Performance of postmortem CT in the diagnosis of natural death from out-of-hospital cardiac arrest.

PURPOSE: Postmortem CT (PMCT) is used widely to identify the cause of death. However, its diagnostic performance in cases of natural death from out-of-hospital cardiac arrest (OHCA) may be unsatisfactory because the cause tends to be cardiogenic and cannot be detected on PMCT images. We retrospectively investigated the diagnostic performance of PMCT in the diagnosis of natural death from OHCA and compared it to that of unnatural death. MATERIALS AND METHODS: Our series included 450 cases; 336 were natural- and 114 were unnatural death cases. Between 2018 and 2022 all underwent non-contrast PMCT to identify the cause of death. Two radiologists reviewed the PMCT images and categorized them as diagnostic (PMCT alone sufficient to determine the cause of death), suggestive (the cause of death was suggested but additional information was needed), and non-diagnostic (the cause of death could not be determined on PMCT images). The diagnostic performance of PMCT was defined by the percentage of diagnosable and suggestive cases and compared between natural- and unnatural death cases. Interobserver agreement for the cause of death on PMCT images was also assessed with the Cohen kappa coefficient of concordance. RESULTS: The diagnostic performance of PMCT for the cause of natural- and unnatural deaths from OHCA was 30.3% and 66.6%, respectively (p < 0.01). The interobserver agreement for the cause of natural- and unnatural deaths on PMCT images was very good with kappa value 0.92 and 0.96, respectively. CONCLUSION: As PMCT identified the cause of natural death by OHCA in only 30% of cases, its diagnostic performance must be improved.

Development of a deep-learning algorithm for age estimation on CT images of the vertebral column.

PURPOSE: The accurate age estimation of cadavers is essential for their identification. However, conventional methods fail to yield adequate age estimation especially in elderly cadavers. We developed a deep learning algorithm for age estimation on CT images of the vertebral column and checked its accuracy. METHOD: For the development of our deep learning algorithm, we included 1,120 CT data of the vertebral column of 140 patients for each of 8 age decades. The deep learning model of regression analysis based on Visual Geometry Group-16 (VGG16) was improved in its estimation accuracy by bagging. To verify its accuracy, we applied our deep learning algorithm to estimate the age of 219 cadavers who had undergone postmortem CT (PMCT). The mean difference and the mean absolute error (MAE), the standard error of the estimate (SEE) between the known- and the estimated age, were calculated. Correlation analysis using the intraclass correlation coefficient (ICC) and Bland-Altman analysis were performed to assess differences between the known- and the estimated age. RESULTS: For the 219 cadavers, the mean difference between the known- and the estimated age was 0.30 years; it was 4.36 years for the MAE, and 5.48 years for the SEE. The ICC (2,1) was 0.96 (95 % confidence interval: 0.95-0.97, p < 0.001). Bland-Altman analysis showed that there were no proportional or fixed errors (p = 0.08 and 0.41). CONCLUSIONS: Our deep learning algorithm for estimating the age of 219 cadavers on CT images of the vertebral column was more accurate than conventional methods and highly useful.

Iodine maps derived from sparse-view kV-switching dual-energy CT equipped with a deep learning reconstruction for diagnosis of hepatocellular carcinoma.

Deep learning-based spectral CT imaging (DL-SCTI) is a novel type of fast kilovolt-switching dual-energy CT equipped with a cascaded deep-learning reconstruction which completes the views missing in the sinogram space and improves the image quality in the image space because it uses deep convolutional neural networks trained on fully sampled dual-energy data acquired via dual kV rotations. We investigated the clinical utility of iodine maps generated from DL-SCTI scans for assessing hepatocellular carcinoma (HCC). In the clinical study, dynamic DL-SCTI scans (tube voltage 135 and 80 kV) were acquired in 52 patients with hypervascular HCCs whose vascularity was confirmed by CT during hepatic arteriography. Virtual monochromatic 70 keV images served as the reference images. Iodine maps were reconstructed using three-material decomposition (fat, healthy liver tissue, iodine). A radiologist calculated the contrast-to-noise ratio (CNR) during the hepatic arterial phase (CNRa) and the equilibrium phase (CNRe). In the phantom study, DL-SCTI scans (tube voltage 135 and 80 kV) were acquired to assess the accuracy of iodine maps; the iodine concentration was known. The CNRa was significantly higher on the iodine maps than on 70 keV images (p < 0.01). The CNRe was significantly higher on 70 keV images than on iodine maps (p < 0.01). The estimated iodine concentration derived from DL-SCTI scans in the phantom study was highly correlated with the known iodine concentration. It was underestimated in small-diameter modules and in large-diameter modules with an iodine concentration of less than 2.0 mgI/ml. Iodine maps generated from DL-SCTI scans can improve the CNR for HCCs during hepatic arterial phase but not during equilibrium phase in comparison with virtual monochromatic 70 keV images. Also, when the lesion is small or the iodine concentration is low, iodine quantification may result in underestimation.

Evaluation of the second-generation whole-heart motion correction algorithm (SSF2) used to demonstrate the aortic annulus on cardiac CT.

The main purpose of pre-transcatheter aortic valve implantation (TAVI) cardiac computed tomography (CT) for patients with severe aortic stenosis is aortic annulus measurements. However, motion artifacts present a technical challenge because they can reduce the measurement accuracy of the aortic annulus. Therefore, we applied the recently developed second-generation whole-heart motion correction algorithm (SnapShot Freeze 2.0, SSF2) to pre-TAVI cardiac CT and investigated its clinical utility by stratified analysis of the patient's heart rate during scanning. We found that SSF2 reconstruction significantly reduced aortic annulus motion artifacts and improved the image quality and measurement accuracy compared to standard reconstruction, especially in patients with high heart rate or a 40% R-R interval (systolic phase). SSF2 may contribute to improving the measurement accuracy of the aortic annulus.

Improvement of Spatial Resolution on Coronary CT Angiography by Using Super-Resolution Deep Learning Reconstruction.

RATIONALE AND OBJECTIVES: Our objective was to compare the image quality of coronary CT angiography reconstructed with super-resolution deep learning reconstruction (SR-DLR) and with hybrid iterative reconstruction (IR) images. MATERIALS AND METHODS: This retrospective study included 100 patients who underwent coronary CT angiography using a 320-detector-row CT scanner. The CT images were reconstructed with hybrid IR and SR-DLR. The standard deviation of the CT number was recorded and the CT attenuation profile across the left main coronary artery was generated to calculate the contrast-to-noise ratio (CNR) and measure the edge rise slope (ERS). Overall image quality was evaluated and plaque detectability was assessed on a 4-point scale (1 = poor, 4 = excellent). For reference, invasive coronary angiography of 14 patients was used. RESULTS: The mean image noise on SR-DLR was significantly lower than on hybrid IR images (15.6 vs 22.9 HU; p < 0.01). The mean CNR was significantly higher and the ERS was steeper on SR-DLR- compared to hybrid IR images (CNR: 32.4 vs 20.4, p < 0.01; ERS: 300.0 vs 198.2 HU/mm, p < 0.01). The image quality score was better on SR-DLR- than on hybrid IR images (3.6 vs 3.1; p < 0.01). SR-DLR increased the detectability of plaques with < 50% stenosis (p < 0.01). CONCLUSION: SR-DLR was superior to hybrid IR with respect to the image noise, the sharpness of coronary artery margins, and plaque detectability.

An introduction to photon-counting detector CT (PCD CT) for radiologists.

The basic performance of photon-counting detector computed tomography (PCD CT) is superior to conventional CT (energy-integrating detector CT: EID CT) because its spatial- and contrast resolution of soft tissues is higher, and artifacts are reduced. Because the X-ray photon energy separation is better with PCD CT than conventional EID-based dual-energy CT, it has the potential to improve virtual monochromatic- and virtual non-contrast images, material decomposition including quantification of the iodine distribution, and K-edge imaging. Therefore, its clinical applicability may be increased. Although the image quality of PCD CT scans is superior to that of EID CT currently, further improvement may be possible. The introduction of iterative image reconstruction and reconstruction with deep convolutional neural networks will be useful.

Global illumination rendering versus volume rendering for the forensic evaluation of stab wounds using computed tomography.

We compared three-dimensional (3D) CT images of stabbing victims subjected to volume-rendering (VR) or global illumination-rendering (GIR), a new technique now available for the reconstruction of 3D CT images. It simulates the complete interactions of photons with the scanned object, thereby providing photorealistic images. The diagnostic value of the images was also compared with that of macroscopic photographs. We used postmortem 3D CT images of 14 stabbing victims who had undergone autopsy and CT studies. The 3D CT images were subjected to GIR or VR and the 3D effect and the smoothness of the skin surface were graded on a 5-point scale. We also compared the 3D CT images of 37 stab wounds with macroscopic photographs. The maximum diameter of the wounds was measured on VR and GIR images and compared with the diameter recorded at autopsy. The overall image-quality scores and the ability to assess the stab wounds were significantly better on GIR than VR images (median scores: VR = 3 vs GIR = 4, p < 0.01). The mean difference between the wound diameter measured on VR and GIR images and at autopsy were both 0.2 cm, respectively. For the assessment of stab wounds, 3D CT images subjected to GIR were superior to VR images. The diagnostic value of 3D CT GIR image was comparable to that of macroscopic photographs.

Computerized Classification of Pneumoconiosis on Digital Chest Radiography Artificial Neural Network with Three Stages.

It is difficult for radiologists to classify pneumoconiosis from category 0 to category 3 on chest radiographs. Therefore, we have developed a computer-aided diagnosis (CAD) system based on a three-stage artificial neural network (ANN) method for classification based on four texture features. The image database consists of 36 chest radiographs classified as category 0 to category 3. Regions of interest (ROIs) with a matrix size of 32 × 32 were selected from chest radiographs. We obtained a gray-level histogram, histogram of gray-level difference, gray-level run-length matrix (GLRLM) feature image, and gray-level co-occurrence matrix (GLCOM) feature image in each ROI. For ROI-based classification, the first ANN was trained with each texture feature. Next, the second ANN was trained with output patterns obtained from the first ANN. Finally, we obtained a case-based classification for distinguishing among four categories with the third ANN method. We determined the performance of the third ANN by receiver operating characteristic (ROC) analysis. The areas under the ROC curve (AUC) of the highest category (severe pneumoconiosis) case and the lowest category (early pneumoconiosis) case were 0.89 ± 0.09 and 0.84 ± 0.12, respectively. The three-stage ANN with four texture features showed the highest performance for classification among the four categories. Our CAD system would be useful for assisting radiologists in classification of pneumoconiosis from category 0 to category 3.

[Development of a computer-aided diagnostic system for detecting multiple sclerosis using magnetic resonance images].

It is of key importance to be able to evaluate the temporal changes seen in multiple sclerosis (MS) lesions in terms of location, shape, and area for estimating MS progression. The purpose of our study was to develop an automated method for detecting potential MS regions based on three types of brain magnetic resonance (MR) images: T1- and T2-weighted images, and fluid attenuated inversion-recovery (FLAIR) images. The brain regions were segmented based on a tri-linear interpolation technique and k-mean clustering technique. True positive regions and false positive regions were classified from three types of MR images using a support vector machine (SVM). We applied our proposed method to 60 slices of 20 MS cases. As a result, the sensitivity for detection of MS regions was 81.8%, with 14.1% false positives per true positive. This method should prove useful for the diagnosis of multiple sclerosis.

Development of CAD based on ANN analysis of power spectra for pneumoconiosis in chest radiographs: effect of three new enhancement methods.

We have been developing a computer-aided detection (CAD) scheme for pneumoconiosis based on a rule-based plus artificial neural network (ANN) analysis of power spectra. In this study, we have developed three enhancement methods for the abnormal patterns to reduce false-positive and false-negative values. The image database consisted of 2 normal and 15 abnormal chest radiographs. The International Labour Organization standard chest radiographs with pneumoconiosis were categorized as subcategory, size, and shape of pneumoconiosis. Regions of interest (ROIs) with a matrix size of 32 × 32 were selected from normal and abnormal lungs. Three new enhanced methods were obtained by window function, top-hat transformation, and gray-level co-occurrence matrix analysis. We calculated the power spectrum (PS) of all ROIs by Fourier transform. For the classification between normal and abnormal ROIs, we applied a combined analysis using the ruled-based plus the ANN method. To evaluate the overall performance of this CAD scheme, we employed ROC analysis for distinguishing between normal and abnormal ROIs. On the chest radiographs of the highest categories (severe pneumoconiosis) and the lowest categories (early pneumoconiosis), this CAD scheme achieved area under the curve (AUC) values of 0.93 ± 0.02 and 0.72 ± 0.03. The combined rule-based plus ANN method with the three new enhanced methods obtained the highest classification performance for distinguishing between abnormal and normal ROIs. Our CAD system based on the three new enhanced methods would be useful in assisting radiologists in the classification of pneumoconiosis.

[Computerized classification of pneumoconiosis radiographs based on grey level co-occurrence matrices].

Pneumoconiosis is diagnosed as categories 0-4 according to the Pneumoconiosis Law. Physicians have difficulty precisely categorizing many chest images. Therefore, we have developed a computerized method for automatically categorizing pneumoconiosis from chest radiographs. First, we extracted the rib edge regions from lung ROIs. Second, texture features were extracted using a dot enhancement filter, line enhancement filter, and grey level co-occurrence matrix. Third, the rib edge regions were removed from these processed images. Finally, we used a support vector machine for feature analysis. In a consistency test, 56 cases (69.7%) were classified correctly, and 45 cases (61.8%) were classified correctly in a validation test. These results show that the proposed features and removal of the rib edge are effective in classifying the profusion of opacities that indicate pneumoconiosis.

Computerized analysis of pneumoconiosis in digital chest radiography: effect of artificial neural network trained with power spectra.

It is difficult for radiologists to classify pneumoconiosis with small nodules on chest radiographs. Therefore, we have developed a computer-aided diagnosis (CAD) system based on the rule-based plus artificial neural network (ANN) method for distinction between normal and abnormal regions of interest (ROIs) selected from chest radiographs with and without pneumoconiosis. The image database consists of 11 normal and 12 abnormal chest radiographs. These abnormal cases included five silicoses, four asbestoses, and three other pneumoconioses. ROIs (matrix size, 32 × 32) were selected from normal and abnormal lungs. We obtained power spectra (PS) by Fourier transform for the frequency analysis. A rule-based method using PS values at 0.179 and 0.357 cycles per millimeter, corresponding to the spatial frequencies of nodular patterns, were employed for identification of obviously normal or obviously abnormal ROIs. Then, ANN was applied for classification of the remaining normal and abnormal ROIs, which were not classified as obviously abnormal or normal by the rule-based method. The classification performance was evaluated by the area under the receiver operating characteristic curve (Az value). The Az value was 0.972 ± 0.012 for the rule-based plus ANN method, which was larger than that of 0.961 ± 0.016 for the ANN method alone (P ≤ 0.15) and that of 0.873 for the rule-based method alone. We have developed a rule-based plus pattern recognition technique based on the ANN for classification of pneumoconiosis on chest radiography. Our CAD system based on PS would be useful to assist radiologists in the classification of pneumoconiosis.

[Relationship between image quality and signal detectability in CR and FPD systems].

With the screen/film X-rays imaging system, Wakamatsu et al. reported that there was a close relationship between the square root of spectral signal-to-noise ratio area and sensitivity measure d' in receiver operating characteristic (ROC) analysis. In this study, we investigated the relationship between image quality and signal detectability in two digital X-ray imaging systems using computed radiography (CR) and a flat panel detector (FPD). We used urethane resin balls with a diameter of 2 mm as a signal for cases samples in ROC analysis. In this experiment, the square root of the spectral signal-to-noise ratio area was closely related to d' in ROC analysis in both digital X-ray imaging systems. In addition, when the exposure dose increased, signal detectability improved, but then saturated at one level. These results suggest that the exposure dose can be reduced when the optimal dose setting can be determined.

[Development of comprehensive image processing technique for differential diagnosis of liver disease by using multi-modality images-pixel-based cross-correlation method using a profile-].

Imaging techniques such as high magnetic field imaging and multidetector-row CT have been markedly improved recently. The final image-reading systems easily produce more than a thousand diagnostic images per patient. Therefore, we developed a comprehensive cross-correlation processing technique using multi-modality images, in order to decrease the considerable time and effort involved in the interpretation of a radiogram (multi-formatted display and/or stack display method, etc). In this scheme, the criteria of an attending radiologist for the differential diagnosis of liver cyst, hemangioma of liver, hepatocellular carcinoma, and metastatic liver cancer on magnetic resonance images with various sequences and CT images with and without contrast enhancement employ a cross-correlation coefficient. Using a one-dimensional cross-correlation method, comprehensive image processing could be also adapted for various artifacts (some depending on modality imaging, and some on patients), which may be encountered at the clinical scene. This comprehensive image-processing technique could assist radiologists in the differential diagnosis of liver diseases.