Low-dose CTs Research Articles

Automated abdominal organ segmentation algorithms for non-enhanced CT for volumetry and 3D radiomics analysis

Abstract Purpose To develop fully-automated abdominal organ segmentation algorithms from non-enhanced abdominal CT and low-dose chest CT and assess their feasibility for automated CT volumetry and 3D radiomics analysis of abdominal solid organs. Methods Fully-automated nnU-Net-based models were developed to segment the liver, spleen, and both kidneys in non-enhanced abdominal CT, and the liver and spleen in low-dose chest CT. 105 abdominal CTs and 60 low-dose chest CTs were used for model development, and 55 abdominal CTs and 10 low-dose chest CTs for external testing. The segmentation performance for each organ was assessed using the Dice similarity coefficients, with manual segmentation results serving as the ground truth. Agreements between ground-truth measurements and model estimates of organ volume and 3D radiomics features were assessed using the Bland–Altman analysis and intraclass correlation coefficients (ICC). Results The models accurately segmented the liver, spleen, right kidney, and left kidney in abdominal CT and the liver and spleen in low-dose chest CT, showing mean Dice similarity coefficients in the external dataset of 0.968, 0.960, 0.952, and 0.958, respectively, in abdominal CT, and 0.969 and 0.960, respectively, in low-dose chest CT. The model-estimated and ground truth volumes of these organs exhibited mean differences between − 0.7% and 2.2%, with excellent agreements. The automatically extracted mean and median Hounsfield units (ICCs, 0.970–0.999 and 0.994–0.999, respectively), uniformity (ICCs, 0.985–0.998), entropy (ICCs, 0.931–0.993), elongation (ICCs, 0.978–0.992), and flatness (ICCs, 0.973–0.997) showed excellent agreement with ground truth measurements for each organ; however, skewness (ICCs, 0.210–0.831), kurtosis (ICCs, 0.053–0.933), and sphericity (ICCs, 0.368–0.819) displayed relatively low and inconsistent agreement. Conclusion Our nnU-Net-based models accurately segmented abdominal solid organs in non-enhanced abdominal and low-dose chest CT, enabling reliable automated measurements of organ volume and specific 3D radiomics features.

Deep Learning-Based Kernel Adaptation Enhances Quantification of Emphysema on Low-Dose Chest CT for Predicting Long-Term Mortality.

The aim of this study was to ascertain the predictive value of quantifying emphysema using low-dose computed tomography (LDCT) post deep learning-based kernel adaptation on long-term mortality. This retrospective study investigated LDCTs obtained from asymptomatic individuals aged 60 years or older during health checkups between February 2009 and December 2016. These LDCTs were reconstructed using a 1- or 1.25-mm slice thickness alongside high-frequency kernels. A deep learning algorithm, capable of generating CT images that resemble standard-dose and low-frequency kernel images, was applied to these LDCTs. To quantify emphysema, the lung volume percentage with an attenuation value less than or equal to -950 Hounsfield units (LAA-950) was gauged before and after kernel adaptation. Low-dose chest CTs with LAA-950 exceeding 6% were deemed emphysema-positive according to the Fleischner Society statement. Survival data were sourced from the National Registry Database at the close of 2021. The risk of nonaccidental death, excluding causes such as injury or poisoning, was explored according to the emphysema quantification results using multivariate Cox proportional hazards models. The study comprised 5178 participants (mean age ± SD, 66 ± 3 years; 3110 males). The median LAA-950 (18.2% vs 2.6%) and the proportion of LDCTs with LAA-950 exceeding 6% (96.3% vs 39.3%) saw a significant decline after kernel adaptation. There was no association between emphysema quantification before kernel adaptation and the risk of nonaccidental death. Nevertheless, after kernel adaptation, higher LAA-950 (hazards ratio for 1% increase, 1.01; P = 0.045) and LAA-950 exceeding 6% (hazards ratio, 1.36; P = 0.008) emerged as independent predictors of nonaccidental death, upon adjusting for age, sex, and smoking status. The application of deep learning for kernel adaptation proves instrumental in quantifying pulmonary emphysema on LDCTs, establishing itself as a potential predictive tool for long-term nonaccidental mortality in asymptomatic individuals.

Accuracy of Computer-Guided Implantology with Pilot Drill Surgical Guide: Retrospective 3D Radiologic Investigation in Partially Edentulous Patients.

Background and Objectives: Implant placement with static navigation enables the reaching of a correct position of implants from an anatomical and prosthetic point of view. Different approaches of static navigation are described in the scientific literature, and the pilot-guided approach is one of the least investigated. The aim of the present study is the evaluation of the accuracy of implant insertion using a pilot drill template. Materials and Methods: Fifteen partially edentulous patients, requiring an implant rehabilitation of at least one implant, were enrolled. Pre- and post-operative low-dose CTs were acquired to measure the differences between final positions of implants and virtually planned ones. Three linear discrepancies (coronal, apical, and depth), two angular ones (bucco-lingual and mesio-distal), and the imprecision area were evaluated. Correlations between accuracy and rehabilitated jaws, sectors, and implant length and diameters were also analyzed. Results: Forty implants were inserted in fifteen patients using pilot drill templates. Mean coronal deviation was 1.08 mm, mean apical deviation was 1.77 mm, mean depth deviation was -0.48 mm, mean bucco-lingual angular deviation was 4.75°, and mean mesio-distal one was 5.22°. The accuracy was statistically influenced only by the rehabilitated jaw for coronal discrepancy and sectors and implant diameter for bucco-lingual angular deviations. Conclusions: The pilot drill template could represent a predictable solution to obtain a correct implant placement. Nonetheless, a safety margin of at least 2 mm should be respected during implant planning to prevent damages to anatomical structures. Therefore, the tool is helpful in order to prosthetically drive the implants; still, great attention must be paid in fully relying on this procedure when approaching dangerous structures such as nerves and vessels.

A Technical Comparison of YOLO-Based Chest Cancer Diagnosis Methods

Cancers have become one of the deadliest diseases in the world, and early diagnosis becomes vital for a patient's survival. As deep learning advances, YOLO has become an attractive tool as it supports real-time interactions. Thus, YOLO is expected to be applied in cancer diagnosis. A technical study of a YOLO-based computer aid diagnosis system for chest cancers is presented in the paper. Four kinds of the image in cancer diagnosis, histopathological images, mammograms, CTs, and Low-dose CTs, are introduced. Three issues of implementing a computer aid diagnosis system (CAD) are discussed and analyzed, including the usage of handcrafted features, the high false positive rate in clinical practice, and difficulty in detecting irregular nodules in spiral CTs. In discussion, the drawback of handcrafted features in the region of interest (ROI) extraction can be addressed by applying extra architectures like ResNet50 as extractors. A trained network can serve as a non-nodule filter to reduce the false positive rate in diagnosis. Image data can be categorized based on morphological features in data preprocessing to train a more sensitive model, then irregular-shape nodules can be detected by CAD.

Evaluation of emphysema on thoracic low-dose CTs through attention-based multiple instance deep learning

In addition to lung cancer, other thoracic abnormalities, such as emphysema, can be visualized within low-dose CT scans that were initially obtained in cancer screening programs, and thus, opportunistic evaluation of these diseases may be highly valuable. However, manual assessment for each scan is tedious and often subjective, thus we have developed an automatic, rapid computer-aided diagnosis system for emphysema using attention-based multiple instance deep learning and 865 LDCTs. In the task of determining if a CT scan presented with emphysema or not, our novel Transfer AMIL approach yielded an area under the ROC curve of 0.94 ± 0.04, which was a statistically significant improvement compared to other methods evaluated in our study following the Delong Test with correction for multiple comparisons. Further, from our novel attention weight curves, we found that the upper lung demonstrated a stronger influence in all scan classes, indicating that the model prioritized upper lobe information. Overall, our novel Transfer AMIL method yielded high performance and provided interpretable information by identifying slices that were most influential to the classification decision, thus demonstrating strong potential for clinical implementation.

Low-dose CT allows for accurate proton therapy dose calculation and plan optimization

Objective. Protons offer a more conformal dose delivery compared to photons, yet they are sensitive to anatomical changes over the course of treatment. To minimize range uncertainties due to anatomical variations, a new CT acquisition at every treatment session would be paramount to enable daily dose calculation and subsequent plan adaptation. However, the series of CT scans results in an additional accumulated patient dose. Reducing CT radiation dose and thereby decreasing the potential risk of radiation exposure to patients is desirable, however, lowering the CT dose results in a lower signal-to-noise ratio and therefore in a reduced quality image. We hypothesized that the signal-to-noise ratio provided by conventional CT protocols is higher than needed for proton dose distribution estimation. In this study, we aim to investigate the effect of CT imaging dose reduction on proton therapy dose calculations and plan optimization. Approach. To verify our hypothesis, a CT dose reduction simulation tool has been developed and validated to simulate lower-dose CT scans from an existing standard-dose scan. The simulated lower-dose CTs were then used for proton dose calculation and plan optimization and the results were compared with those of the standard-dose scan. The same strategy was adopted to investigate the effect of CT dose reduction on water equivalent thickness (WET) calculation to quantify CT noise accumulation during integration along the beam. Main results. The similarity between the dose distributions acquired from the low-dose and standard-dose CTs was evaluated by the dose-volume histogram and the 3D Gamma analysis. The results on an anthropomorphic head phantom and three patient cases indicate that CT imaging dose reduction up to 90% does not have a significant effect on proton dose calculation and plan optimization. The relative error was employed to evaluate the similarity between WET maps and was found to be less than 1% after reducing the CT imaging dose by 90%. Significance. The results suggest the possibility of using low-dose CT for proton therapy dose estimation, since the dose distributions acquired from the standard-dose and low-dose CTs are clinically equivalent.

A deep learning algorithm for detecting lytic bone lesions of multiple myeloma on CT.

Whole-body low-dose CT is the recommended initial imaging modality to evaluate bone destruction as a result of multiple myeloma. Accurate interpretation of these scans to detect small lytic bone lesions is time intensive. A functional deep learning) algorithm to detect lytic lesions on CTs could improve the value of these CTs for myeloma imaging. Our objectives were to develop a DL algorithm and determine its performance at detecting lytic lesions of multiple myeloma. Axial slices (2-mm section thickness) from whole-body low-dose CT scans of subjects with biochemically confirmed plasma cell dyscrasias were included in the study. Data were split into train and test sets at the patient level targeting a 90%/10% split. Two musculoskeletal radiologists annotated lytic lesions on the images with bounding boxes. Subsequently, we developed a two-step deep learning model comprising bone segmentation followed by lesion detection. Unet and "You Look Only Once" (YOLO) models were used as bone segmentation and lesion detection algorithms, respectively. Diagnostic performance was determined using the area under the receiver operating characteristic curve (AUROC). Forty whole-body low-dose CTs from 40 subjects yielded 2193 image slices. A total of 5640 lytic lesions were annotated. The two-step model achieved a sensitivity of 91.6% and a specificity of 84.6%. Lesion detection AUROC was 90.4%. We developed a deep learning model that detects lytic bone lesions of multiple myeloma on whole-body low-dose CTs with high performance. External validation is required prior to widespread adoption in clinical practice.

Improving reproducibility and performance of radiomics in low-dose CT using cycle GANs.

BackgroundAs a means to extract biomarkers from medical imaging, radiomics has attracted increased attention from researchers. However, reproducibility and performance of radiomics in low‐dose CT scans are still poor, mostly due to noise. Deep learning generative models can be used to denoise these images and in turn improve radiomics’ reproducibility and performance. However, most generative models are trained on paired data, which can be difficult or impossible to collect.PurposeIn this article, we investigate the possibility of denoising low‐dose CTs using cycle generative adversarial networks (GANs) to improve radiomics reproducibility and performance based on unpaired datasets.Methods and materialsTwo cycle GANs were trained: (1) from paired data, by simulating low‐dose CTs (i.e., introducing noise) from high‐dose CTs and (2) from unpaired real low dose CTs. To accelerate convergence, during GAN training, a slice‐paired training strategy was introduced. The trained GANs were applied to three scenarios: (1) improving radiomics reproducibility in simulated low‐dose CT images and (2) same‐day repeat low dose CTs (RIDER dataset), and (3) improving radiomics performance in survival prediction. Cycle GAN results were compared with a conditional GAN (CGAN) and an encoder–decoder network (EDN) trained on simulated paired data.ResultsThe cycle GAN trained on simulated data improved concordance correlation coefficients (CCC) of radiomic features from 0.87 (95%CI, [0.833,0.901]) to 0.93 (95%CI, [0.916,0.949]) on simulated noise CT and from 0.89 (95%CI, [0.881,0.914]) to 0.92 (95%CI, [0.908,0.937]) on the RIDER dataset, as well improving the area under the receiver operating characteristic curve (AUC) of survival prediction from 0.52 (95%CI, [0.511,0.538]) to 0.59 (95%CI, [0.578,0.602]). The cycle GAN trained on real data increased the CCCs of features in RIDER to 0.95 (95%CI, [0.933,0.961]) and the AUC of survival prediction to 0.58 (95%CI, [0.576,0.596]).ConclusionThe results show that cycle GANs trained on both simulated and real data can improve radiomics’ reproducibility and performance in low‐dose CT and achieve similar results compared to CGANs and EDNs.

Evaluation of Alternative Diagnostic Follow-up Intervals for Lung Reporting and Data System Criteria on the Effectiveness of Lung Cancer Screening.

The ACR developed the Lung CT Screening Reporting and Data System (Lung-RADS) to standardize the diagnostic follow-up of suspicious screening findings. A retrospective analysis showed that Lung-RADS would have reduced the false-positive rate in the National Lung Screening Trial, but the optimal timing of follow-up examinations has not been established. In this study, we assess the effectiveness of alternative diagnostic follow-up intervals on lung cancer screening. We used the Lung Cancer Outcome Simulator to estimate population-level outcomes of alternative diagnostic follow-up intervals for Lung-RADS categories 3 and 4A. The Lung Cancer Outcome Simulator is a microsimulation model developed within the Cancer Intervention and Surveillance Modeling Network Consortium to evaluate outcomes of national screening guidelines. Here, among the evaluated outcomes are percentage of mortality reduction, screens performed, lung cancer deaths averted, screen-detected cases, and average number of screens and follow-ups per death averted. The recommended 3-month follow-up interval for Lung-RADS category 4A is optimal. However, for Lung-RADS category 3, a 5-month, instead of the recommended 6-month, follow-up interval yielded a higher mortality reduction (0.08% for men versus 0.05% for women), and a higher number of deaths averted (36 versus 27), a higher number of screen-detected cases (13 versus 7), and a lower number of combined low-dose CTs and diagnostic follow-ups per death avoided (8 versus 5), per one million general population. Sensitivity analysis of nodule progression threshold verifies a higher mortality reduction with a 1-month earlier follow-up for Lung-RADS3. One-month earlier diagnostic follow-ups for individuals with Lung-RADS category 3 nodules may result in a higher mortality reduction and warrants further investigation.

Additional ultra-low-dose protocols in coronary computed tomography angiography scans to early detection of lung cancer

Abstract Background Chest CT screening improves lung cancer detection at early stages for high-risk populations, such as heavy-smokers. However, CT screening also has some considerable disadvantages as radiation exposure, patient distress and overdiagnosis. Tobacco is also a major risk factor for coronary heart disease. Purpose To assess whether an additional chest ultra-low-dose CT scan to the coronary CT angiography protocol can be used for lung cancer screening among patients with suspected coronary artery disease. Methods 980 patients underwent coronary CT angiography for assessment of coronary artery disease, additionally undergoing ultra-low-dose CT screening to early diagnosis of lung cancer in the same scanner (80kVp and 15mAs). Patients presenting pulmonary nodules were followed-up for two years, repeating low-dose CTs in intervals of 3, 6, or 12 months based on nodule size and growth rate in accordance with National Comprehensive Cancer Network guidelines. Results Ultra-low-dose CT identified 437 patients with solitary pulmonary nodules (45%), with a mean diameter of 5.40±4.10mm. One hundred fifty two were >6mm, and in 83% (n=126) of these cases they were false positive findings, further confirmed by follow-up (n=114), resection (n=3), or biopsy (n=9). Lung cancer was detected in twenty nine patients due to CT screening (diagnostic yield: 2.9%). Among these, 20 cases could not be detected in the cardiac field of view. Most patients were in early stages of the disease, and eight patients diagnosed at advanced stages (IIIB and IV). The addition of the ultra-low-dose CT scan represented a radiation dose increment of 1.32±0.29% (effective dose, 0.12±0.04mSv). Conclusions Lung cancer might be detected using additional ultra-low-dose protocols in coronary CT angiography scans among patients with suspected coronary artery disease. Funding Acknowledgement Type of funding sources: None.

Automatic pulmonary vessel segmentation on noncontrast chest CT: deep learning algorithm developed using spatiotemporally matched virtual noncontrast images and low-keV contrast-enhanced vessel maps.

ObjectivesTo develop a deep learning–based pulmonary vessel segmentation algorithm (DLVS) from noncontrast chest CT and to investigate its clinical implications in assessing vascular remodeling of chronic obstructive lung disease (COPD) patients.MethodsFor development, 104 pulmonary CT angiography scans (49,054 slices) using a dual-source CT were collected, and spatiotemporally matched virtual noncontrast and 50-keV images were generated. Vessel maps were extracted from the 50-keV images. The 3-dimensional U-Net-based DLVS was trained to segment pulmonary vessels (with a vessel map as the output) from virtual noncontrast images (as the input). For external validation, vendor-independent noncontrast CT images (n = 14) and the VESSEL 12 challenge open dataset (n = 3) were used. For each case, 200 points were selected including 20 intra-lesional points, and the probability value for each point was extracted. For clinical validation, we included 281 COPD patients with low-dose noncontrast CTs. The DLVS-calculated volume of vessels with a cross-sectional area < 5 mm2 (PVV5) and the PVV5 divided by total vessel volume (%PVV5) were measured.ResultsDLVS correctly segmented 99.1% of the intravascular points (1,387/1,400) and 93.1% of the extravascular points (1,309/1,400). The areas-under-the receiver-operating characteristic curve (AUROCs) were 0.977 and 0.969 for the two external validation datasets. For the COPD patients, both PPV5 and %PPV5 successfully differentiated severe patients whose FEV1 < 50 (AUROCs; 0.715 and 0.804) and were significantly correlated with the emphysema index (Ps < .05).ConclusionsDLVS successfully segmented pulmonary vessels on noncontrast chest CT by utilizing spatiotemporally matched 50-keV images from a dual-source CT scanner and showed promising clinical applicability in COPD.Key Points • We developed a deep learning pulmonary vessel segmentation algorithm using virtual noncontrast images and 50-keV enhanced images produced by a dual-source CT scanner. • Our algorithm successfully segmented vessels on diseased lungs. • Our algorithm showed promising results in assessing the loss of small vessel density in COPD patients. Supplementary InformationThe online version contains supplementary material available at 10.1007/s00330-021-08036-z.

Artificial intelligence in COPD: Possible applications and future prospects.

Artificial intelligence in COPD: Possible applications and future prospects.

Interobserver variability in Lung CT Screening Reporting and Data System categorisation in subsolid nodule-enriched lung cancer screening CTs.

To assess interobserver agreement in Lung CT Screening Reporting and Data System (Lung-RADS) categorisation in subsolid nodule-enriched low-dose screening CTs. A retrospective review of low-dose screening CT reports from 2013 to 2017 using keyword searches for subsolid nodules identified 54 baseline CT scans. With an additional 108 negative screening CT scans, a total of 162 CT scans were categorised according to the Lung-RADS by two fellowship-trained thoracic radiologists in consensus. We randomly selected 20, 20, 10, and 10 scans from categories 1/2, 3, 4A, and 4B CT scans, respectively, to ensure balanced category representation. Five radiologists classified the 60 CT scans into Lung-RADS categories. The frequencies of concordance and minor and major discordance were calculated, with major discordance defined as at least 6 months of management discrepancy. We used Cohen's κ statistics to analyse reader agreement. An average of 60.3% (181 of 300) of all cases and 45.0% (90 of 200) of positive screens were correctly categorised. The minor and major discordance rates were 12.3% and 27.3% overall and 18.5% and 36.5% in positive screens, respectively. The concordance rate was significantly higher among experienced thoracic radiologists. Overall, the interobserver agreement was moderate (mean κ, 0.45; 95% confidence interval: 0.40-0.51). The proportion of part-solid risk-dominant nodules was significantly higher in cases with low rates of accurate categorisation. This retrospective study observed variable accuracy and moderate interobserver agreement in radiologist categorisation of subsolid nodules in screening CTs. This inconsistency may affect management recommendations for lung cancer screening. • Diagnostic performance for Lung-RADS categorisation is variable among radiologists with fair to moderate interobserver agreement in subsolid nodule-enriched CT scans. • Experienced thoracic radiologists showed more accurate and consistent Lung-RADS categorisation than radiology residents. • The relative abundance of part-solid nodules was a potential factor related to increased disagreement in Lung-RADS categorisation.

Scoring Osteoarthritis Reliably in Large Joints and the Spine Using Whole-Body CT: OsteoArthritis Computed Tomography-Score (OACT-Score).

A standardized method to assess structural osteoarthritis (OA) burden thorough the body lacks from literature. Such a method can be valuable in developing personalized treatments for OA. We developed a reliable scoring system to evaluate OA in large joints and the spine—the OsteoArthritis Computed Tomography (OACT) score, using a convenience sample of 197 whole-body low-dose non-contrast CTs. An atlas, containing example images as reference points for training and scoring, are presented. Each joint was graded between 0–3. The total OA burden was calculated by summing scores of individual joints. Intra- and inter-observer reliability was tested 25 randomly selected scans (N = 600 joints). Intra-observer reliability and inter-observer reliability between three observers was assessed using intraclass correlation coefficient (ICC) and square-weighted kappa statistics. The square-weighted kappa for intra-observer reliability for OACT-score at joint-level ranged from 0.79 to 0.95; the ICC for the total OA grade was 0.97 (95%-CI, 0.94 to 0.99). Square-weighted kappa for interobserver reliability ranged from 0.48 to 0.95; the ICC for the total OA grade was 0.95 (95%-CI, 0.90 to 0.98). The OACT score, a new reproducible CT-based grading system reflecting OA burden in large joints and the spine, has a satisfactory reproducibility. The atlas can be used for research purposes, training, educational purposes and systemic grading of OA on CT-scans.

Variability in interpretation of low-dose chest CT using computerized assessment in a nationwide lung cancer screening program: comparison of prospective reading at individual institutions and retrospective central reading.

To evaluate the degree of variability in computer-assisted interpretation of low-dose chest CTs (LDCTs) among radiologists in a nationwide lung cancer screening (LCS) program, through comparison with a retrospective interpretation from a central laboratory. Consecutive baseline LDCTs (n = 3353) from a nationwide LCS program were investigated. In the institutional reading, 20 radiologists in 14 institutions interpreted LDCTs using computer-aided detection and semi-automated segmentation systems for lung nodules. In the retrospective central review, a single radiologist re-interpreted all LDCTs using the same system, recording any non-calcified nodules ≥ 3mm without arbitrary rejection of semi-automated segmentation to minimize the intervention of radiologist's discretion. Positive results (requiring additional follow-up LDCTs or diagnostic procedures) were initially classified by the lung CT screening reporting and data system (Lung-RADS) during the interpretation, while the classifications based on the volumetric criteria from the Dutch-Belgian lung cancer screening trial (NELSON) were retrospectively applied. Variabilities in positive rates were assessed with coefficients of variation (CVs). In the institutional reading, positive rates by the Lung-RADS ranged from 7.5 to 43.3%, and those by the NELSON ranged from 11.4 to 45.0% across radiologists. The central review exhibited higher positive rates by Lung-RADS (20.0% vs. 27.3%; p < .001) and the NELSON (23.1% vs. 37.0%; p < .001), and lower inter-institution variability (CV, 0.30 vs. 0.12, p = .003 by Lung-RADS; CV, 0.25 vs. 0.12, p = .014 by the NELSON) compared to the institutional reading. Considerable inter-institution variability in the interpretation of LCS results is caused by different usage of the computer-assisted system. • Considerable variability existed in the interpretation of screening LDCT among radiologists partly from the different usage of the computerized system. • A retrospective reading of low-dose chest CTs in the central laboratory resulted in reduced variability but an increased positive rate.

Machine Learning Facilitates Hotspot Classification in PSMA-PET/CT with Nuclear Medicine Specialist Accuracy.

Gallium-68 prostate-specific membrane antigen positron emission tomography (68Ga-PSMA-PET) is a highly sensitive method to detect prostate cancer (PC) metastases. Visual discrimination between malignant and physiologic/unspecific tracer accumulation by a nuclear medicine (NM) specialist is essential for image interpretation. In the future, automated machine learning (ML)-based tools will assist physicians in image analysis. The aim of this work was to develop a tool for analysis of 68Ga-PSMA-PET images and to compare its efficacy to that of human readers. Five different ML methods were compared and tested on multiple positron emission tomography/computed tomography (PET/CT) data-sets. Forty textural features extracted from both PET- and low-dose CT data were analyzed. In total, 2419 hotspots from 72 patients were included. Comparing results from human readers to those of ML-based analyses, up to 98% area under the curve (AUC), 94% sensitivity (SE), and 89% specificity (SP) were achieved. Interestingly, textural features assessed in native low-dose CT increased the accuracy significantly. Thus, ML based on 68Ga-PSMA-PET/CT radiomics features can classify hotspots with high precision, comparable to that of experienced NM physicians. Additionally, the superiority of multimodal ML-based analysis considering all PET and low-dose CT features was shown. Morphological features seemed to be of special additional importance even though they were extracted from native low-dose CTs.

Towards large-scale case-finding: training and validation of residual networks for detection of chronic obstructive pulmonary disease using low-dose CT

Chronic obstructive pulmonary disease (COPD) is underdiagnosed in the community. Thoracic CT scans are widely used for diagnostic and screening purposes for lung cancer. In this proof-of-concept study, we aimed to evaluate a software pipeline for the automated detection of COPD, based on deep learning and a dataset of low-dose CTs that were performed for early detection of lung cancer. We examined the use of deep residual networks, a type of artificial residual network, for the automated detection of COPD. Three versions of the residual networks were independently trained to perform COPD diagnosis using random subsets of CT scans collected from the PanCan study, which enrolled ex-smokers and current smokers at high risk of lung cancer, and evaluated the networks using three-fold cross-validation experiments. External validation was performed using 2153 CT scans acquired from a separate cohort of individuals with COPD in the ECLIPSE study. Spirometric data were used to define COPD, with stages defined according to the GOLD criteria. The best performing networks achieved an area under the receiver operating characteristic curve (AUC) of 0·889 (SD 0·017) in three-fold cross-validation experiments. When the same set of networks was applied to the ECLIPSE cohort without any modifications to the trained models, they achieved an AUC of 0·886 (0·017), a positive predictive value of 0·847 (0·056), and a negative predictive value of 0·755 (0·097), which is a greater performance than the best quantitative CT measure, the percentage of lung volumes of less than or equal to -950 Hounsfield units (AUC 0·742). Our proposed approach could identify patients with COPD among ex-smokers and current smokers without a previous diagnosis of COPD, with clinically acceptable performance. The use of deep residual networks on chest CT scans could be an effective case-finding tool for COPD detection and diagnosis, particularly in ex-smokers and current smokers who are being screened for lung cancer. Data Science Institute, University of British Columbia; Canadian Institutes of Health Research.

Computed Tomography Radiation Exposure Among Referred Kidney Stone Patients: Results from the Registry for Stones of the Kidney and Ureter.

Purpose: Kidney stone patients routinely have CT scans during diagnostic work-up before being referred to a tertiary center. How often these patients exceed the recommended dose limits for occupational radiation exposure of >100 mSv for 5 years and >50 mSv in a single year from CT alone remains unknown. This study aimed to quantify radiation doses from CTs received by stone patients before their evaluation at a tertiary care stone clinic. Methods: From November 2015 to March 2017, consecutive new patients enrolled into the Registry for Stones of the Kidney and Ureter (ReSKU™) had the dose-length product of every available CT abdomen/pelvis within 5 years of their initial visit recorded, allowing for an effective dose (EDose) calculation. Multivariate logistic regression analysis identified factors associated with exceeding recommended dose limits. Models were created to test radiation reducing effects of low-dose and phase-reduction CT protocols. Results: Of 343 noncontrast CTs performed, only 29 (8%) were low-dose CTs (calculated EDose <4 mSv). Among 389 total patients, 101 (26%) and 25 (6%) had an EDose >20 mSv and >50 mSv/year, respectively. Increased body mass index, number of scans, and multiphase scans were associated with exceeding exposure thresholds (p < 0.01). The implementation of a low-dose CT protocol decreased the estimated number of scans contributing to overexposure by >50%. Conclusions: Stone patients referred to a tertiary stone center may receive excessive radiation from CT scans alone. Unnecessary phases and underutilization of low-dose CT protocols continue to take place. Enacting new approaches to CT protocols may spare stone patients from exceeding recommended dose limits.

Ultra‐low‐dose CT image denoising using modified BM3D scheme tailored to data statistics

It is important to enhance image quality for low-dose CT acquisitions to push the ALARA boundary. Current state-of-the-art block-matching three-dimensional (BM3D) denoising scheme assumes white Gaussian noise (WGN) model. This study proposes a novel filtering module to be incorporated into the BM3D framework for ultra-low-dose CT denoising, by accounting for its specific power spectral properties. In the current BM3D algorithm, the Wiener filtering is applied in the transform domain to a post-thresholding signal for enhanced denoising. However, unlike most natural/synthetic images, low-dose CTs do not obey the ideal Gaussian noise model. Based on the specific noise properties of ultra-low-dose CT, we derive the optimal transform-domain coefficients of Wiener filter based on the minimum mean-square-error (MMSE) criterion, taking the noise spectrum and the signal/noise cross spectrum into consideration. In the absence of ground-truth signal, the hard-thresholding denoising module in the previous stage is used as a plug-in estimator. We evaluate the denoising performance on thoracic CT image datasets containing paired full-dose and ultra-low-dose images simulated by a well-validated clinical engine (or pipeline). We also assess its clinical implication by applying the denoising methods to the emphysema quantification task. Our modified BM3D method is compared with the current one, using peak signal-to-noise ratio (PSNR) and emphysema scoring results as evaluation metrics. The noise in ultra-low-dose CT presented distinct non-Gaussian characteristics and was correlated with image intensity. Performance evaluation showed that the current Wiener filter in basic BM3D algorithm yielded little denoising enhancement on ultra-low-dose CT images. In contrast, the proposed Wiener filter achieved (1.46, 1.91)dB performance gain in mean and median peak signal-to-noise ratio (PSNR) for 5%-dose image denoising and (0.93, 0.95)dB improvement for 10% dose. A paired t-test of the PSNRs between denoising using the current and the proposed Wiener filters demonstrated statistically significant improvement, yielding P-values of 1.45E-12 and 1.34E-7 on 5% and 10%-dose images, respectively. In addition, emphysema quantification on the denoised images using the modified BM3D method also had statistically significant advantage over that using the current BM3D scheme, resulting in a P-value of 6.30E-5 with the commonly used measure. This work tailors the Wiener filter in BM3D algorithm to data statistics and demonstrates statistically significant performance improvement on ultra-low-dose CT image denoising and a subsequent emphysema quantification task. Such performance gain is more pronounced with a lower dose level. The development and rationale are generally enough for other image denoising tasks when the WGN assumption is violated.

Low-dose CT imaging of the acute abdomen using model-based iterative reconstruction: a prospective study.

Performance of a modified abdominopelvic CT protocol reconstructed using full iterative reconstruction (IR) was assessed for imaging patients presenting with acute abdominal symptoms. Fifty-seven patients (17 male, 40 female; mean age of 56.5 ± 8years) were prospectively studied. Low-dose (LD) and conventional-dose (CD) CTs were contemporaneously acquired between November 2015 and March 2016. The LD and CD protocols imparted radiation exposures approximating 10-20% and 80-90% those of routine abdominopelvic CT, respectively. The LD images were reconstructed with model-based iterative reconstruction (MBIR), and CD images with hybrid IR (40% adaptive statistical iterative reconstruction (ASIR)). Image quality was assessed quantitatively and qualitatively. Independent clinical interpretations were performed with a 6-week delay between reviews. A 74.7% mean radiation dose reduction was achieved: LD effective dose (ED) 2.38 ± 1.78mSv (size-specific dose estimate (SSDE) 3.77 ± 1.97mGy); CD ED 7.04 ± 4.89mSv (SSDE 10.74 ± 5.5mGy). LD-MBIR images had significantly lower objective and subjective image noise compared with CD-ASIR (p < 0.0001). Noise reduction for LD-MBIR studies was greater for patients with BMI < 25kg/m2 than those with BMI ≥ 25kg/m2 (5.36 ± 3.2Hounsfield units (HU) vs. 4.05 ± 3.1HU, p < 0.0001). CD-ASIR studies had significantly better contrast resolution, and diagnostic acceptability (p < 0.0001 for all). LD-MBIR studies had significantly lower streak artifact (p < 0.0001). There was no difference in sensitivity for primary findings between the low-dose and conventional protocols with the exception of one case of enteritis. Low-dose abdominopelvic CT performed with MBIR is a feasible radiation dose reduction strategy for imaging patients presenting with acute abdominal pain.