Multicentre external validation of a commercial artificial intelligence software to analyse chest radiographs in health screening environments with low disease prevalence.

OBJECTIVES: To externally validate the performance of a commercial AI software program for interpreting CXRs in a large, consecutive, real-world cohort from primary healthcare centres. METHODS: A total of 3047 CXRs were collected from two primary healthcare centres, characterised by low disease prevalence, between January and December 2018. All CXRs were labelled as normal or abnormal according to CT findings. Four radiology residents read all CXRs twice with and without AI assistance. The performances of the AI and readers with and without AI assistance were measured in terms of area under the receiver operating characteristic curve (AUROC), sensitivity, and specificity. RESULTS: The prevalence of clinically significant lesions was 2.2% (68 of 3047). The AUROC, sensitivity, and specificity of the AI were 0.648 (95% confidence interval [CI] 0.630-0.665), 35.3% (CI, 24.7-47.8), and 94.2% (CI, 93.3-95.0), respectively. AI detected 12 of 41 pneumonia, 3 of 5 tuberculosis, and 9 of 22 tumours. AI-undetected lesions tended to be smaller than true-positive lesions. The readers' AUROCs ranged from 0.534-0.676 without AI and 0.571-0.688 with AI (all p values < 0.05). For all readers, the mean reading time was 2.96-10.27 s longer with AI assistance (all p values < 0.05). CONCLUSIONS: The performance of commercial AI in these high-volume, low-prevalence settings was poorer than expected, although it modestly boosted the performance of less-experienced readers. The technical prowess of AI demonstrated in experimental settings and approved by regulatory bodies may not directly translate to real-world practice, especially where the demand for AI assistance is highest. KEY POINTS: • This study shows the limited applicability of commercial AI software for detecting abnormalities in CXRs in a health screening population. • When using AI software in a specific clinical setting that differs from the training setting, it is necessary to adjust the threshold or perform additional training with such data that reflects this environment well. • Prospective test accuracy studies, randomised controlled trials, or cohort studies are needed to examine AI software to be implemented in real clinical practice.

Impact of image denoising on image quality, quantitative parameters and sensitivity of ultra-low-dose volume perfusion CT imaging.

OBJECTIVES: To examine the impact of denoising on ultra-low-dose volume perfusion CT (ULD-VPCT) imaging in acute stroke. METHODS: Simulated ULD-VPCT data sets at 20 % dose rate were generated from perfusion data sets of 20 patients with suspected ischemic stroke acquired at 80 kVp/180 mAs. Four data sets were generated from each ULD-VPCT data set: not-denoised (ND); denoised using spatiotemporal filter (D1); denoised using quanta-stream diffusion technique (D2); combination of both methods (D1 + D2). Signal-to-noise ratio (SNR) was measured in the resulting 100 data sets. Image quality, presence/absence of ischemic lesions, CBV and CBF scores according to a modified ASPECTS score were assessed by two blinded readers. RESULTS: SNR and qualitative scores were highest for D1 + D2 and lowest for ND (all p ≤ 0.001). In 25 % of the patients, ND maps were not assessable and therefore excluded from further analyses. Compared to original data sets, in D2 and D1 + D2, readers correctly identified all patients with ischemic lesions (sensitivity 1.0, kappa 1.0). Lesion size was most accurately estimated for D1 + D2 with a sensitivity of 1.0 (CBV) and 0.94 (CBF) and an inter-rater agreement of 1.0 and 0.92, respectively. CONCLUSION: An appropriate combination of denoising techniques applied in ULD-VPCT produces diagnostically sufficient perfusion maps at substantially reduced dose rates as low as 20 % of the normal scan. KEY POINTS: Perfusion-CT is an accurate tool for the detection of brain ischemias. The high associated radiation doses are a major drawback of brain perfusion CT. Decreasing tube current in perfusion CT increases image noise and deteriorates image quality. Combination of different image-denoising techniques produces sufficient image quality from ultra-low-dose perfusion CT.

Effects of radiation dose reduction in Volume Perfusion CT imaging of acute ischemic stroke.

PURPOSE: To examine the influence of radiation dose reduction on image quality and sensitivity of Volume Perfusion CT (VPCT) maps regarding the detection of ischemic brain lesions. METHODS AND MATERIALS: VPCT data of 20 patients with suspected ischemic stroke acquired at 80 kV and 180 mAs were included. Using realistic reduced-dose simulation, low-dose VPCT datasets with 144 mAs, 108 mAs, 72 mAs and 36 mAs (80 %, 60 %, 40 % and 20 % of the original levels) were generated, resulting in a total of 100 datasets. Perfusion maps were created and signal-to-noise-ratio (SNR) measurements were performed. Qualitative analyses were conducted by two blinded readers, who also assessed the presence/absence of ischemic lesions and scored CBV and CBF maps using a modified ASPECTS-score. RESULTS: SNR of all low-dose datasets were significantly lower than those of the original datasets (p < .05). All datasets down to 72 mAs (40 %) yielded sufficient image quality and high sensitivity with excellent inter-observer-agreements, whereas 36 mAs datasets (20 %) yielded poor image quality in 15 % of the cases with lower sensitivity and inter-observer-agreements. CONCLUSION: Low-dose VPCT using decreased tube currents down to 72 mAs (40 % of original radiation dose) produces sufficient perfusion maps for the detection of ischemic brain lesions. KEY POINTS: • Perfusion CT is highly accurate for the detection of ischemic brain lesions • Perfusion CT results in high radiation exposure, therefore low-dose protocols are required • Reduction of tube current down to 72 mAs produces sufficient perfusion maps.

Is a Camera-Type Portable X-Ray Device Clinically Feasible in Chest Imaging?: Image Quality Comparison with Chest Radiographs Taken with Traditional Mobile Digital X-Ray Devices.

Purpose: To evaluate whether the image quality of chest radiographs obtained using a camera-type portable X-ray device is appropriate for clinical practice by comparing them with traditional mobile digital X-ray devices. Materials and Methods: Eighty-six patients who visited our emergency department and underwent endotracheal intubation, central venous catheterization, or nasogastric tube insertion were included in the study. Two radiologists scored images captured with traditional mobile devices before insertion and those captured with camera-type devices after insertion. Identification of the inserted instruments was evaluated on a 5-point scale, and the overall image quality was evaluated on a total of 20 points scale. Results: The identification score of the instruments was 4.67 ± 0.71. The overall image quality score was 19.70 ± 0.72 and 15.02 ± 3.31 (p < 0.001) for the mobile and camera-type devices, respectively. The scores of the camera-type device were significantly lower than those of the mobile device in terms of the detailed items of respiratory motion artifacts, trachea and bronchus, pulmonary vessels, posterior cardiac blood vessels, thoracic intervertebral disc space, subdiaphragmatic vessels, and diaphragm (p = 0.013 for the item of diaphragm, p < 0.001 for the other detailed items). Conclusion: Although caution is required for general diagnostic purposes as image quality degrades, a camera-type device can be used to evaluate the inserted instruments in chest radiographs.

Evaluation of deep learning-based quantitative computed tomography for opportunistic osteoporosis screening.

To evaluate diagnostic efficacy of deep learning (DL)-based automated bone mineral density (BMD) measurement for opportunistic screening of osteoporosis with routine computed tomography (CT) scans. A DL-based automated quantitative computed tomography (DL-QCT) solution was evaluated with 112 routine clinical CT scans from 84 patients who underwent either chest (N:39), lumbar spine (N:34), or abdominal CT (N:39) scan. The automated BMD measurements (DL-BMD) on L1 and L2 vertebral bodies from DL-QCT were validated with manual BMD (m-BMD) measurement from conventional asynchronous QCT using Pearson's correlation and intraclass correlation. Receiver operating characteristic curve (ROC) analysis identified the diagnostic ability of DL-BMD for low BMD and osteoporosis, determined by dual-energy X-ray absorptiometry (DXA) and m-BMD. Excellent concordance were seen between m-BMD and DL-BMD in total CT scans (r = 0.961/0.979). The ROC-derived AUC of DL-BMD compared to that of central DXA for the low-BMD and osteoporosis patients was 0.847 and 0.770 respectively. The sensitivity, specificity, and accuracy of DL-BMD compared to central DXA for low BMD were 75.0%, 75.0%, and 75.0%, respectively, and those for osteoporosis were 68.0%, 80.5%, and 77.7%. The AUC of DL-BMD compared to the m-BMD for low BMD and osteoporosis diagnosis were 0.990 and 0.943, respectively. The sensitivity, specificity, and accuracy of DL-BMD compared to m-BMD for low BMD were 95.5%, 93.5%, and 94.6%, and those for osteoporosis were 88.2%, 94.5%, and 92.9%, respectively. DL-BMD exhibited excellent agreement with m-BMD on L1 and L2 vertebrae in the various routine clinical CT scans and had comparable diagnostic performance for detecting the low-BMD and osteoporosis on conventional QCT.

Clinical validation of a deep-learning-based bone age software in healthy Korean children.

PURPOSE: Bone age (BA) is needed to assess developmental status and growth disorders. We evaluated the clinical performance of a deep-learning-based BA software to estimate the chronological age (CA) of healthy Korean children. METHODS: This retrospective study included 371 healthy children (217 boys, 154 girls), aged between 4 and 17 years, who visited the Department of Pediatrics for health check-ups between January 2017 and December 2018. A total of 553 left-hand radiographs from 371 healthy Korean children were evaluated using a commercial deep-learning-based BA software (BoneAge, Vuno, Seoul, Korea). The clinical performance of the deep learning (DL) software was determined using the concordance rate and Bland-Altman analysis via comparison with the CA. RESULTS: A 2-sample t-test (P<0.001) and Fisher exact test (P=0.011) showed a significant difference between the normal CA and the BA estimated by the DL software. There was good correlation between the 2 variables (r=0.96, P<0.001); however, the root mean square error was 15.4 months. With a 12-month cutoff, the concordance rate was 58.8%. The Bland-Altman plot showed that the DL software tended to underestimate the BA compared with the CA, especially in children under the age of 8.3 years. CONCLUSION: The DL-based BA software showed a low concordance rate and a tendency to underestimate the BA in healthy Korean children.

Assessing lung fluid status using noninvasive bioelectrical impedance analysis in patients with acute heart failure: A pilot study.

BACKGROUND: Outpatient monitoring of pulmonary congestion in heart failure (HF) patients may reduce hospitalization rates. This study tested the feasibility of non-invasive high-frequency bioelectrical impedance analysis (HF-BIA) for estimating lung fluid status. METHODS: This prospective study included 70 participants: 50 with acute HF (HF group) and 20 without HF (control group). All participants underwent a supine chest CT scan to measure lung fluid content with lung density analysis software. Concurrently, direct segmental multi-frequency BIA was performed to assess the edema index (EI) of the trunk, entire body, and extremities. RESULTS: The correlation coefficients between lung fluid content and EI measured using HF-BIA were r = 0.566 (p < 0.001) and r = 0.550 (p < 0.001) for the trunk and whole body, respectively. In the HF group, the trunk EI (0.402 ± 0.015) and whole body EI (0.402 ± 0.016) were significantly higher than those of the control group (trunk EI, 0.383 ± 0.007; whole body EI, 0.383 ± 0.007; all p < 0.001). The lung fluid content was significantly higher in the HF than that in the control group (23.7 ± 5.3 vs. 15.5 ± 2.8%, p < 0.001). The log value of NT pro-BNP was significantly correlated with trunk EI (r = 0.688, p < 0.001) and whole-body EI (r = 0.675, p < 0.001) measured by HF-BIA, and the lung fluid content analyzed by CT (r = 0.686, p < 0.001). CONCLUSIONS: BIA-based EI measurements of the trunk and whole body significantly correlated with lung fluid content and NT pro-BNP levels. Non-invasive BIA could be a promising screening tool for lung fluid status monitoring in acute HF patients.

Deep learning-based pectoralis muscle volume segmentation method from chest computed tomography image using sagittal range detection and axial slice-based segmentation.

The pectoralis muscle is an important indicator of respiratory muscle function and has been linked to various parenchymal biomarkers, such as airflow limitation severity and diffusing capacity for carbon monoxide, which are widely used in diagnosing parenchymal diseases, including asthma and chronic obstructive pulmonary disease. Pectoralis muscle segmentation is a method for measuring muscle volume and mass for various applications. The segmentation method is based on deep-learning techniques that combine a muscle area detection model and a segmentation model. The training dataset for the detection model comprised multichannel images of patients, whereas the segmentation model was trained on 7,796 cases of the computed tomography (CT) image dataset of 1,841 patients. The dataset was expanded incrementally through an active learning process. The performance of the model was evaluated by comparing the segmentation results with manual annotations by radiologists and the volumetric differences between the CT image datasets of the same patients. The results indicated that the machine learning model is promising in segmenting the pectoralis major muscle, with good agreement between the automatic segmentation and manual annotations by radiologists. The training accuracy and loss values of the validation set were 0.9954 and 0.0725, respectively, and for segmentation, the loss value was 0.0579. This study shows the potential clinical usefulness of the machine learning model for pectoralis major muscle segmentation as a quantitative biomarker for various parenchymal and muscular diseases.

Accuracy of two deep learning-based reconstruction methods compared with an adaptive statistical iterative reconstruction method for solid and ground-glass nodule volumetry on low-dose and ultra-low-dose chest computed tomography: A phantom study.

No published studies have evaluated the accuracy of volumetric measurement of solid nodules and ground-glass nodules on low-dose or ultra-low-dose chest computed tomography, reconstructed using deep learning-based algorithms. This is an important issue in lung cancer screening. Our study aimed to investigate the accuracy of semiautomatic volume measurement of solid nodules and ground-glass nodules, using two deep learning-based image reconstruction algorithms (Truefidelity and ClariCT.AI), compared with iterative reconstruction (ASiR-V) in low-dose and ultra-low-dose settings. We performed computed tomography scans of solid nodules and ground-glass nodules of different diameters placed in a phantom at four radiation doses (120 kVp/220 mA, 120 kVp/90 mA, 120 kVp/40 mA, and 80 kVp/40 mA). Each scan was reconstructed using Truefidelity, ClariCT.AI, and ASiR-V. The solid nodule and ground-glass nodule volumes were measured semiautomatically. The gold-standard volumes could be calculated using the diameter since all nodule phantoms are perfectly spherical. Subsequently, absolute percentage measurement errors of the measured volumes were calculated. Image noise was also calculated. Across all nodules at all dose settings, the absolute percentage measurement errors of Truefidelity and ClariCT.AI were less than 11%; they were significantly lower with Truefidelity or ClariCT.AI than with ASiR-V (all P<0.05). The absolute percentage measurement errors for the smallest solid nodule (3 mm) reconstructed by Truefidelity or ClariCT.AI at all dose settings were significantly lower than those of this nodule reconstructed by ASiR-V (all P<0.05). Furthermore, the lowest absolute percentage measurement errors for ground-glass nodules were observed with Truefidelity or ClariCT.AI at all dose settings. The absolute percentage measurement errors for ground-glass nodules reconstructed with Truefidelity at ultra-low-dose settings were significantly lower than those of all sizes of ground-glass nodules reconstructed with ASiR-V (all P<0.05). Image noise was lowest with Truefidelity (all P<0.05). In conclusion, the deep learning-based algorithms were more accurate for volume measurements of both solid nodules and ground-glass nodules than ASiR-V at both low-dose and ultra-low-dose settings.

The spectrum of magnetic resonance imaging proton density fat fraction (MRI-PDFF), magnetic resonance spectroscopy (MRS), and two different histopathologic methods (artificial intelligence vs. pathologist) in quantifying hepatic steatosis.

Background: The grade of hepatic steatosis is assessed semi-quantitatively and graded as a discrete value. However, the proton density fat fraction (PDFF) measured by magnetic resonance imaging (MRI) and FF measured by MR spectroscopy (FFMRS) are continuous values. Therefore, a quantitative histopathologic method may be needed. This study aimed to (I) provide a spectrum of values of MRI-PDFF, FFMRS, and FFs measured by two different histopathologic methods [artificial intelligence (AI) and pathologist], (II) to evaluate the correlation among them, and (III) to evaluate the diagnostic performance of MRI-PDFF and MRS for grading hepatic steatosis. Methods: Forty-seven patients who underwent liver biopsy and MRI for nonalcoholic steatohepatitis (NASH) evaluation were included. The agreement between MRI-PDFF and MRS was evaluated through Bland-Altman analysis. Correlations among MRI-PDFF, MRS, and two different histopathologic methods were assessed using Pearson correlation coefficient (r). The diagnostic performance of MRI-PDFF and MRS was assessed using receiver operating characteristic curve analyses and the area under the curve (AUC) were obtained. Results: The means±standard deviation of MRI-PDFF, FFMRS, FF measured by pathologist (FFpathologist), and FF measured by AI (FFAI) were 12.04±6.37, 14.01±6.16, 34.26±19.69, and 6.79±4.37 (%), respectively. Bland-Altman bias [mean of MRS - (MRI-PDFF) differences] was 2.06%. MRI-PDFF and MRS had a very strong correlation (r=0.983, P<0.001). The two different histopathologic methods also showed a very strong correlation (r=0.872, P<0.001). Both MRI-PDFF and MRS demonstrated a strong correlation with FFpathologist (r=0.701, P<0.001 and r=0.709, P<0.001, respectively) and with FFAI (r=0.700, P<0.001 and r=0.690, P<0.001, respectively). The AUCs of MRI-PDFF for grading ≥S2 and ≥S3 were 0.846 and 0.855, respectively. The AUCs of MRS for grading ≥S2 and ≥S3 were 0.860 and 0.878, respectively. Conclusions: Since MRS and MRI-PDFF demonstrated a strong correlation with each other and with the two different histopathologic methods, they can be used as an alternative noninvasive reference standard in nonalcoholic fatty liver disease (NAFLD) patients. However, these preliminary results should be interpreted with caution until they are validated in further studies.

Usefulness of Pulmonary Rehabilitation in Non-Small Cell Lung Cancer Patients Based on Pulmonary Function Tests and Muscle Analysis Using Computed Tomography Images.

PURPOSE: The usefulness of rehabilitation in patients with reduced lung function before lung surgery remains unclear, and there is no adequate method for evaluating the effect of rehabilitation. We aimed to evaluate the usefulness of rehabilitation in patients with non-small cell lung cancer (NSCLC) undergoing lung cancer surgery. MATERIALS AND METHODS: We retrospectively analyzed the medical records of NSCLC patients at Korea University Guro Hospital between 2018 and 2020. Patients were divided into two groups depending on whether they underwent rehabilitation. Pulmonary function test (PFT) data and muscle determined using chest computed tomography (CT) images were analyzed. Because the baseline characteristics were different between the two groups, propensity score matching was performed. RESULTS: Of 325 patients, 75 (23.1%) and 250 (76.9%) were included in the rehabilitation and non-rehabilitation (control) groups, respectively. The rehabilitation group had a worse general condition at baseline. After propensity score matching, 45 patients remained in each group. Pulmonary function (forced expiratory volume in 1 second, %) (p=0.001) and the Hounsfield unit of erector spinae muscle (p=0.001) were better preserved in the rehabilitation group. Muscle loss of 3.4% and 0.6% was observed in the control and rehabilitation groups, respectively (p=0.003). In addition, the incidence of embolic events was lower in the rehabilitation group (p=0.044). CONCLUSION: Pulmonary rehabilitation is useful in patients with NSCLC undergoing lung surgery. Pulmonary rehabilitation preserves lung function, muscle and reduces embolic events after surgery. Pulmonary rehabilitation is recommended for patients with NSCLC undergoing surgery.

External validation of deep learning-based bone-age software: a preliminary study with real world data.

Artificial intelligence (AI) is increasingly being used in bone-age (BA) assessment due to its complicated and lengthy nature. We aimed to evaluate the clinical performance of a commercially available deep learning (DL)-based software for BA assessment using a real-world data. From Nov. 2018 to Feb. 2019, 474 children (35 boys, 439 girls, age 4-17 years) were enrolled. We compared the BA estimated by DL software (DL-BA) with that independently estimated by 3 reviewers (R1: Musculoskeletal radiologist, R2: Radiology resident, R3: Pediatric endocrinologist) using the traditional Greulich-Pyle atlas, then to his/her chronological age (CA). A paired t-test, Pearson's correlation coefficient, Bland-Altman plot, mean absolute error (MAE) and root mean square error (RMSE) were used for the statistical analysis. The intraclass correlation coefficient (ICC) was used for inter-rater variation. There were significant differences between DL-BA and each reviewer's BA (P < 0.025), but the correlation was good with one another (r = 0.983, P < 0.025). RMSE (MAE) values were 10.09 (7.21), 10.76 (7.88) and 13.06 (10.06) months between DL-BA and R1, R2, R3 BA. Compared with the CA, RMSE (MAE) values were 13.54 (11.06), 15.18 (12.11), 16.19 (12.78) and 19.53 (17.71) months for DL-BA, R1, R2, R3 BA, respectively. Bland-Altman plots revealed the software and reviewers' tendency to overestimate the BA in general. ICC values between 3 reviewers were 0.97, 0.85 and 0.86, and the overall ICC value was 0.93. The BA estimated by DL-based software showed statistically similar, or even better performance than that of reviewers' compared to the chronological age in the real world clinic.

Multifragmentary patellar fracture has a distinct fracture pattern which makes coronal split, inferior pole, or satellite fragments.

The present study aimed to map the location and frequency of fracture lines on the coronal articular and sagittal planes in multifragmentary patellar fractures. 66 multifragmentary patellar fractures were digitally reconstructed using the 3D CT mapping technique. The coronal articular surface and midsagittal fracture maps were produced by superimposing each case over a single template. Each fracture line was classified based on the initial displacement and orientation. We evaluated the frequency and direction of the fracture line, coronal split fragment area, and satellite and inferior pole fragment presence. Coronal articular surface fracture mapping identified primary horizontal fracture lines between the middle and inferior one-third of the articular surface in 63 patients (95.4%). Secondary horizontal fracture lines running on the inferior border of the articular facet were confirmed (83.3%). Secondary vertical fracture lines creating satellite fragments were mostly located on the periphery of the bilateral facet. Midsagittal fracture mapping of primary and secondary horizontal fracture lines with the main coronal fracture line revealed a predominantly X-shaped fracture map. The consequent coronal split fragment and inferior pole fracture were combined in most cases. In conclusion, the multifragmentary patellar fracture has a distinct pattern which makes coronal split, inferior pole, or satellite fragments.

Multilevel analysis of spatiotemporal association features for differentiation of tumor enhancement patterns in breast DCE-MRI.

PURPOSE: Analyzing spatiotemporal enhancement patterns is an important task for the differential diagnosis of breast tumors in dynamic contrast-enhanced MRI (DCE-MRI), and yet remains challenging because of complexities in analyzing the time-series of three-dimensional image data. The authors propose a novel approach to breast MRI computer-aided diagnosis (CAD) using a multilevel analysis of spatiotemporal association features for tumor enhancement patterns in DCE-MRI. METHODS: A database of 171 cases consisting of 111 malignant and 60 benign tumors was used. Time-series contrast-enhanced MR images were obtained from two different types of MR scanners and protocols. The images were first registered for motion compensation, and then tumor regions were segmented using a fuzzy c-means clustering-based method. Spatiotemporal associations of tumor enhancement patterns were analyzed at three levels: Mapping of pixelwise kinetic features within a tumor, extraction of spatial association features from kinetic feature maps, and extraction of kinetic association features at the spatial feature level. A total of 84 initial features were extracted. Predictable values of these features were evaluated with an area under the ROC curve, and were compared between the spatiotemporal association features and a subset of simple form features which do not reflect spatiotemporal association. Several optimized feature sets were identified among the spatiotemporal association feature group or among the simple feature group based on a feature ranking criterion using a support vector machine based recursive feature elimination algorithm. A least-squares support vector machine (LS-SVM) classifier was used for tumor differentiation and the performances were evaluated using a leave-one-out testing. RESULTS: Predictable values of the extracted single features ranged in 0.52-0.75. By applying multilevel analysis strategy, the spatiotemporal association features became more informative in predicting tumor malignancy, which was shown by a statistical testing in ten spatiotemporal association features. By using a LS-SVM classifier with the optimized second and third level feature set, the CAD scheme showed Az of 0.88 in classification of malignant and benign tumors. When this performance was compared to the same LS-SVM classifier with simple form features which do not reflect spatiotemporal association, there was a statistically significant difference (0.88 vs 0.79, p <0.05), suggesting that the multilevel analysis strategy yields a significant performance improvement. CONCLUSIONS: The results suggest that the multilevel analysis strategy characterizes the complex tumor enhancement patterns effectively with the spatiotemporal association features, which in turn leads to an improved tumor differentiation. The proposed CAD scheme has a potential for improving diagnostic performance in breast DCE-MRI.

Attenuation profile matching: An accurate and scan parameter-robust measurement method for small airway dimensions in low-dose CT scans.

PURPOSE: The dimensions of small airways with an internal diameter of less than 2-3 mm are important biomarkers for the evaluation of pulmonary diseases, such as asthma and chronic obstructive pulmonary disease (COPD). The resolution limitations of CT systems, however, have remained a barrier to be of use for determining the small airway dimensions. We present a novel approach, called the attenuation profile matching (APM) method, which allows for the accurate determination of the small airway dimension while being robust to varying CT scan parameters. METHOD: For generating the synthetic attenuation profiles of an airway, we acquired and employed the point spread functions of a CT system by calculating its convolution with numerical airway models with varying wall thicknesses. The dimensions of a given airway were determined as per the numerical model yielding minimum error between the measured and the synthetic attenuation profiles across the airway. RESULTS: In a phantom study with airway tubes, the APM method proved to be highly accurate in determining airway wall dimensions. The measurement error for the smallest tube (0.6 mm thickness, 3 mm diameter) was merely 0.02 mm (3.3%) in wall thickness and 0.17 mm (5.6%) in lumen diameter. In a pilot clinical test, the APM method was able to distinguish the airway wall thicknesses of COPD cases (1.16 ± 0.23 mm) from those of normal subjects (0.6 ± 0.18 mm), while the measurements using the full width at half maximum method substantially overlapped (1.45 ± 0.32 mm vs. 1.28 ± 0.30 mm, respectively) and were barely distinguishable from each other. CONCLUSION: Our proposed APM method has the potential to overcome the resolution limitations of current CT systems and accurately determine the small airway dimensions in COPD patients.