Machine learning prediction of hepatic steatosis using body composition parameters: A UK Biobank Study.

Non-alcoholic fatty liver disease (NAFLD) has emerged as the most prevalent chronic liver disease worldwide, yet detection has remained largely based on surrogate serum biomarkers, elastography or biopsy. In this study, we used a total of 2959 participants from the UK biobank cohort and established the association of dual-energy X-ray absorptiometry (DXA)-derived body composition parameters and leveraged machine learning models to predict NAFLD. Hepatic steatosis reference was based on MRI-PDFF which has been extensively validated previously. We found several significant associations with traditional measurements such as abdominal obesity, as defined by waist-to-hip ratio (OR = 2.50 (male), 3.35 (female)), android-gynoid ratio (OR = 3.35 (male), 6.39 (female)) and waist circumference (OR = 1.79 (male), 3.80 (female)) with hepatic steatosis. Similarly, A Body Shape Index (Quantile 4 OR = 1.89 (male), 5.81 (female)), and for fat mass index, both overweight (OR = 6.93 (male), 2.83 (female)) and obese (OR = 14.12 (male), 5.32 (female)) categories were likewise significantly associated with hepatic steatosis. DXA parameters were shown to be highly associated such as visceral adipose tissue mass (OR = 8.37 (male), 19.03 (female)), trunk fat mass (OR = 8.64 (male), 25.69 (female)) and android fat mass (OR = 7.93 (male), 21.77 (female)) with NAFLD. We trained machine learning classifiers with logistic regression and two histogram-based gradient boosting ensembles for the prediction of hepatic steatosis using traditional body composition indices and DXA parameters which achieved reasonable performance (AUC = 0.83-0.87). Based on SHapley Additive exPlanations (SHAP) analysis, DXA parameters that had the largest contribution to the classifiers were the features predicted with significant association with NAFLD. Overall, this study underscores the potential utility of DXA as a practical and potentially opportunistic method for the screening of hepatic steatosis.

Deep Learning in Breast Cancer Imaging: A Decade of Progress and Future Directions.

Breast cancer has reached the highest incidence rate worldwide among all malignancies since 2020. Breast imaging plays a significant role in early diagnosis and intervention to improve the outcome of breast cancer patients. In the past decade, deep learning has shown remarkable progress in breast cancer imaging analysis, holding great promise in interpreting the rich information and complex context of breast imaging modalities. Considering the rapid improvement in deep learning technology and the increasing severity of breast cancer, it is critical to summarize past progress and identify future challenges to be addressed. This paper provides an extensive review of deep learning-based breast cancer imaging research, covering studies on mammograms, ultrasound, magnetic resonance imaging, and digital pathology images over the past decade. The major deep learning methods and applications on imaging-based screening, diagnosis, treatment response prediction, and prognosis are elaborated and discussed. Drawn from the findings of this survey, we present a comprehensive discussion of the challenges and potential avenues for future research in deep learning-based breast cancer imaging.

Metabolic biomarkers using nuclear magnetic resonance metabolomics assay for the prediction of aging-related disease risk and mortality: a prospective, longitudinal, observational, cohort study based on the UK Biobank.

The identification of metabolic biomarkers for aging-related diseases and mortality is of significant interest in the field of longevity. In this study, we investigated the associations between nuclear magnetic resonance (NMR) metabolomics biomarkers and aging-related diseases as well as mortality using the UK Biobank dataset. We analyzed NMR samples from approximately 110,000 participants and used multi-head machine learning classification models to predict the incidence of aging-related diseases. Cox regression models were then applied to assess the relevance of NMR biomarkers to the risk of death due to aging-related diseases. Additionally, we conducted survival analyses to evaluate the potential improvements of NMR in predicting survival and identify the biomarkers most strongly associated with negative health outcomes by dividing participants into health, disease, and death groups for all age groups. Our analysis revealed specific metabolomics profiles that were associated with the incidence of age-related diseases, and the most significant biomarker was intermediate density lipoprotein cholesteryl (IDL-CE). In addition, NMR biomarkers could provide additional contributions to relevant mortality risk prediction when combined with conventional risk factors, by improving the C-index from 0.813 to 0.833, with 17 NMR biomarkers significantly contributing to disease-related death, such as monounsaturated fatty acids (MUFA), linoleic acid (LA), glycoprotein acetyls (GlycA), and omega-3. Moreover, the value of free cholesterol in very large HDL particles (XL-HDL-FC) in the healthy control group demonstrated significantly higher values than the disease and death group across all age groups. This study highlights the potential of NMR metabolomics profiling as a valuable tool for identifying metabolic biomarkers associated with aging-related diseases and mortality risk, which could have practical implications for aging-related disease risk and mortality prediction.

Radiologic extranodal extension for nodal staging in nasopharyngeal carcinoma.

PURPOSE: Extranodal extension (ENE) has the potential to add value to the current nodal staging system (N8th) for predicting outcome in nasopharyngeal carcinoma (NPC). This study aimed to incorporate ENE, as well as cervical nodal necrosis (CNN) to the current stage N3 and evaluated their impact on outcome prediction. The findings were validated on an external cohort. METHODS & MATERIALS: Pre-treatment MRI of 750 patients from the internal cohort were retrospectively reviewed. Predictive values of six modified nodal staging systems that incorporated four patterns of ENE and two patterns of CNN to the current stage N3 for disease-free survival (DFS) were compared with that of N8th using multivariate cox-regression and concordance statistics in the internal cohort. Performance of stage N3 for predicting disease recurrence was calculated. An external cohort of 179 patients was used to validate the findings. RESULTS: Incorporation of advanced ENE, which infiltrates into adjacent muscle/skin/salivary glands outperformed the other five modifications for predicting outcomes (p < 0.01) and achieved a significantly higher c-index for 5-year DFS (0.69 vs 0.72) (p < 0.01) when compared with that of N8th staging system. By adding advanced ENE to the current N3 increased the sensitivity for predicting disease recurrence from 22.4 % to 47.1 %. The finding was validated in the external cohort (5-year DFS 0.65 vs. 0.72, p < 0.01; sensitivity of stage N3 increased from 14.0 % to 41.9 % for disease recurrence). CONCLUSION: Results from two centre cohorts confirmed that the radiological advanced ENE should be considered as a criterion for stage N3 disease in NPC.

Design and synthesis of chiral DOTA-based MRI contrast agents with remarkable relaxivities.

Due to the adverse effects of de-metallation in past concerning FDA-approved gadolinium-based contrast agents (GBCAs), researchers have been focusing on developing safer and more efficient alternatives that could avoid toxicity caused by free gadolinium ions. Herein, two chiral GBCAs, Gd-LS with sulfonate groups and Gd-T with hydroxyl groups, are reported as potential candidates for magnetic reasonance imaging (MRI). The r1 relaxivities of TSAP, SAP isomers of Gd-LS and SAP isomer of Gd-T at 1.4 T, 37 °C in water are 7.4 mM-1s-1, 14.5 mM-1s-1 and 5.2 mM-1s-1, respectively. Results show that the hydrophilic functional groups introduced to the chiral macrocyclic scaffold of Gd-T and Gd-LS both give constructive influences on the second-sphere relaxivity and enhance the overall r1 value. Both cases indicate that the design of GBCAs should also focus on the optimal window in Solomon-Bloembergen-Morgan (SBM) theory and the effects caused by the second-sphere and outer-sphere relaxivity.

Photon-Counting CT Material Decomposition in Bone Imaging.

The accurate screening of osteoporosis is important for identifying persons at risk. The diagnosis of bone conditions using dual X-ray absorptiometry is limited to extracting areal bone mineral density (BMD) and fails to provide any structural information. Computed tomography (CT) is excellent for morphological imaging but not ideal for material quantification. Advanced photon-counting detector CT (PCD-CT) possesses high spectral sensitivity and material decomposition capabilities to simultaneously determine qualitative and quantitative information. In this study, we explored the diagnostic utility of PCD-CT to provide high-resolution 3-D imaging of bone microarchitecture and composition for the sensitive diagnosis of bone in untreated and ovariectomized rats. PCD-CT accurately decomposed the calcium content within hydroxyapatite phantoms (r = 0.99). MicroCT analysis of tibial bone revealed significant differences in the morphological parameters between the untreated and ovariectomized samples. However, differences in the structural parameters of the mandible between the treatment groups were not observed. BMD determined with microCT and calcium concentration decomposed using PCD-CT differed significantly between the treatment groups in both the tibia and mandible. Quantitative analysis with PCD-CT is sensitive in determining the distribution of calcium and water components in bone and may have utility in the screening and diagnosis of bone conditions such as osteoporosis.

Towards visceral fat estimation at population scale: correlation of visceral adipose tissue assessment using three-dimensional cross-sectional imaging with BIA, DXA, and single-slice CT.

Background: In terms of assessing obesity-associated risk, quantification of visceral adipose tissue (VAT) has become increasingly important in risk assessment for cardiovascular and metabolic diseases. However, differences exist in the accuracy of various modalities, with a lack of up-to-date comparison with three-dimensional whole volume assessment. Aims: Using CT or MRI three-dimensional whole volume VAT as a reference, we evaluated the correlation of various commonly used modalities and techniques namely body impedance analysis (BIA), dual-energy x-ray absorptiometry (DXA) as well as single slice CT to establish how these methods compare. Methods: We designed the study in two parts. First, we performed an intra-individual comparison of the 4558 participants from the UK Biobank cohorts with matching data of MRI abdominal body composition, DXA with VAT estimation, and BIA. Second, we evaluated 174 CT scans from the publicly available dataset to assess the correlation of the commonly used single-slice technique compared to three-dimensional VAT volume. Results: Across the UK Biobank cohort, the DXA-derived VAT measurement correlated better (R2 0.94, p<0.0001) than BIA (R2 0.49, p<0.0001) with reference three-dimensional volume on MRI. However, DXA-derived VAT correlation was worse for participants with a BMI of < 20 (R2 = 0.62, p=0.0013). A commonly used single slice method on CT demonstrated a modest correlation (R2 between 0.51 - 0.64), with best values at L3- and L4 (R2 L3 = 0.63, p<0.0001; L4 = 0.64, p<0.0001) compared to reference three-dimensional volume. Combining multiple slices yielded a better correlation, with a strong correlation when L2-L3 levels were combined (R2 = 0.92, p<0.0001). Conclusion: When deployed at scale, DXA-derived VAT volume measurement shows excellent correlation with three-dimensional volume on MRI based on the UK Biobank cohort. Whereas a single slice CT technique demonstrated moderate correlation with three-dimensional volume on CT, with a stronger correlation achieved when multiple levels were combined.

Optimal time for early therapeutic response prediction in nasopharyngeal carcinoma with functional magnetic resonance imaging.

Background and Purpose: Physiological changes in tumour occur much earlier than morphological changes. They can potentially be used as biomarkers for therapeutic response prediction. This study aimed to investigate the optimal time for early therapeutic response prediction with multi-parametric magnetic resonance imaging (MRI) in patients with nasopharyngeal carcinoma (NPC) receiving concurrent chemo-radiotherapy (CCRT). Material and Methods: Twenty-seven NPC patients were divided into the responder (N = 23) and the poor-responder (N = 4) groups by their primary tumour post-treatment shrinkages. Single-voxel proton MR spectroscopy (1H-MRS), diffusion-weighted (DW) and dynamic contrast-enhanced (DCE) MRI were scanned at baseline, weekly during CCRT and post-CCRT. The median choline peak in 1H-MRS, the median apparent diffusion coefficient (ADC) in DW-MRI, the median influx rate constant (Ktrans), reflux rate constant (Kep), volume of extravascular-extracellular space per unit volume (Ve), and initial area under the time-intensity curve for the first 60 s (iAUC60) in DCE-MRI were compared between the two groups with the Mann-Whitney tests for any significant difference at different time points. Results: In DW-MRI, the percentage increase in ADC from baseline to week-1 for the responders (median = 11.39%, IQR = 18.13%) was higher than the poor-responders (median = 4.91%, IQR = 7.86%) (p = 0.027). In DCE-MRI, the iAUC60 on week-2 was found significantly higher in the poor-responders (median = 0.398, IQR = 0.051) than the responders (median = 0.192, IQR = 0.111) (p = 0.012). No significant difference was found in median choline peaks in 1H-MRS at all time points. Conclusion: Early perfusion and diffusion changes occurred in primary tumours of NPC patients treated with CCRT. The DW-MRI on week-1 and the DCE-MRI on week-2 were the optimal time points for early therapeutic response prediction.

Lung Cancer Screening in Asia: An Expert Consensus Report.

INTRODUCTION: The incidence and mortality of lung cancer are highest in Asia compared with Europe and USA, with the incidence and mortality rates being 34.4 and 28.1 per 100,000 respectively in East Asia. Diagnosing lung cancer at early stages makes the disease amenable to curative treatment and reduces mortality. In some areas in Asia, limited availability of robust diagnostic tools and treatment modalities, along with variations in specific health care investment and policies, make it necessary to have a more specific approach for screening, early detection, diagnosis, and treatment of patients with lung cancer in Asia compared with the West. METHOD: A group of 19 advisors across different specialties from 11 Asian countries, met on a virtual Steering Committee meeting, to discuss and recommend the most affordable and accessible lung cancer screening modalities and their implementation, for the Asian population. RESULTS: Significant risk factors identified for lung cancer in smokers in Asia include age 50 to 75 years and smoking history of more than or equal to 20 pack-years. Family history is the most common risk factor for nonsmokers. Low-dose computed tomography screening is recommended once a year for patients with screening-detected abnormality and persistent exposure to risk factors. However, for high-risk heavy smokers and nonsmokers with risk factors, reassessment scans are recommended at an initial interval of 6 to 12 months with subsequent lengthening of reassessment intervals, and it should be stopped in patients more than 80 years of age or are unable or unwilling to undergo curative treatment. CONCLUSIONS: Asian countries face several challenges in implementing low-dose computed tomography screening, such as economic limitations, lack of efforts for early detection, and lack of specific government programs. Various strategies are suggested to overcome these challenges in Asia.

Intraventricular 4D flow cardiovascular magnetic resonance for assessing patients with heart failure with preserved ejection fraction: a pilot study.

Diagnosing heart failure with preserved ejection fraction (HFpEF) remains challenging. Intraventricular four-dimensional flow (4D flow) phase-contrast cardiovascular magnetic resonance (CMR) can assess different components of left ventricular (LV) flow including direct flow, delayed ejection, retained inflow and residual volume. This could be utilised to identify HFpEF. This study investigated if intraventricular 4D flow CMR could differentiate HFpEF patients from non-HFpEF and asymptomatic controls. Suspected HFpEF patients and asymptomatic controls were recruited prospectively. HFpEF patients were confirmed using European Society of Cardiology (ESC) 2021 expert recommendations. Non-HFpEF patients were diagnosed if suspected HFpEF patients did not fulfil ESC 2021 criteria. LV direct flow, delayed ejection, retained inflow and residual volume were obtained from 4D flow CMR images. Receiver operating characteristic (ROC) curves were plotted. 63 subjects (25 HFpEF patients, 22 non-HFpEF patients and 16 asymptomatic controls) were included in this study. 46% were male, mean age 69.8 ± 9.1 years. CMR 4D flow derived LV direct flow and residual volume could differentiate HFpEF vs combined group of non-HFpEF and asymptomatic controls (p < 0.001 for both) as well as HFpEF vs non-HFpEF patients (p = 0.021 and p = 0.005, respectively). Among the 4 parameters, direct flow had the largest area under curve (AUC) of 0.781 when comparing HFpEF vs combined group of non-HFpEF and asymptomatic controls, while residual volume had the largest AUC of 0.740 when comparing HFpEF and non-HFpEF patients. CMR 4D flow derived LV direct flow and residual volume show promise in differentiating HFpEF patients from non-HFpEF patients.

Diagnostic accuracy of cardiovascular magnetic resonance strain analysis and atrial size to identify heart failure with preserved ejection fraction.

Aims: Heart failure with preserved ejection fraction (HFpEF) continues to be a diagnostic challenge. Cardiac magnetic resonance atrial measurement, feature tracking (CMR-FT), tagging has long been suggested to diagnose HFpEF and potentially complement echocardiography especially when echocardiography is indeterminate. Data supporting the use of CMR atrial measurements, CMR-FT or tagging, are absent. Our aim is to conduct a prospective case-control study assessing the diagnostic accuracy of CMR atrial volume/area, CMR-FT, and tagging to diagnose HFpEF amongst patients suspected of having HFpEF. Methods and results: One hundred and twenty-one suspected HFpEF patients were prospectively recruited from four centres. Patients underwent echocardiography, CMR, and N-terminal pro-B-type natriuretic peptide (NT-proBNP) measurements within 24 h to diagnose HFpEF. Patients without HFpEF diagnosis underwent catheter pressure measurements or stress echocardiography to confirm HFpEF or non-HFpEF. Area under the curve (AUC) was determined by comparing HFpEF with non-HFpEF patients. Fifty-three HFpEF (median age 78 years, interquartile range 74-82 years) and thirty-eight non-HFpEF (median age 70 years, interquartile range 64-76 years) were recruited. Cardiac magnetic resonance left atrial (LA) reservoir strain (ResS), LA area index (LAAi), and LA volume index (LAVi) had the highest diagnostic accuracy (AUCs 0.803, 0.815, and 0.776, respectively). Left atrial ResS, LAAi, and LAVi had significantly better diagnostic accuracy than CMR-FT left ventricle (LV)/right ventricle (RV) parameters and tagging (P < 0.01). Tagging circumferential and radial strain had poor diagnostic accuracy (AUC 0.644 and 0.541, respectively). Conclusion: Cardiac magnetic resonance LA ResS, LAAi, and LAVi have the highest diagnostic accuracy to identify HFpEF patients from non-HFpEF patients amongst clinically suspected HFpEF patients. Cardiac magnetic resonance feature tracking LV/RV parameters and tagging had low diagnostic accuracy to diagnose HFpEF.

Image-spectral decomposition extended-learning assisted by sparsity for multi-energy computed tomography reconstruction.

Background: Multi-energy computed tomography (CT) provides multiple channel-wise reconstructed images, and they can be used for material identification and k-edge imaging. Nonetheless, the projection datasets are frequently corrupted by various noises (e.g., electronic, Poisson) in the acquisition process, resulting in lower signal-noise-ratio (SNR) measurements. Multi-energy CT images have local sparsity, nonlocal self-similarity in spatial dimension, and correlation in spectral dimension. Methods: In this paper, we propose an image-spectral decomposition extended-learning assisted by sparsity (IDEAS) method to fully exploit these intrinsic priors for multi-energy CT image reconstruction. Particularly, a nonlocal low-rank Tucker decomposition (TD) is employed to utilize the correlation and nonlocal self-similarity priors. Moreover, considering the advantages of multi-task tensor dictionary learning (TDL) in sparse representation, an adaptive spatial dictionary and an adaptive spectral dictionary are trained during the iterative reconstruction process. Furthermore, a weighted total variation (TV) regularization term is employed to encourage local sparsity. Results: Numerical simulation, physical phantom, and preclinical mouse experiments are performed to validate the proposed IDEAS algorithm. Specifically, in the simulation experiments, the proposed IDEAS reconstructed high-quality images that are very close to the references. For example, the root mean square error (RMSE) of IDEAS image in energy bin 1 is as low as 0.0672, while the RMSE of other methods are higher than 0.0843. Besides, the structural similarity (SSIM) of IDEAS reconstructed image in energy bin 1 is greater than 0.98. For material decomposition, the RMSE of IDEAS bone component is as low as 0.0152, and other methods are higher than 0.0199. In addition, the computational cost of IDEAS is as low as 98.8 s for one iteration, and the competing tensor decomposition method is higher than 327 s. Conclusions: To further improve the quality of the reconstructed multi-energy CT images, multiple prior regularizations are introduced to the multi-energy CT reconstructed model, leading to an IDEAS method. Both qualitative and quantitative evaluation of our results confirm the outstanding performance of the proposed algorithm compared to the state-of-the-arts.

Joint Cancer Segmentation and PI-RADS Classification on Multiparametric MRI Using MiniSegCaps Network.

MRI is the primary imaging approach for diagnosing prostate cancer. Prostate Imaging Reporting and Data System (PI-RADS) on multiparametric MRI (mpMRI) provides fundamental MRI interpretation guidelines but suffers from inter-reader variability. Deep learning networks show great promise in automatic lesion segmentation and classification, which help to ease the burden on radiologists and reduce inter-reader variability. In this study, we proposed a novel multi-branch network, MiniSegCaps, for prostate cancer segmentation and PI-RADS classification on mpMRI. MiniSeg branch outputted the segmentation in conjunction with PI-RADS prediction, guided by the attention map from the CapsuleNet. CapsuleNet branch exploited the relative spatial information of prostate cancer to anatomical structures, such as the zonal location of the lesion, which also reduced the sample size requirement in training due to its equivariance properties. In addition, a gated recurrent unit (GRU) is adopted to exploit spatial knowledge across slices, improving through-plane consistency. Based on the clinical reports, we established a prostate mpMRI database from 462 patients paired with radiologically estimated annotations. MiniSegCaps was trained and evaluated with fivefold cross-validation. On 93 testing cases, our model achieved a 0.712 dice coefficient on lesion segmentation, 89.18% accuracy, and 92.52% sensitivity on PI-RADS classification (PI-RADS ≥ 4) in patient-level evaluation, significantly outperforming existing methods. In addition, a graphical user interface (GUI) integrated into the clinical workflow can automatically produce diagnosis reports based on the results from MiniSegCaps.

Computed Tomography of Cartilage: An Exploration of Novel Cationic Bismuth Contrast Agent.

Accurate diagnosis of minor cartilage injuries with delayed contrast-enhanced computed tomography (CECT) is challenging as poor diffusion and toxicity issues limit the usage of common CT contrast agents. Hence, the design of safe contrast agents with physiochemical properties suitable for fast, deep cartilage imaging is imminent. Herein, a novel cationic bismuth contrast agent (Bi-DOTAPXD) based on dodecane tetraacetic acid (DOTA) was synthesized and examined for CECT of cartilage. The complex was designed to improve diagnosis by utilising a net-positive charge for enhanced permeability through cartilage, inherent low-toxicity and high X-ray attenuation of bismuth. Osteochondral plugs (n = 12), excised from visually intact porcine articular cartilage were immersed in Bi-DOTAPXD (8 mg/mL) and Gd-DOTAPXD (10 mg/mL) contrast agents and scanned with a high-resolution microcomputed tomography scanner at multiple time-points. The mean Bi-DOTAPXD and Gd-DOTAPXD partitions at 45-min time-point were 85.7 ± 35.1 and 69.8 ± 30.2%, and the partitions correlated with the histopathological analysis of cartilage proteoglycan (PG) content (r) at 0.657 and 0.632, respectively. The time diffusion constants (τ) for Bi-DOTAPXD and Gd-DOTA were 121 and 159 min, respectively. Diffusion Bi-DOTAPXD and Gd-DOTAPXD reflected inter-sample variation in cartilage PG content. Cationic Bi-DOTAPXD may have the potential as a CT agent for the diagnosis of cartilage.

Semisupervised white matter hyperintensities segmentation on MRI.

This study proposed a semisupervised loss function named level-set loss (LSLoss) for cerebral white matter hyperintensities (WMHs) segmentation on fluid-attenuated inversion recovery images. The training procedure did not require manually labeled WMH masks. Our image preprocessing steps included biased field correction, skull stripping, and white matter segmentation. With the proposed LSLoss, we trained a V-Net using the MRI images from both local and public databases. Local databases were the small vessel disease cohort (HKU-SVD, n = 360) and the multiple sclerosis cohort (HKU-MS, n = 20) from our institutional imaging center. Public databases were the Medical Image Computing Computer-assisted Intervention (MICCAI) WMH challenge database (MICCAI-WMH, n = 60) and the normal control cohort of the Alzheimer's Disease Neuroimaging Initiative database (ADNI-CN, n = 15). We achieved an overall dice similarity coefficient (DSC) of 0.81 on the HKU-SVD testing set (n = 20), DSC = 0.77 on the HKU-MS testing set (n = 5), and DSC = 0.78 on MICCAI-WMH testing set (n = 30). The segmentation results obtained by our semisupervised V-Net were comparable with the supervised methods and outperformed the unsupervised methods in the literature.

A Multimodal Deep Learning Approach to Predicting Systemic Diseases from Oral Conditions.

Background: It is known that oral diseases such as periodontal (gum) disease are closely linked to various systemic diseases and disorders. Deep learning advances have the potential to make major contributions to healthcare, particularly in the domains that rely on medical imaging. Incorporating non-imaging information based on clinical and laboratory data may allow clinicians to make more comprehensive and accurate decisions. Methods: Here, we developed a multimodal deep learning method to predict systemic diseases and disorders from oral health conditions. A dual-loss autoencoder was used in the first phase to extract periodontal disease-related features from 1188 panoramic radiographs. Then, in the second phase, we fused the image features with the demographic data and clinical information taken from electronic health records (EHR) to predict systemic diseases. We used receiver operation characteristics (ROC) and accuracy to evaluate our model. The model was further validated by an unseen test dataset. Findings: According to our findings, the top three most accurately predicted chapters, in order, are the Chapters III, VI and IX. The results indicated that the proposed model could predict systemic diseases belonging to Chapters III, VI and IX, with AUC values of 0.92 (95% CI, 0.90-94), 0.87 (95% CI, 0.84-89) and 0.78 (95% CI, 0.75-81), respectively. To assess the robustness of the models, we performed the evaluation on the unseen test dataset for these chapters and the results showed an accuracy of 0.88, 0.82 and 0.72 for Chapters III, VI and IX, respectively. Interpretation: The present study shows that the combination of panoramic radiograph and clinical oral features could be considered to train a fusion deep learning model for predicting systemic diseases and disorders.

Spectral CT reconstruction via Spectral-Image Tensor and Bidirectional Image-gradient minimization.

It is challenging to obtain good image quality in spectral computed tomography (CT) as the photon-number for the photon-counting detectors is limited for each narrow energy bin. This results in a lower signal to noise ratio (SNR) for the projections. To handle this issue, we first formulate the weight bidirectional image gradient with L0-norm constraint of spectral CT image. Then, as a new regularizer, bidirectional image gradient with L0-norm constraint is introduced into the tensor decomposition model, generating the Spectral-Image Tensor and Bidirectional Image-gradient Minimization (SITBIM) algorithm. Finally, the split-Bregman method is employed to optimize the proposed SITBIM mathematical model. The experiments on the numerical mouse phantom and real mouse experiments are designed to validate and evaluate the SITBIM method. The results demonstrate that the SITBIM can outperform other state-of-the-art methods (including TVM, TV + LR, SSCMF and NLCTF). INDEX TERMS: -spectral CT, image reconstruction, tensor decomposition, unidirectional image gradient, image similarity.

Rethinking Annotation Granularity for Overcoming Shortcuts in Deep Learning-based Radiograph Diagnosis: A Multicenter Study.

Purpose: To evaluate the ability of fine-grained annotations to overcome shortcut learning in deep learning (DL)-based diagnosis using chest radiographs. Materials and Methods: Two DL models were developed using radiograph-level annotations (disease present: yes or no) and fine-grained lesion-level annotations (lesion bounding boxes), respectively named CheXNet and CheXDet. A total of 34 501 chest radiographs obtained from January 2005 to September 2019 were retrospectively collected and annotated regarding cardiomegaly, pleural effusion, mass, nodule, pneumonia, pneumothorax, tuberculosis, fracture, and aortic calcification. The internal classification performance and lesion localization performance of the models were compared on a testing set (n = 2922); external classification performance was compared on National Institutes of Health (NIH) Google (n = 4376) and PadChest (n = 24 536) datasets; and external lesion localization performance was compared on the NIH ChestX-ray14 dataset (n = 880). The models were also compared with radiologist performance on a subset of the internal testing set (n = 496). Performance was evaluated using receiver operating characteristic (ROC) curve analysis. Results: Given sufficient training data, both models performed similarly to radiologists. CheXDet achieved significant improvement for external classification, such as classifying fracture on NIH Google (CheXDet area under the ROC curve [AUC], 0.67; CheXNet AUC, 0.51; P < .001) and PadChest (CheXDet AUC, 0.78; CheXNet AUC, 0.55; P < .001). CheXDet achieved higher lesion detection performance than CheXNet for most abnormalities on all datasets, such as detecting pneumothorax on the internal set (CheXDet jackknife alternative free-response ROC [JAFROC] figure of merit [FOM], 0.87; CheXNet JAFROC FOM, 0.13; P < .001) and NIH ChestX-ray14 (CheXDet JAFROC FOM, 0.55; CheXNet JAFROC FOM, 0.04; P < .001). Conclusion: Fine-grained annotations overcame shortcut learning and enabled DL models to identify correct lesion patterns, improving the generalizability of the models.Keywords: Computer-aided Diagnosis, Conventional Radiography, Convolutional Neural Network (CNN), Deep Learning Algorithms, Machine Learning Algorithms, Localization Supplemental material is available for this article © RSNA, 2022.

Early stage NSCLS patients' prognostic prediction with multi-information using transformer and graph neural network model.

Background: We proposed a population graph with Transformer-generated and clinical features for the purpose of predicting overall survival (OS) and recurrence-free survival (RFS) for patients with early stage non-small cell lung carcinomas and to compare this model with traditional models. Methods: The study included 1705 patients with lung cancer (stages I and II), and a public data set for external validation (n=127). We proposed a graph with edges representing non-imaging patient characteristics and nodes representing imaging tumour region characteristics generated by a pretrained Vision Transformer. The model was compared with a TNM model and a ResNet-Graph model. To evaluate the models' performance, the area under the receiver operator characteristic curve (ROC-AUC) was calculated for both OS and RFS prediction. The Kaplan-Meier method was used to generate prognostic and survival estimates for low- and high-risk groups, along with net reclassification improvement (NRI), integrated discrimination improvement (IDI), and decision curve analysis. An additional subanalysis was conducted to examine the relationship between clinical data and imaging features associated with risk prediction. Results: Our model achieved AUC values of 0.785 (95% confidence interval [CI]: 0.716-0.855) and 0.695 (95% CI: 0.603-0.787) on the testing and external data sets for OS prediction, and 0.726 (95% CI: 0.653-0.800) and 0.700 (95% CI: 0.615-0.785) for RFS prediction. Additional survival analyses indicated that our model outperformed the present TNM and ResNet-Graph models in terms of net benefit for survival prediction. Conclusions: Our Transformer-Graph model was effective at predicting survival in patients with early stage lung cancer, which was constructed using both imaging and non-imaging clinical features. Some high-risk patients were distinguishable by using a similarity score function defined by non-imaging characteristics such as age, gender, histology type, and tumour location, while Transformer-generated features demonstrated additional benefits for patients whose non-imaging characteristics were non-discriminatory for survival outcomes. Funding: The study was supported by the National Natural Science Foundation of China (91959126, 8210071009), and Science and Technology Commission of Shanghai Municipality (20XD1403000, 21YF1438200).