies was 8mm ± 4mm. For respiratory motion 蠅 8mm (2 patients) there was a strong correlation between the data driven technique and external device; method 1, r = 0.65, 0.73; method 2, r = 0.61, 0.60. For motion 蠅 7mm and < 8mm (1 patient) there was a moderate correlation for method 2, r = 0.42. On these studies respiratory motion was clearly apparent when viewing the rebinned data from both signals. For the remaining 4 patients, respiratory motion was < 7mm and there was a weak correlation with the external device; method 1, r ≤ 0.15; method 2, r ≤ 0.37. The magnitude of patient motion was 5mm ± 2mm. There was a strong correlation between patient motion derived using our previously validated technique3 and this extended method for motion 蠅 5mm (r 蠅 0.7) and a weak correlation when motion was < 5mm (r ≤ 0.3).
Conclusions A technique has been developed that is capable of detecting the respiratory signal on MPI images on a CZT solid state gamma camera. The curve that is generated can be analysed to infer the translation of the heart. Correlation with an external device has been demonstrated when the translation of the heart due to respiration is 蠅 7mm. The technique requires further assessment if its implications for diagnosis are to be characterised, but its principal advantage is that no external, physical motion detection device is required. Characterisation of such motion could lead to improved interpretation of associated artefacts.
[1] J Nucl Med Technol. 2008;36(3):155-61. [2] J Nucl Med. 2002;43(9):1259-67. [3] J Nucl Cardiol. 2015; doi:10.1007/s12350-015-0314-1.
This abstract presents independent research funded by the National Institute for Health Research (NIHR).The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
0
References
2
Citations
NaN
KQI