Mathieu Rubeaux | Cedars-Sinai Medical Center (original) (raw)
Uploads
Papers by Mathieu Rubeaux
Journal of Nuclear Medicine, 2016
Precise definition of the mitral valve plane (VP) during segmentation of the left ventricle (LV) ... more Precise definition of the mitral valve plane (VP) during segmentation of the left ventricle (LV) for single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI) quantification often requires manual adjustment, which affects quantification of perfusion. We developed a machine learning approach using support vector machines (SVM) for automatic VP placement. 392 consecutive patients undergoing fast SPECT MPI (99m)Tc-tetrofosmin stress (5 min, 300-350 MBq)/rest (5 min, 750-1050 MBq) attenuation corrected (AC) by computed tomography (CT) and same‑day coronary CT angiography (CCTA), including 48 patients with invasive angiography (ICA) and no known coronary artery disease, were studied. LV was segmented by standard clinical software (Quantitative Perfusion SPECT) by two experts, adjusting VP if needed. Two-class SVM models were computed from expert positions using 10‑fold cross validation to separate the patients used for training and for validation. SVM probability estimates were used to compute the best VP position. Automatic VP localization in AC and non-AC images were compared to expert placement on CCTA. Stress and rest total perfusion deficits (TPD) and detection of per‑vessel obstructive stenosis by ICA were also compared. VP agreement (bias, 95% confidence interval) between SVM and experts was lower than inter-expert agreement for stress-AC (1, -5-7 mm) vs. (0, -8-8 mm) and rest-AC (1, -7-10 mm) vs. (0, -10-10 mm) (p<0.01), and similar for stress non-AC (1, -4‒6 mm) vs. (0, -6-5 mm) and rest non-AC (2, -7-10 mm) vs. (-1, -9-7 mm) images (P = non-significant (NS)). For regional detection of obstructive stenosis, ischemic‑TPD areas under receiver operating characteristic curve (AUCs) (95% confidence interval) of the two experts 0.79 (0.7-0.87), 0.81 (0.73-0.89) and SVM 0.82 (0.74-0.9) for AC data were the same (P = NS) and higher than for unadjusted VP 0.63 (0.53-0.73) (p<0.01). Similarly, for non-AC data, AUCs of experts 0.77 (0.69-0.89), 0.8 (0.72-0.88) and SVM 0.79 (0.71-0.87) were the same (P = NS) and higher than for unadjusted VP 0.65 (0.56-0.75) (p<0.01). Machine learning with SVM allows automatic and accurate VP localization, decreasing user-dependence in SPECT MPI quantification.
Journal of nuclear medicine : official publication, Society of Nuclear Medicine, 2017
Atherothrombotic events in coronary arteries are most often due to rupture of unstable plaque res... more Atherothrombotic events in coronary arteries are most often due to rupture of unstable plaque resulting in myocardial infarction. Radiolabeled molecular imaging tracers directed toward cellular targets that are unique to unstable plaque can serve as a powerful tool for identifying high-risk patients and for assessing the potential of new therapeutic approaches. Two commonly available radiopharmaceuticals-(18)F-FDG and (18)F-NaF-have been used in clinical research for imaging coronary artery plaque, and ongoing clinical studies are testing whether there is an association between (18)F-NaF uptake and future atherothrombotic events. Other, less available, tracers that target macrophages, endothelial cells, and apoptotic cells have also been tested in small groups of patients. Adoption of molecular imaging of coronary plaque into clinical practice will depend on overcoming major hurdles, ultimately including evidence that the detection of unstable plaque can change patient management an...
Journal of nuclear medicine : official publication, Society of Nuclear Medicine, Jan 15, 2015
Ruptured coronary atherosclerotic plaques commonly cause acute myocardial infarction. It has been... more Ruptured coronary atherosclerotic plaques commonly cause acute myocardial infarction. It has been recently shown that active microcalcification in the coronary arteries, one of the features that characterizes vulnerable plaques at risk of rupture, can be imaged using (18)F-sodium fluoride ((18)F-NaF) PET. We aimed to determine whether a motion correction technique applied to gated (18)F-NaF PET images could enhance image quality and improve uptake estimates. Seventeen patients with myocardial infarction (n = 7) and stable angina (n = 10) underwent (18)F-NaF PET and prospective coronary CT angiography (CCTA). PET data were reconstructed in 4 different ways: (i) one gated bin (end-diastolic phase with 25% of the counts), (ii) 4 gated bins (consecutive 25% segments), (iii) 10 gated bins (consecutive 10% segments), and (iv) ungated. Subsequently, gated PET images were registered using a local and non-linear motion correction method guided by the extracted coronary arteries from CT angio...
Journal of nuclear medicine : official publication, Society of Nuclear Medicine, Jan 24, 2015
A novel PET radiotracer, flurpiridaz F-18, has undergone Phase II clinical trial evaluation as a ... more A novel PET radiotracer, flurpiridaz F-18, has undergone Phase II clinical trial evaluation as a high-resolution PET cardiac perfusion imaging agent. In a subgroup of patients imaged with this agent, we assessed the feasibility and benefit of simultaneous correction of respiratory and cardiac motion. In 16 patients, PET imaging was performed on a 4-ring Siemens Biograph-64 in dual cardiac and respiratory gating mode. Four sets of data were reconstructed with high-definition reconstruction (HD•PET): (i) ungated and (ii) 8-bin ECG-gated images using 5-min acquisition, (iii) optimal respiratory gating (ORG)-as developed for oncological imaging- using a narrow range of breathing amplitude around end-expiration level with 35% of the counts in a 7-min acquisition, and (iv) 4-bin respiratory-gated and 8-bin ECG-gated images (32 bins in total) using the 7-min acquisition (dual-gating, using all data). Motion-frozen (MF) registration algorithms were applied to ECG-gated and dual-gated data, ...
Journal of Nuclear Medicine
Objectives Hybrid PET/Computed Tomography Angiography (CTA) using 18F-sodium fluoride (18F-NaF) t... more Objectives Hybrid PET/Computed Tomography Angiography (CTA) using 18F-sodium fluoride (18F-NaF) tracer has been shown to identify ruptured and high-risk coronary atherosclerotic plaques1. However, distinction between 18F-NaF positive and negative lesions is challenging due to cardiac motion, low signal-to-noise ratio and small size of the lesions with PET uptake. We examined the feasibility of cardiac motion-frozen (MF) coronary 18F-NaF PET by nonlinear registration to aid this novel application. Methods 11 patients with acute myocardial infarction and stable angina underwent 18F-NaF PET-CT and CTA. Gated PET images, reconstructed with time-of-flight information and resolution recovery, were fused with CTA images. A total of 23 18F-NaF positive and 16 18F-NaF negative (reference) lesions were manually delineated by experienced observers guided by CTA. MF images were created from the 4 gates of PET data by a nonlinear level-set registration algorithm applied on a region of interest s...
In Image-Guided Radiation Therapy of prostate cancer, the CBCT scan acquired at each treatment fr... more In Image-Guided Radiation Therapy of prostate cancer, the CBCT scan acquired at each treatment fraction could be used to estimate a cumula-tive dose distribution thanks to non-rigid registration. However, this cumulative dose estimation is highly sensitive to non-rigid registration errors. For this rea-son, validation of the registration algorithm with organ overlap measures or visual assessment is not sufficient. In this paper, we describe the construction of a nu-merical phantom based on a finite element model of the prostate and the neighbor organs which can be used to assess the non-rigid registration accuracy. Prelimi-nary results show the potential of this phantom to better characterize registration algorithms than traditional Dice score.
Mutual Information (MI) has been extensively studied as similarity measure for the registration o... more Mutual Information (MI) has been extensively studied as similarity measure for the registration of medical images, and it has been found to be especially robust for multimodal image registration. However, MI estimators are known i) to have a very high variance and ii) to be computationally costly. In order to overcome these drawbacks, we propose a new similarity measure based
2010 2nd International Conference on Image Processing Theory, Tools and Applications, 2010
Medical Imaging 2013: Image-Guided Procedures, Robotic Interventions, and Modeling, 2013
Journal of Nuclear Medicine, 2016
Precise definition of the mitral valve plane (VP) during segmentation of the left ventricle (LV) ... more Precise definition of the mitral valve plane (VP) during segmentation of the left ventricle (LV) for single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI) quantification often requires manual adjustment, which affects quantification of perfusion. We developed a machine learning approach using support vector machines (SVM) for automatic VP placement. 392 consecutive patients undergoing fast SPECT MPI (99m)Tc-tetrofosmin stress (5 min, 300-350 MBq)/rest (5 min, 750-1050 MBq) attenuation corrected (AC) by computed tomography (CT) and same‑day coronary CT angiography (CCTA), including 48 patients with invasive angiography (ICA) and no known coronary artery disease, were studied. LV was segmented by standard clinical software (Quantitative Perfusion SPECT) by two experts, adjusting VP if needed. Two-class SVM models were computed from expert positions using 10‑fold cross validation to separate the patients used for training and for validation. SVM probability estimates were used to compute the best VP position. Automatic VP localization in AC and non-AC images were compared to expert placement on CCTA. Stress and rest total perfusion deficits (TPD) and detection of per‑vessel obstructive stenosis by ICA were also compared. VP agreement (bias, 95% confidence interval) between SVM and experts was lower than inter-expert agreement for stress-AC (1, -5-7 mm) vs. (0, -8-8 mm) and rest-AC (1, -7-10 mm) vs. (0, -10-10 mm) (p<0.01), and similar for stress non-AC (1, -4‒6 mm) vs. (0, -6-5 mm) and rest non-AC (2, -7-10 mm) vs. (-1, -9-7 mm) images (P = non-significant (NS)). For regional detection of obstructive stenosis, ischemic‑TPD areas under receiver operating characteristic curve (AUCs) (95% confidence interval) of the two experts 0.79 (0.7-0.87), 0.81 (0.73-0.89) and SVM 0.82 (0.74-0.9) for AC data were the same (P = NS) and higher than for unadjusted VP 0.63 (0.53-0.73) (p<0.01). Similarly, for non-AC data, AUCs of experts 0.77 (0.69-0.89), 0.8 (0.72-0.88) and SVM 0.79 (0.71-0.87) were the same (P = NS) and higher than for unadjusted VP 0.65 (0.56-0.75) (p<0.01). Machine learning with SVM allows automatic and accurate VP localization, decreasing user-dependence in SPECT MPI quantification.
Journal of nuclear medicine : official publication, Society of Nuclear Medicine, 2017
Atherothrombotic events in coronary arteries are most often due to rupture of unstable plaque res... more Atherothrombotic events in coronary arteries are most often due to rupture of unstable plaque resulting in myocardial infarction. Radiolabeled molecular imaging tracers directed toward cellular targets that are unique to unstable plaque can serve as a powerful tool for identifying high-risk patients and for assessing the potential of new therapeutic approaches. Two commonly available radiopharmaceuticals-(18)F-FDG and (18)F-NaF-have been used in clinical research for imaging coronary artery plaque, and ongoing clinical studies are testing whether there is an association between (18)F-NaF uptake and future atherothrombotic events. Other, less available, tracers that target macrophages, endothelial cells, and apoptotic cells have also been tested in small groups of patients. Adoption of molecular imaging of coronary plaque into clinical practice will depend on overcoming major hurdles, ultimately including evidence that the detection of unstable plaque can change patient management an...
Journal of nuclear medicine : official publication, Society of Nuclear Medicine, Jan 15, 2015
Ruptured coronary atherosclerotic plaques commonly cause acute myocardial infarction. It has been... more Ruptured coronary atherosclerotic plaques commonly cause acute myocardial infarction. It has been recently shown that active microcalcification in the coronary arteries, one of the features that characterizes vulnerable plaques at risk of rupture, can be imaged using (18)F-sodium fluoride ((18)F-NaF) PET. We aimed to determine whether a motion correction technique applied to gated (18)F-NaF PET images could enhance image quality and improve uptake estimates. Seventeen patients with myocardial infarction (n = 7) and stable angina (n = 10) underwent (18)F-NaF PET and prospective coronary CT angiography (CCTA). PET data were reconstructed in 4 different ways: (i) one gated bin (end-diastolic phase with 25% of the counts), (ii) 4 gated bins (consecutive 25% segments), (iii) 10 gated bins (consecutive 10% segments), and (iv) ungated. Subsequently, gated PET images were registered using a local and non-linear motion correction method guided by the extracted coronary arteries from CT angio...
Journal of nuclear medicine : official publication, Society of Nuclear Medicine, Jan 24, 2015
A novel PET radiotracer, flurpiridaz F-18, has undergone Phase II clinical trial evaluation as a ... more A novel PET radiotracer, flurpiridaz F-18, has undergone Phase II clinical trial evaluation as a high-resolution PET cardiac perfusion imaging agent. In a subgroup of patients imaged with this agent, we assessed the feasibility and benefit of simultaneous correction of respiratory and cardiac motion. In 16 patients, PET imaging was performed on a 4-ring Siemens Biograph-64 in dual cardiac and respiratory gating mode. Four sets of data were reconstructed with high-definition reconstruction (HD•PET): (i) ungated and (ii) 8-bin ECG-gated images using 5-min acquisition, (iii) optimal respiratory gating (ORG)-as developed for oncological imaging- using a narrow range of breathing amplitude around end-expiration level with 35% of the counts in a 7-min acquisition, and (iv) 4-bin respiratory-gated and 8-bin ECG-gated images (32 bins in total) using the 7-min acquisition (dual-gating, using all data). Motion-frozen (MF) registration algorithms were applied to ECG-gated and dual-gated data, ...
Journal of Nuclear Medicine
Objectives Hybrid PET/Computed Tomography Angiography (CTA) using 18F-sodium fluoride (18F-NaF) t... more Objectives Hybrid PET/Computed Tomography Angiography (CTA) using 18F-sodium fluoride (18F-NaF) tracer has been shown to identify ruptured and high-risk coronary atherosclerotic plaques1. However, distinction between 18F-NaF positive and negative lesions is challenging due to cardiac motion, low signal-to-noise ratio and small size of the lesions with PET uptake. We examined the feasibility of cardiac motion-frozen (MF) coronary 18F-NaF PET by nonlinear registration to aid this novel application. Methods 11 patients with acute myocardial infarction and stable angina underwent 18F-NaF PET-CT and CTA. Gated PET images, reconstructed with time-of-flight information and resolution recovery, were fused with CTA images. A total of 23 18F-NaF positive and 16 18F-NaF negative (reference) lesions were manually delineated by experienced observers guided by CTA. MF images were created from the 4 gates of PET data by a nonlinear level-set registration algorithm applied on a region of interest s...
In Image-Guided Radiation Therapy of prostate cancer, the CBCT scan acquired at each treatment fr... more In Image-Guided Radiation Therapy of prostate cancer, the CBCT scan acquired at each treatment fraction could be used to estimate a cumula-tive dose distribution thanks to non-rigid registration. However, this cumulative dose estimation is highly sensitive to non-rigid registration errors. For this rea-son, validation of the registration algorithm with organ overlap measures or visual assessment is not sufficient. In this paper, we describe the construction of a nu-merical phantom based on a finite element model of the prostate and the neighbor organs which can be used to assess the non-rigid registration accuracy. Prelimi-nary results show the potential of this phantom to better characterize registration algorithms than traditional Dice score.
Mutual Information (MI) has been extensively studied as similarity measure for the registration o... more Mutual Information (MI) has been extensively studied as similarity measure for the registration of medical images, and it has been found to be especially robust for multimodal image registration. However, MI estimators are known i) to have a very high variance and ii) to be computationally costly. In order to overcome these drawbacks, we propose a new similarity measure based
2010 2nd International Conference on Image Processing Theory, Tools and Applications, 2010
Medical Imaging 2013: Image-Guided Procedures, Robotic Interventions, and Modeling, 2013