Limits...
Optimisation of quantitative lung SPECT applied to mild COPD: a software phantom simulation study.

Norberg P, Olsson A, Alm Carlsson G, Sandborg M, Gustafsson A - EJNMMI Res (2015)

Bottom Line: Sixty-four reconstruction updates and a small kernel size should be used when the whole lung is analysed, and for the reduced lung a greater number of updates and a larger kernel size are needed.A LEHR collimator and 125 (99m)Tc MBq together with an optimal combination of cutoff frequency, number of updates and kernel size, gave the best result.Suboptimal selections of either cutoff frequency, number of updates and kernel size will reduce the imaging system's ability to detect mild COPD in the lung phantom.

View Article: PubMed Central - PubMed

Affiliation: Medical Radiation Physics, Department of Medical and Health Sciences, Linköping University, Linköping, 581 83 Sweden ; Center for Medical Image Science and Visualization (CMIV), Linköping University, Linköping, 581 83 Sweden.

ABSTRACT

Background: The amount of inhomogeneities in a (99m)Tc Technegas single-photon emission computed tomography (SPECT) lung image, caused by reduced ventilation in lung regions affected by chronic obstructive pulmonary disease (COPD), is correlated to disease advancement. A quantitative analysis method, the CVT method, measuring these inhomogeneities was proposed in earlier work. To detect mild COPD, which is a difficult task, optimised parameter values are needed.

Methods: In this work, the CVT method was optimised with respect to the parameter values of acquisition, reconstruction and analysis. The ordered subset expectation maximisation (OSEM) algorithm was used for reconstructing the lung SPECT images. As a first step towards clinical application of the CVT method in detecting mild COPD, this study was based on simulated SPECT images of an advanced anthropomorphic lung software phantom including respiratory and cardiac motion, where the mild COPD lung had an overall ventilation reduction of 5%.

Results: The best separation between healthy and mild COPD lung images as determined using the CVT measure of ventilation inhomogeneity and 125 MBq (99m)Tc was obtained using a low-energy high-resolution collimator (LEHR) and a power 6 Butterworth post-filter with a cutoff frequency of 0.6 to 0.7 cm(-1). Sixty-four reconstruction updates and a small kernel size should be used when the whole lung is analysed, and for the reduced lung a greater number of updates and a larger kernel size are needed.

Conclusions: A LEHR collimator and 125 (99m)Tc MBq together with an optimal combination of cutoff frequency, number of updates and kernel size, gave the best result. Suboptimal selections of either cutoff frequency, number of updates and kernel size will reduce the imaging system's ability to detect mild COPD in the lung phantom.

No MeSH data available.


Related in: MedlinePlus

pvalues for all the LEHR-125 MBq designs. Designs for the whole lung are shown in the left column and for the reduced lung in the right column. In row (a) the designs are placed in order of rank based on resulting p values. In row (b) the designs are grouped by kernel size, row (c) by iteration number and in row (d) by cutoff frequency of the Butterworth filter. The lowest p values are encircled.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4385278&req=5

Fig1: pvalues for all the LEHR-125 MBq designs. Designs for the whole lung are shown in the left column and for the reduced lung in the right column. In row (a) the designs are placed in order of rank based on resulting p values. In row (b) the designs are grouped by kernel size, row (c) by iteration number and in row (d) by cutoff frequency of the Butterworth filter. The lowest p values are encircled.

Mentions: When the whole lung is analysed, the lowest p values result from using LEHR-125 MBq designs. LEHR-125 MBq designs placed in order of rank based on resulting p values and thereafter grouped by kernel size show the lowest p value for the smallest kernel size, i.e. with an edge length of 1.0 cm (see Figure 1b). When the designs are grouped by number of iterations, a minimum is found at four iterations, i.e. 64 updates (see Figure 1c). When grouped by cutoff frequency of the Butterworth filter, a minimum at 0.6 to 0.7 cm−1 is found (see Figure 1d). The lowest p value for these designs is p = 4.1 × 10−13. Designs generating the 10 lowest p values are listed in Table 4. The lowest p value for a LEGP-125 MBq design is p = 8.0 × 10−4, a LEGP-25 MBq design p = 0.016 and for a LEHR-25 MBq design p = 0.020. Also shown in Figure 1 is that a LEHR-125 MBq design can result in a higher p value than the best LEGP-125 MBq design, e.g. if a kernel edge length of 3.0 cm, 288 updates and a cutoff frequency of 0.4 cm−1 are chosen, p = 0.0026 will result.Figure 1


Optimisation of quantitative lung SPECT applied to mild COPD: a software phantom simulation study.

Norberg P, Olsson A, Alm Carlsson G, Sandborg M, Gustafsson A - EJNMMI Res (2015)

pvalues for all the LEHR-125 MBq designs. Designs for the whole lung are shown in the left column and for the reduced lung in the right column. In row (a) the designs are placed in order of rank based on resulting p values. In row (b) the designs are grouped by kernel size, row (c) by iteration number and in row (d) by cutoff frequency of the Butterworth filter. The lowest p values are encircled.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4385278&req=5

Fig1: pvalues for all the LEHR-125 MBq designs. Designs for the whole lung are shown in the left column and for the reduced lung in the right column. In row (a) the designs are placed in order of rank based on resulting p values. In row (b) the designs are grouped by kernel size, row (c) by iteration number and in row (d) by cutoff frequency of the Butterworth filter. The lowest p values are encircled.
Mentions: When the whole lung is analysed, the lowest p values result from using LEHR-125 MBq designs. LEHR-125 MBq designs placed in order of rank based on resulting p values and thereafter grouped by kernel size show the lowest p value for the smallest kernel size, i.e. with an edge length of 1.0 cm (see Figure 1b). When the designs are grouped by number of iterations, a minimum is found at four iterations, i.e. 64 updates (see Figure 1c). When grouped by cutoff frequency of the Butterworth filter, a minimum at 0.6 to 0.7 cm−1 is found (see Figure 1d). The lowest p value for these designs is p = 4.1 × 10−13. Designs generating the 10 lowest p values are listed in Table 4. The lowest p value for a LEGP-125 MBq design is p = 8.0 × 10−4, a LEGP-25 MBq design p = 0.016 and for a LEHR-25 MBq design p = 0.020. Also shown in Figure 1 is that a LEHR-125 MBq design can result in a higher p value than the best LEGP-125 MBq design, e.g. if a kernel edge length of 3.0 cm, 288 updates and a cutoff frequency of 0.4 cm−1 are chosen, p = 0.0026 will result.Figure 1

Bottom Line: Sixty-four reconstruction updates and a small kernel size should be used when the whole lung is analysed, and for the reduced lung a greater number of updates and a larger kernel size are needed.A LEHR collimator and 125 (99m)Tc MBq together with an optimal combination of cutoff frequency, number of updates and kernel size, gave the best result.Suboptimal selections of either cutoff frequency, number of updates and kernel size will reduce the imaging system's ability to detect mild COPD in the lung phantom.

View Article: PubMed Central - PubMed

Affiliation: Medical Radiation Physics, Department of Medical and Health Sciences, Linköping University, Linköping, 581 83 Sweden ; Center for Medical Image Science and Visualization (CMIV), Linköping University, Linköping, 581 83 Sweden.

ABSTRACT

Background: The amount of inhomogeneities in a (99m)Tc Technegas single-photon emission computed tomography (SPECT) lung image, caused by reduced ventilation in lung regions affected by chronic obstructive pulmonary disease (COPD), is correlated to disease advancement. A quantitative analysis method, the CVT method, measuring these inhomogeneities was proposed in earlier work. To detect mild COPD, which is a difficult task, optimised parameter values are needed.

Methods: In this work, the CVT method was optimised with respect to the parameter values of acquisition, reconstruction and analysis. The ordered subset expectation maximisation (OSEM) algorithm was used for reconstructing the lung SPECT images. As a first step towards clinical application of the CVT method in detecting mild COPD, this study was based on simulated SPECT images of an advanced anthropomorphic lung software phantom including respiratory and cardiac motion, where the mild COPD lung had an overall ventilation reduction of 5%.

Results: The best separation between healthy and mild COPD lung images as determined using the CVT measure of ventilation inhomogeneity and 125 MBq (99m)Tc was obtained using a low-energy high-resolution collimator (LEHR) and a power 6 Butterworth post-filter with a cutoff frequency of 0.6 to 0.7 cm(-1). Sixty-four reconstruction updates and a small kernel size should be used when the whole lung is analysed, and for the reduced lung a greater number of updates and a larger kernel size are needed.

Conclusions: A LEHR collimator and 125 (99m)Tc MBq together with an optimal combination of cutoff frequency, number of updates and kernel size, gave the best result. Suboptimal selections of either cutoff frequency, number of updates and kernel size will reduce the imaging system's ability to detect mild COPD in the lung phantom.

No MeSH data available.


Related in: MedlinePlus