In addressing the challenge of motion artifacts in Positron Emission Tomography (PET) lung scans, our studyintroduces the Triple Equivariant Motion Transformer (TEMT), an innovative, unsupervised, deep-learningbasedfr...In addressing the challenge of motion artifacts in Positron Emission Tomography (PET) lung scans, our studyintroduces the Triple Equivariant Motion Transformer (TEMT), an innovative, unsupervised, deep-learningbasedframework for efficient respiratory motion correction in PET imaging. Unlike traditional techniques,which segment PET data into bins throughout a respiratory cycle and often face issues such as inefficiency andoveremphasis on certain artifacts, TEMT employs Convolutional Neural Networks (CNNs) for effective featureextraction and motion decomposition.TEMT’s unique approach involves transforming motion sequences into Liegroup domains to highlight fundamental motion patterns, coupled with employing competitive weighting forprecise target deformation field generation. Our empirical evaluations confirm TEMT’s superior performancein handling diverse PET lung datasets compared to existing image registration networks. Experimental resultsdemonstrate that TEMT achieved Dice indices of 91.40%, 85.41%, 79.78%, and 72.16% on simulated geometricphantom data, lung voxel phantom data, cardiopulmonary voxel phantom data, and clinical data, respectively. Tofacilitate further research and practical application, the TEMT framework, along with its implementation detailsand part of the simulation data, is made publicly accessible at https://github.com/yehaowei/temt.展开更多
BACKGROUND Hepatopulmonary syndrome (HPS) is an arterial oxygenation defect induced by intrapulmonary vascular dilatation (IPVD) in the setting of liver disease and/or portal hypertension.This syndrome occurs most oft...BACKGROUND Hepatopulmonary syndrome (HPS) is an arterial oxygenation defect induced by intrapulmonary vascular dilatation (IPVD) in the setting of liver disease and/or portal hypertension.This syndrome occurs most often in cirrhotic patients(4%-32%) and has been shown to be detrimental to functional status,quality of life,and survival.The diagnosis of HPS in the setting of liver disease and/or portal hypertension requires the demonstration of IPVD (i.e.,diffuse or localized abnormally dilated pulmonary capillaries and pulmonary and pleural arteriovenous communications) and arterial oxygenation defects,preferably by contrast-enhanced echocardiography and measurement of the alveolar-arterial oxygen gradient,respectively.AIM To compare brain and whole-body uptake of technetium for diagnosing HPS.METHODS Sixty-nine patients with chronic liver disease and/or portal hypertension were prospectively included.Brain uptake and whole-body uptake were calculated using the geometric mean of technetium counts in the brain and lungs and in the entire body and lungs,respectively.RESULTS Thirty-two (46%) patients had IPVD as detected by contrast-enhancedechocardiography.The demographics and clinical characteristics of the patients with and without IPVD were not significantly different with the exception of the creatinine level (0.71±0.18 mg/dL vs 0.83±0.23 mg/dL;P=0.041),alveolararterial oxygen gradient (23.2±13.3 mmHg vs 16.4±14.1 mmHg;P=0.043),and arterial partial pressure of oxygen (81.0±12.1 mmHg vs 90.1±12.8 mmHg;P=0.004).Whole-body uptake was significantly higher in patients with IPVD than in patients without IPVD (48.0%±6.1%vs 40.1%±8.1%;P=0.001).The area under the curve of whole-body uptake for detecting IPVD was significantly higher than that of brain uptake (0.75 vs 0.54;P=0.025).The optimal cut-off values of brain uptake and whole-body uptake for detecting IPVD were 5.7%and 42.5%,respectively,based on Youden’s index.The sensitivity,specificity,and accuracy of brain uptake> 5.7%and whole-body uptake> 42.5%for detecting IPVD were23%,89%,and 59%and 100%,52%,and 74%,respectively.CONCLUSION Whole-body uptake is superior to brain uptake for diagnosing HPS.展开更多
Background: Computed tomography (CT) and bronchoscopy have been shown to improve the detection rates of peripheral and central lung cancers (LC), respectively. However, the performance of the combination of CT and bro...Background: Computed tomography (CT) and bronchoscopy have been shown to improve the detection rates of peripheral and central lung cancers (LC), respectively. However, the performance of the combination of CT and bronchoscopy in detecting LC, in high-risk patients, is not clear. Patients & Methods: This prospective study included 205 high-risk patients with a history of at least 2 of the following risk factors: (1) heavy smoking;(2) aero-digestive cancer;(3) pulmonary asbestosis or;(4) chronic obstructive pulmonary disease. Patients were offered chest X-ray, sputum cytology, conventional white-light followed by autofluorescence beonchoscopy (WL/AFB) and low-dose spiral CT both at baseline and follow-up visits. Results: Seven patients (3.4%) were diagnosed with LC or carcinoma in-situ (CIS) at baseline: CT evaluation detected 5 LC/CIS, while WL/AFB evaluation also identified 5 LC/CIS, 2 of which were not detected on CT. Six (85%) of these baseline lesions were early stage (0/IA). The relative-sensitivity of CT with WL/ AFB was 40% better than CT alone. On four year follow-up, 20 patients (9.8%) were diagnosed with an LC/CIS. CT with WL/AFB detected 19 cases (95%), whereas CT alone detected 15 cases (75%). Conclusion: Bimodality surveillance with spiral CT and WL/AFB can improve the detection of early stage LCs among high-risk展开更多
Purpose: To quantitatively evaluate four different Proton SFUD PBS initial planning strategies for lung mobile tumor. Methods and Materials: A virtual lung patient’s four-dimensional computed tomography (4DCT) was ge...Purpose: To quantitatively evaluate four different Proton SFUD PBS initial planning strategies for lung mobile tumor. Methods and Materials: A virtual lung patient’s four-dimensional computed tomography (4DCT) was generated in this study. To avoid the uncertainties from target delineation and imaging artifacts, a sphere with diameter of 3 cm representing a rigid mobile target (GTV) was inserted into the right side of the lung. The target motion is set in superior-inferior (SI) direction from ?5 mm to 5 mm. Four SFUD planning strategies were used based on: 1) Maximum-In-tensity-Projection Image (MIP-CT);2) CT_average with ITV overridden to muscle density (CTavg_muscle);3) CT_average with ITV overridden to tumor density (CTavg_tumor);4) CT_average without any override density (CTavg_only). Dose distributions were recalculated on each individual phase and accumulated together to assess the “actual” treatment. To estimate the impact of proton range uncertainties, +/?3.5% CT calibration curve was applied to the 4DCT phase images. Results: Comparing initial plan to the dose accumulation: MIP-CT based GTV D98 degraded 2.42 Gy (60.10 Gy vs 57.68 Gy). Heart D1 increased 6.19 Gy (1.88 Gy vs 8.07 Gy);CTavg_tumor based GTV D98 degraded 0.34 Gy (60.07 Gy vs 59.73 Gy). Heart D1 increased 2.24 Gy (3.74 Gy vs 5.98 Gy);CTavg_muscle based initial GTV D98 degraded 0.31 Gy (60.4 Gy vs 60.19 Gy). Heart D1 increased 3.44 Gy (4.38 Gy vs 7.82 Gy);CTavg_only based Initial GTV D98 degraded 6.63 Gy (60.11 Gy vs 53.48 Gy). Heart D1 increased 0.30 Gy (2.69 Gy vs 2.96 Gy);in the presence of ±3.5% range uncertainties, CTavg_tumor based plan’s accumulated GTV D98 degraded to 57.99 Gy (+3.5%) 59.38 Gy (?3.5%), and CTavg_muscle based plan’s accumulated GTV D98 degraded to 59.37 Gy (+3.5%) 59.37 Gy (?3.5%). Conclusion: This study shows that CTavg_Tumor and CTavg_Muscle based planning strategies provide the most robust GTV coverage. However, clinicians need to be aware that the actual dose to OARs at distal end of target may increase. The study also indicates that the current SFUD PBS planning strategy might not be sufficient to compensate the CT calibration uncertainty.展开更多
Early detection of lung nodule is of great importance for the successful diagnosis and treatment of lung cancer. Many researchers have tried with diverse methods, such as thresholding, computer-aided diagnosis system,...Early detection of lung nodule is of great importance for the successful diagnosis and treatment of lung cancer. Many researchers have tried with diverse methods, such as thresholding, computer-aided diagnosis system, pattern recognition technique, backpropagation algorithm, etc. Recently, convolutional neural network (CNN) finds promising applications in many areas. In this research, we investigated 3D CNN to detect early lung cancer using LUNA 16 dataset. At first, we preprocessed raw image using thresholding technique. Then we used Vanilla 3D CNN classifier to determine whether the image is cancerous or non-cancerous. The experimental results show that the proposed method can achieve a detection accuracy of about 80% and it is a satisfactory performance compared to the existing technique.展开更多
Objective: To evaluate the lung CT scan as a possible predictive diagnostic method for COVID-19 in the Cameroonian context. Methods: We designed a cross sectional study. Suspected cases of COVID-19 during the first wa...Objective: To evaluate the lung CT scan as a possible predictive diagnostic method for COVID-19 in the Cameroonian context. Methods: We designed a cross sectional study. Suspected cases of COVID-19 during the first wave at the national social insurance fund (NSIF) hospital were screened with both COVID-19 with lung CT scan and a PCR test. Univariate analysis was performed for sample description and multivariate analysis to assess the correlation between positive results for the PCR and other parameters. We estimated the optimum threshold of sensitivity/specificity, and area under curve using the empirical method and package. Results: A total of 62 suspected COVID-19 cases were recorded, predominantly males (Sex Ratio = 2.2) with a median age of 58.5 (IQR = 19.7). Among our 62 patients, 29 (46.8%) were confirmed COVID-19 cases with positive PCR results. All the patients had a thorax CT scan with a median impairment of 40% (IQR = 20%). The optimum threshold estimate for CT scan for COVID-19 infection diagnosis was 60% (95% CI = 25% - 80%). Overall, the sensitivity and specificity estimates were 0.30 (95% CI = 0.15 - 0.49) and 0.87 (95% CI = 0.70 - 0.96), respectively, leading to an Area Under Curve (AUC) estimate of 0.59 (95% CI = 0.46, 0.71). Conclusion: In this setting, lung CT scan was neither sensitive nor specific to predict COVID-19 disease.展开更多
基金the National Natural Science Foundation of China(No.82160347)Yunnan Provincial Science and Technology Department(No.202102AE090031)Yunnan Key Laboratory of Smart City in Cyberspace Security(No.202105AG070010).
文摘In addressing the challenge of motion artifacts in Positron Emission Tomography (PET) lung scans, our studyintroduces the Triple Equivariant Motion Transformer (TEMT), an innovative, unsupervised, deep-learningbasedframework for efficient respiratory motion correction in PET imaging. Unlike traditional techniques,which segment PET data into bins throughout a respiratory cycle and often face issues such as inefficiency andoveremphasis on certain artifacts, TEMT employs Convolutional Neural Networks (CNNs) for effective featureextraction and motion decomposition.TEMT’s unique approach involves transforming motion sequences into Liegroup domains to highlight fundamental motion patterns, coupled with employing competitive weighting forprecise target deformation field generation. Our empirical evaluations confirm TEMT’s superior performancein handling diverse PET lung datasets compared to existing image registration networks. Experimental resultsdemonstrate that TEMT achieved Dice indices of 91.40%, 85.41%, 79.78%, and 72.16% on simulated geometricphantom data, lung voxel phantom data, cardiopulmonary voxel phantom data, and clinical data, respectively. Tofacilitate further research and practical application, the TEMT framework, along with its implementation detailsand part of the simulation data, is made publicly accessible at https://github.com/yehaowei/temt.
基金Supported by National Key R and D Program of China,No.2017YFC0107800CAMS Initiative for Innovative Medicine,No.2016-12M-2-004
文摘BACKGROUND Hepatopulmonary syndrome (HPS) is an arterial oxygenation defect induced by intrapulmonary vascular dilatation (IPVD) in the setting of liver disease and/or portal hypertension.This syndrome occurs most often in cirrhotic patients(4%-32%) and has been shown to be detrimental to functional status,quality of life,and survival.The diagnosis of HPS in the setting of liver disease and/or portal hypertension requires the demonstration of IPVD (i.e.,diffuse or localized abnormally dilated pulmonary capillaries and pulmonary and pleural arteriovenous communications) and arterial oxygenation defects,preferably by contrast-enhanced echocardiography and measurement of the alveolar-arterial oxygen gradient,respectively.AIM To compare brain and whole-body uptake of technetium for diagnosing HPS.METHODS Sixty-nine patients with chronic liver disease and/or portal hypertension were prospectively included.Brain uptake and whole-body uptake were calculated using the geometric mean of technetium counts in the brain and lungs and in the entire body and lungs,respectively.RESULTS Thirty-two (46%) patients had IPVD as detected by contrast-enhancedechocardiography.The demographics and clinical characteristics of the patients with and without IPVD were not significantly different with the exception of the creatinine level (0.71±0.18 mg/dL vs 0.83±0.23 mg/dL;P=0.041),alveolararterial oxygen gradient (23.2±13.3 mmHg vs 16.4±14.1 mmHg;P=0.043),and arterial partial pressure of oxygen (81.0±12.1 mmHg vs 90.1±12.8 mmHg;P=0.004).Whole-body uptake was significantly higher in patients with IPVD than in patients without IPVD (48.0%±6.1%vs 40.1%±8.1%;P=0.001).The area under the curve of whole-body uptake for detecting IPVD was significantly higher than that of brain uptake (0.75 vs 0.54;P=0.025).The optimal cut-off values of brain uptake and whole-body uptake for detecting IPVD were 5.7%and 42.5%,respectively,based on Youden’s index.The sensitivity,specificity,and accuracy of brain uptake> 5.7%and whole-body uptake> 42.5%for detecting IPVD were23%,89%,and 59%and 100%,52%,and 74%,respectively.CONCLUSION Whole-body uptake is superior to brain uptake for diagnosing HPS.
文摘Background: Computed tomography (CT) and bronchoscopy have been shown to improve the detection rates of peripheral and central lung cancers (LC), respectively. However, the performance of the combination of CT and bronchoscopy in detecting LC, in high-risk patients, is not clear. Patients & Methods: This prospective study included 205 high-risk patients with a history of at least 2 of the following risk factors: (1) heavy smoking;(2) aero-digestive cancer;(3) pulmonary asbestosis or;(4) chronic obstructive pulmonary disease. Patients were offered chest X-ray, sputum cytology, conventional white-light followed by autofluorescence beonchoscopy (WL/AFB) and low-dose spiral CT both at baseline and follow-up visits. Results: Seven patients (3.4%) were diagnosed with LC or carcinoma in-situ (CIS) at baseline: CT evaluation detected 5 LC/CIS, while WL/AFB evaluation also identified 5 LC/CIS, 2 of which were not detected on CT. Six (85%) of these baseline lesions were early stage (0/IA). The relative-sensitivity of CT with WL/ AFB was 40% better than CT alone. On four year follow-up, 20 patients (9.8%) were diagnosed with an LC/CIS. CT with WL/AFB detected 19 cases (95%), whereas CT alone detected 15 cases (75%). Conclusion: Bimodality surveillance with spiral CT and WL/AFB can improve the detection of early stage LCs among high-risk
文摘Purpose: To quantitatively evaluate four different Proton SFUD PBS initial planning strategies for lung mobile tumor. Methods and Materials: A virtual lung patient’s four-dimensional computed tomography (4DCT) was generated in this study. To avoid the uncertainties from target delineation and imaging artifacts, a sphere with diameter of 3 cm representing a rigid mobile target (GTV) was inserted into the right side of the lung. The target motion is set in superior-inferior (SI) direction from ?5 mm to 5 mm. Four SFUD planning strategies were used based on: 1) Maximum-In-tensity-Projection Image (MIP-CT);2) CT_average with ITV overridden to muscle density (CTavg_muscle);3) CT_average with ITV overridden to tumor density (CTavg_tumor);4) CT_average without any override density (CTavg_only). Dose distributions were recalculated on each individual phase and accumulated together to assess the “actual” treatment. To estimate the impact of proton range uncertainties, +/?3.5% CT calibration curve was applied to the 4DCT phase images. Results: Comparing initial plan to the dose accumulation: MIP-CT based GTV D98 degraded 2.42 Gy (60.10 Gy vs 57.68 Gy). Heart D1 increased 6.19 Gy (1.88 Gy vs 8.07 Gy);CTavg_tumor based GTV D98 degraded 0.34 Gy (60.07 Gy vs 59.73 Gy). Heart D1 increased 2.24 Gy (3.74 Gy vs 5.98 Gy);CTavg_muscle based initial GTV D98 degraded 0.31 Gy (60.4 Gy vs 60.19 Gy). Heart D1 increased 3.44 Gy (4.38 Gy vs 7.82 Gy);CTavg_only based Initial GTV D98 degraded 6.63 Gy (60.11 Gy vs 53.48 Gy). Heart D1 increased 0.30 Gy (2.69 Gy vs 2.96 Gy);in the presence of ±3.5% range uncertainties, CTavg_tumor based plan’s accumulated GTV D98 degraded to 57.99 Gy (+3.5%) 59.38 Gy (?3.5%), and CTavg_muscle based plan’s accumulated GTV D98 degraded to 59.37 Gy (+3.5%) 59.37 Gy (?3.5%). Conclusion: This study shows that CTavg_Tumor and CTavg_Muscle based planning strategies provide the most robust GTV coverage. However, clinicians need to be aware that the actual dose to OARs at distal end of target may increase. The study also indicates that the current SFUD PBS planning strategy might not be sufficient to compensate the CT calibration uncertainty.
文摘Early detection of lung nodule is of great importance for the successful diagnosis and treatment of lung cancer. Many researchers have tried with diverse methods, such as thresholding, computer-aided diagnosis system, pattern recognition technique, backpropagation algorithm, etc. Recently, convolutional neural network (CNN) finds promising applications in many areas. In this research, we investigated 3D CNN to detect early lung cancer using LUNA 16 dataset. At first, we preprocessed raw image using thresholding technique. Then we used Vanilla 3D CNN classifier to determine whether the image is cancerous or non-cancerous. The experimental results show that the proposed method can achieve a detection accuracy of about 80% and it is a satisfactory performance compared to the existing technique.
文摘Objective: To evaluate the lung CT scan as a possible predictive diagnostic method for COVID-19 in the Cameroonian context. Methods: We designed a cross sectional study. Suspected cases of COVID-19 during the first wave at the national social insurance fund (NSIF) hospital were screened with both COVID-19 with lung CT scan and a PCR test. Univariate analysis was performed for sample description and multivariate analysis to assess the correlation between positive results for the PCR and other parameters. We estimated the optimum threshold of sensitivity/specificity, and area under curve using the empirical method and package. Results: A total of 62 suspected COVID-19 cases were recorded, predominantly males (Sex Ratio = 2.2) with a median age of 58.5 (IQR = 19.7). Among our 62 patients, 29 (46.8%) were confirmed COVID-19 cases with positive PCR results. All the patients had a thorax CT scan with a median impairment of 40% (IQR = 20%). The optimum threshold estimate for CT scan for COVID-19 infection diagnosis was 60% (95% CI = 25% - 80%). Overall, the sensitivity and specificity estimates were 0.30 (95% CI = 0.15 - 0.49) and 0.87 (95% CI = 0.70 - 0.96), respectively, leading to an Area Under Curve (AUC) estimate of 0.59 (95% CI = 0.46, 0.71). Conclusion: In this setting, lung CT scan was neither sensitive nor specific to predict COVID-19 disease.
