Impact of missing attenuation and scatter corrections on 99m Tc-MAA SPECT 3D dosimetry for liver radioembolization using the patient relative calibration methodology: A retrospective investigation on clinical images

Francesca Botta, Mahila Ferrari, Carlo Chiesa, Sara Vitali, Francesco Guerriero, Maria Chiara De Nile, Marta Mira, Leda Lorenzon, Massimiliano Pacilio, Marta Cremonesi

Research output: Contribution to journalArticle

Abstract

PURPOSE: To investigate the clinical implication of performing pre-treatment dosimetry for 90 Y-microspheres liver radioembolization on 99m Tc-MAA SPECT images reconstructed without attenuation or scatter correction and quantified with the patient relative calibration methodology.

METHODS: Twenty-five patients treated with SIR-Spheres® at Istituto Europeo di Oncologia and 31 patients treated with TheraSphere® at Istituto Nazionale Tumori were considered. For each acquired 99m Tc-MAA SPECT, four reconstructions were performed: with attenuation and scatter correction (AC_SC), only attenuation (AC_NoSC), only scatter (NoAC_SC) and without corrections (NoAC_NoSC). Absorbed dose maps were calculated from the activity maps, quantified applying the patient relative calibration to the SPECT images. Whole Liver (WL) and Tumor (T) regions were drawn on CT images. Injected Liver (IL) region was defined including the voxels receiving absorbed dose >3.8 Gy/GBq. Whole Healthy Liver (WHL) and Healthy Injected Liver (HIL) regions were obtained as WHL = WL - T and HIL = IL - T. Average absorbed dose to WHL and HIL were calculated, and the injection activity was derived following each Institute's procedure. The values obtained from AC_NoSC, NoAC_SC and NoAC_NoSC images were compared to the reference value suggested by AC_SC images using Bland-Altman analysis and Wilcoxon paired test (5% significance threshold). Absorbed-dose maps were compared to the reference map (AC_SC) in global terms using the Voxel Normalized Mean Square Error (%VNMSE), and at voxel level by calculating for each voxel the normalized difference with the reference value. The uncertainty affecting absorbed dose at voxel level was accounted for in the comparison; to this purpose, the voxel counts fluctuation due to Poisson and reconstruction noise was estimated from SPECT images of a water phantom acquired and reconstructed as patient images.

RESULTS: NoAC_SC images lead to activity prescriptions not significantly different from the reference AC_SC images; the individual differences (<0.1 GBq for all IEO patients, <0.6 GBq for all but one INT patients) were comparable to the uncertainty affecting activity measurement. AC_NoSC and NoAC_NoSC images, instead, yielded significantly different activity prescriptions and wider 95% confidence intervals in the Bland-Altman analysis. Concerning the absorbed dose map, AC_NoSC images had the smallest %VNMSE value and the highest fraction of voxels differing less than 2 standard deviations from AC_SC.

CONCLUSIONS: The patient relative calibration methodology can compensate for the missing attenuation correction when performing healthy liver pre-treatment dosimetry: safe treatments can be planned even on NoAC_SC images, suggesting activities comparable to AC_SC images. Scatter correction is recommended due to its heavy impact on healthy liver dosimetry.

Original languageEnglish
Pages (from-to)1684-1698
Number of pages15
JournalMedical Physics
Volume45
Issue number4
DOIs
Publication statusPublished - Apr 2018

Fingerprint

Technetium Tc 99m Aggregated Albumin
Single-Photon Emission-Computed Tomography
Calibration
Liver
Uncertainty
Prescriptions
Reference Values
Microspheres

Keywords

  • Adult
  • Calibration
  • Embolization, Therapeutic
  • Female
  • Humans
  • Image Processing, Computer-Assisted
  • Liver/diagnostic imaging
  • Liver Neoplasms/diagnostic imaging
  • Male
  • Middle Aged
  • Monte Carlo Method
  • Phantoms, Imaging
  • Radiometry
  • Retrospective Studies
  • Scattering, Radiation
  • Signal-To-Noise Ratio
  • Technetium Tc 99m Aggregated Albumin
  • Tomography, Emission-Computed, Single-Photon
  • Uncertainty

Cite this

@article{a4c1b8831ac840d2af9e5795a2e9c4af,
title = "Impact of missing attenuation and scatter corrections on 99m Tc-MAA SPECT 3D dosimetry for liver radioembolization using the patient relative calibration methodology: A retrospective investigation on clinical images",
abstract = "PURPOSE: To investigate the clinical implication of performing pre-treatment dosimetry for 90 Y-microspheres liver radioembolization on 99m Tc-MAA SPECT images reconstructed without attenuation or scatter correction and quantified with the patient relative calibration methodology.METHODS: Twenty-five patients treated with SIR-Spheres{\circledR} at Istituto Europeo di Oncologia and 31 patients treated with TheraSphere{\circledR} at Istituto Nazionale Tumori were considered. For each acquired 99m Tc-MAA SPECT, four reconstructions were performed: with attenuation and scatter correction (AC_SC), only attenuation (AC_NoSC), only scatter (NoAC_SC) and without corrections (NoAC_NoSC). Absorbed dose maps were calculated from the activity maps, quantified applying the patient relative calibration to the SPECT images. Whole Liver (WL) and Tumor (T) regions were drawn on CT images. Injected Liver (IL) region was defined including the voxels receiving absorbed dose >3.8 Gy/GBq. Whole Healthy Liver (WHL) and Healthy Injected Liver (HIL) regions were obtained as WHL = WL - T and HIL = IL - T. Average absorbed dose to WHL and HIL were calculated, and the injection activity was derived following each Institute's procedure. The values obtained from AC_NoSC, NoAC_SC and NoAC_NoSC images were compared to the reference value suggested by AC_SC images using Bland-Altman analysis and Wilcoxon paired test (5{\%} significance threshold). Absorbed-dose maps were compared to the reference map (AC_SC) in global terms using the Voxel Normalized Mean Square Error ({\%}VNMSE), and at voxel level by calculating for each voxel the normalized difference with the reference value. The uncertainty affecting absorbed dose at voxel level was accounted for in the comparison; to this purpose, the voxel counts fluctuation due to Poisson and reconstruction noise was estimated from SPECT images of a water phantom acquired and reconstructed as patient images.RESULTS: NoAC_SC images lead to activity prescriptions not significantly different from the reference AC_SC images; the individual differences (<0.1 GBq for all IEO patients, <0.6 GBq for all but one INT patients) were comparable to the uncertainty affecting activity measurement. AC_NoSC and NoAC_NoSC images, instead, yielded significantly different activity prescriptions and wider 95{\%} confidence intervals in the Bland-Altman analysis. Concerning the absorbed dose map, AC_NoSC images had the smallest {\%}VNMSE value and the highest fraction of voxels differing less than 2 standard deviations from AC_SC.CONCLUSIONS: The patient relative calibration methodology can compensate for the missing attenuation correction when performing healthy liver pre-treatment dosimetry: safe treatments can be planned even on NoAC_SC images, suggesting activities comparable to AC_SC images. Scatter correction is recommended due to its heavy impact on healthy liver dosimetry.",
keywords = "Adult, Calibration, Embolization, Therapeutic, Female, Humans, Image Processing, Computer-Assisted, Liver/diagnostic imaging, Liver Neoplasms/diagnostic imaging, Male, Middle Aged, Monte Carlo Method, Phantoms, Imaging, Radiometry, Retrospective Studies, Scattering, Radiation, Signal-To-Noise Ratio, Technetium Tc 99m Aggregated Albumin, Tomography, Emission-Computed, Single-Photon, Uncertainty",
author = "Francesca Botta and Mahila Ferrari and Carlo Chiesa and Sara Vitali and Francesco Guerriero and Nile, {Maria Chiara De} and Marta Mira and Leda Lorenzon and Massimiliano Pacilio and Marta Cremonesi",
note = "{\circledC} 2018 American Association of Physicists in Medicine.",
year = "2018",
month = "4",
doi = "10.1002/mp.12774",
language = "English",
volume = "45",
pages = "1684--1698",
journal = "Medical Physics",
issn = "0094-2405",
publisher = "John Wiley and Sons Ltd",
number = "4",

}

TY - JOUR

T1 - Impact of missing attenuation and scatter corrections on 99m Tc-MAA SPECT 3D dosimetry for liver radioembolization using the patient relative calibration methodology

T2 - A retrospective investigation on clinical images

AU - Botta, Francesca

AU - Ferrari, Mahila

AU - Chiesa, Carlo

AU - Vitali, Sara

AU - Guerriero, Francesco

AU - Nile, Maria Chiara De

AU - Mira, Marta

AU - Lorenzon, Leda

AU - Pacilio, Massimiliano

AU - Cremonesi, Marta

N1 - © 2018 American Association of Physicists in Medicine.

PY - 2018/4

Y1 - 2018/4

N2 - PURPOSE: To investigate the clinical implication of performing pre-treatment dosimetry for 90 Y-microspheres liver radioembolization on 99m Tc-MAA SPECT images reconstructed without attenuation or scatter correction and quantified with the patient relative calibration methodology.METHODS: Twenty-five patients treated with SIR-Spheres® at Istituto Europeo di Oncologia and 31 patients treated with TheraSphere® at Istituto Nazionale Tumori were considered. For each acquired 99m Tc-MAA SPECT, four reconstructions were performed: with attenuation and scatter correction (AC_SC), only attenuation (AC_NoSC), only scatter (NoAC_SC) and without corrections (NoAC_NoSC). Absorbed dose maps were calculated from the activity maps, quantified applying the patient relative calibration to the SPECT images. Whole Liver (WL) and Tumor (T) regions were drawn on CT images. Injected Liver (IL) region was defined including the voxels receiving absorbed dose >3.8 Gy/GBq. Whole Healthy Liver (WHL) and Healthy Injected Liver (HIL) regions were obtained as WHL = WL - T and HIL = IL - T. Average absorbed dose to WHL and HIL were calculated, and the injection activity was derived following each Institute's procedure. The values obtained from AC_NoSC, NoAC_SC and NoAC_NoSC images were compared to the reference value suggested by AC_SC images using Bland-Altman analysis and Wilcoxon paired test (5% significance threshold). Absorbed-dose maps were compared to the reference map (AC_SC) in global terms using the Voxel Normalized Mean Square Error (%VNMSE), and at voxel level by calculating for each voxel the normalized difference with the reference value. The uncertainty affecting absorbed dose at voxel level was accounted for in the comparison; to this purpose, the voxel counts fluctuation due to Poisson and reconstruction noise was estimated from SPECT images of a water phantom acquired and reconstructed as patient images.RESULTS: NoAC_SC images lead to activity prescriptions not significantly different from the reference AC_SC images; the individual differences (<0.1 GBq for all IEO patients, <0.6 GBq for all but one INT patients) were comparable to the uncertainty affecting activity measurement. AC_NoSC and NoAC_NoSC images, instead, yielded significantly different activity prescriptions and wider 95% confidence intervals in the Bland-Altman analysis. Concerning the absorbed dose map, AC_NoSC images had the smallest %VNMSE value and the highest fraction of voxels differing less than 2 standard deviations from AC_SC.CONCLUSIONS: The patient relative calibration methodology can compensate for the missing attenuation correction when performing healthy liver pre-treatment dosimetry: safe treatments can be planned even on NoAC_SC images, suggesting activities comparable to AC_SC images. Scatter correction is recommended due to its heavy impact on healthy liver dosimetry.

AB - PURPOSE: To investigate the clinical implication of performing pre-treatment dosimetry for 90 Y-microspheres liver radioembolization on 99m Tc-MAA SPECT images reconstructed without attenuation or scatter correction and quantified with the patient relative calibration methodology.METHODS: Twenty-five patients treated with SIR-Spheres® at Istituto Europeo di Oncologia and 31 patients treated with TheraSphere® at Istituto Nazionale Tumori were considered. For each acquired 99m Tc-MAA SPECT, four reconstructions were performed: with attenuation and scatter correction (AC_SC), only attenuation (AC_NoSC), only scatter (NoAC_SC) and without corrections (NoAC_NoSC). Absorbed dose maps were calculated from the activity maps, quantified applying the patient relative calibration to the SPECT images. Whole Liver (WL) and Tumor (T) regions were drawn on CT images. Injected Liver (IL) region was defined including the voxels receiving absorbed dose >3.8 Gy/GBq. Whole Healthy Liver (WHL) and Healthy Injected Liver (HIL) regions were obtained as WHL = WL - T and HIL = IL - T. Average absorbed dose to WHL and HIL were calculated, and the injection activity was derived following each Institute's procedure. The values obtained from AC_NoSC, NoAC_SC and NoAC_NoSC images were compared to the reference value suggested by AC_SC images using Bland-Altman analysis and Wilcoxon paired test (5% significance threshold). Absorbed-dose maps were compared to the reference map (AC_SC) in global terms using the Voxel Normalized Mean Square Error (%VNMSE), and at voxel level by calculating for each voxel the normalized difference with the reference value. The uncertainty affecting absorbed dose at voxel level was accounted for in the comparison; to this purpose, the voxel counts fluctuation due to Poisson and reconstruction noise was estimated from SPECT images of a water phantom acquired and reconstructed as patient images.RESULTS: NoAC_SC images lead to activity prescriptions not significantly different from the reference AC_SC images; the individual differences (<0.1 GBq for all IEO patients, <0.6 GBq for all but one INT patients) were comparable to the uncertainty affecting activity measurement. AC_NoSC and NoAC_NoSC images, instead, yielded significantly different activity prescriptions and wider 95% confidence intervals in the Bland-Altman analysis. Concerning the absorbed dose map, AC_NoSC images had the smallest %VNMSE value and the highest fraction of voxels differing less than 2 standard deviations from AC_SC.CONCLUSIONS: The patient relative calibration methodology can compensate for the missing attenuation correction when performing healthy liver pre-treatment dosimetry: safe treatments can be planned even on NoAC_SC images, suggesting activities comparable to AC_SC images. Scatter correction is recommended due to its heavy impact on healthy liver dosimetry.

KW - Adult

KW - Calibration

KW - Embolization, Therapeutic

KW - Female

KW - Humans

KW - Image Processing, Computer-Assisted

KW - Liver/diagnostic imaging

KW - Liver Neoplasms/diagnostic imaging

KW - Male

KW - Middle Aged

KW - Monte Carlo Method

KW - Phantoms, Imaging

KW - Radiometry

KW - Retrospective Studies

KW - Scattering, Radiation

KW - Signal-To-Noise Ratio

KW - Technetium Tc 99m Aggregated Albumin

KW - Tomography, Emission-Computed, Single-Photon

KW - Uncertainty

U2 - 10.1002/mp.12774

DO - 10.1002/mp.12774

M3 - Article

C2 - 29383733

VL - 45

SP - 1684

EP - 1698

JO - Medical Physics

JF - Medical Physics

SN - 0094-2405

IS - 4

ER -