Yongsu Yoon

Estimation of body parameters, height and body weight, is important for biological profiling in forensic investigations. This study aimed to develop a deep learning model for automated height and weight estimation using postmortem computed tomography (PMCT) scout view images acquired in forensic practice. A retrospective analysis was conducted in 3,124 cases that underwent autopsy and PMCT. Scout view images were preprocessed with geometric correction based on PMCT table height. Three CT window settings were fused into RGB channels, enhancing anatomical and environmental information. A convolutional neural network based on a pre-trained VGG16 architecture was trained to simultaneously estimate height and weight. Model performance was evaluated on a test dataset (n = 624) using mean absolute error (MAE), mean absolute percentage error (MAPE), correlation analysis, and Bland-Altman analysis. The multi-window RGB fusion model achieved MAE values of 2.39 cm for height and 1.83 kg for weight, outperforming the single-window model. Predicted and reference values were strongly correlated. Bland-Altman analysis demonstrated narrow limits of agreement. Deep learning applied to PMCT scout view images enables automatic estimation of height and weight with high accuracy. The proposed multi-window RGB fusion with geometric correction offers a practical tool for body parameter estimation in forensic investigations.

Sphenoid sinus fluid is considered a supportive indicator of drowning in forensic medicine, but traditional manual assessment on postmortem computed tomography (PMCT) is labor-intensive and observer-dependent. Efficient, reproducible methods for quantitative evaluation are needed in forensic practice. This study developed deep learning-based approaches for the automated segmentation and volumetric estimation of sphenoid sinus fluid using PMCT images from 165 autopsy-confirmed drowning cases. Three U-Net-based models (2D, 2.5D, and 3D) were developed and evaluated against manually annotated reference standards. In the test dataset, mean Dice coefficients were 0.866 (2D), 0.869 (2.5D), and 0.798 (3D). Volumetric estimates showed no statistically significant differences from the reference standard, with strong correlations (Spearman's ρ = 0.976-0.988). Mean absolute errors were 0.218 (2D), 0.206 (2.5D), and 0.310 ml (3D). The 2.5D approach provided the most balanced performance between segmentation accuracy and volumetric estimation. These findings demonstrate the feasibility of automated PMCT-based segmentation and volumetric quantification of sphenoid sinus fluid, enabling quantitative assessment on PMCT images prior to autopsy.

The global increase in computed tomography (CT) use, highlighted by a 40% growth in South Korea over the past decade, has made CT a significant source of medical radiation exposure, emphasizing the need for accurate effective dose (ED) estimation. This study aimed to develop population-specific effective dose conversion factors (k-factors) for brain CT examinations across the range of tube voltages used in Korean hospitals. Clinical dose parameters were obtained from the Korean National CT Dose Index Registry (KNCTDIR), which compiles large-scale dose-length product (DLP) data from 45 hospitals nationwide. The mean, maximum, and minimum kVp and DLP values were selected to represent typical clinical variations. Monte Carlo simulations were performed using GATE version 10.0b8 with Korean-sized XCAT phantoms for adult and pediatric groups. Organ and effective doses were calculated following ICRP 103 tissue-weighting factors, and k-factors were derived for each age, sex, and voltage condition. The results showed consistent k-factors across the evaluated voltage range, with only minimal sex-related differences. Infants had the highest coefficients (0.0029 mSv/mGy·cm), while pediatric k-factors were lower and remained relatively stable from ages 2 to 15 years. Comparisons with previous Korean and international studies revealed notable quantitative differences, emphasizing the need for updated, population-specific coefficients. The revised k-factors facilitate practical and consistent effective-dose estimation in Korean brain CT procedures.

Background: Iodinated contrast agents are widely used in computed tomography (CT) imaging; however, they can cause adverse drug reactions (ADRs) ranging from mild hypersensitivity to severe anaphylaxis. While several clinical risk factors have been identified, large–scale studies incorporating environmental variables remain limited. This study aimed to assess the prevalence and predictors of contrast agent-related ADRs over a 10-year period. Methods: We retrospectively analyzed 221,962 adult outpatients who underwent contrast-enhanced CT between January 2014 and December 2023 at a single tertiary center: Patient characteristics, clinical conditions (e.g., hypertension, allergy history), contrast agent types, premedication status, seasonal trends, temperature, and humidity were examined. ADRs were categorized as mild, moderate, or severe based on American College of Radiology (ACR) guidelines. Logistic regression was used to identify independent predictors. Results: The overall prevalence of ADRs was 0.64% (1423 cases). ADRs were more frequent in females, younger patients, and those receiving premedication. Seasonal and environmental patterns were evident: higher ADR rates occurred in summer and autumn, with positive correlations to ambient temperature and humidity. Among contrast agents, Ioversol (1.4%) and Iomeprol (1.2%) showed the highest ADR rates. The prevalence of mild ADRs increased in the post–COVID-19 period, while that of moderate reactions declined. Conclusions: This real–world study identified multiple clinical and environmental factors associated with ADRs to iodinated contrast agents in CT imaging. The findings suggest the importance of individualized risk assessment and the consideration of environmental factors when planning contrast administration.

In digital radiography (DR) systems, the sensitivity index, also known as the exposure indicator, exhibits unique behavior, depending on the manufacturer. The exposure index (EI) was introduced by the International Electrotechnical Commission to standardize the sensitivity indices of DR systems produced by different manufacturers for general radiography. This EI value is directly proportional to the dose incident on the imaging detector, providing valuable insight into exposure dose levels and radiographic noise. Recently, technological advancements have enabled some DR systems to display the EI value on the console immediately after exposure. This study investigated the characteristics of the displayed EI value and radiographic noise for an indirect flat-panel detector across four X-ray beam qualities based on added aluminum filters: RQA3, 5, 7, and 9. The displayed EI values for all the X-ray beam qualities proportionally increased to the dose incident on the detector. However, the displayed EI values for RQA3 deviated from those for RQA5, 7, and 9 at the same doses incident on the imaging detector. Furthermore, the Wiener spectrum (WS) was used to evaluate radiographic noise. For similar EI values, the WS of RQA3, which had the highest dose incident on the imaging detector, was slightly lower than those of other X-ray beam qualities. Our findings suggest that the relationship between the displayed EI and radiographic noise varies with the X-ray beam quality.

Prompt personal identification is required during disasters that can result in many casualties. To rapidly estimate sex based on skull structure, this study applied deep learning using two-dimensional silhouette images, obtained from head postmortem computed tomography (PMCT), to enhance the outline shape of the skull. We investigated the process of sex estimation using silhouette images viewed from different angles and majority votes. A total of 264 PMCT cases (132 cases for each sex) were used for transfer learning with two deep-learning models (AlexNet and VGG16). VGG16 exhibited the highest accuracy (89.8%) for lateral projections. The accuracy improved to 91.7% when implementing a majority vote based on the results of multiple projection angles. Moreover, silhouette images can be obtained from simple and popular X-ray imaging in addition to PMCT. Thus, this study demonstrated the feasibility of sex estimation by combining silhouette images with deep learning. The results implied that X-ray images can be used for personal identification.

Postmortem computed tomography (PMCT) is used in forensic medicine worldwide due to its ability to non-invasively visualize injuries, hemorrhage, and estimate volume. In the autopsy of infants, assessing nutritional conditions such as weight is crucial for identifying neglect. This study aims to evaluate the usefulness of retrospectively estimating the weight of Korean infants using PMCT-based volume and multiplication factors, even when the body has been cremated. A total of 44 cases of infant death (under 12 months) were analyzed. PMCT images were obtained before autopsy. Autopsy records and documentation provided by the police at the time of autopsy were reviewed to determine the weight (g) of the infant. PMCT-based infant volumes (mL) were estimated using a three-dimensional semi-automatic segmentation method. Multiplication factors (g/mL) were calculated by dividing the weight recorded at autopsy by the PMCT-based volume, yielding a mean of 1.047 g/mL, ranging from 1.014 g/mL to 1.085 g/mL. The mean absolute error compared to weights recorded at autopsy was 95 g. Significant discrepancies were observed between weights recorded at the scene or medical center and those measured at autopsy. This study demonstrates that PMCT-based weight estimation for Korean infants is a reliable method and has the potential for retrospectively validating incorrect weight measurements and addressing inconsistencies in recorded weight data.

Post-mortem (PM) imaging has potential for identifying individuals by comparing ante-mortem (AM) and PM images. Radiographic images of bones contain significant information for personal identification. However, PM images are affected by soft tissue decomposition; therefore, it is desirable to extract only images of bones that change little over time. This study evaluated the effectiveness of U-Net for bone image extraction from two-dimensional (2D) X-ray images. Two types of pseudo 2D X-ray images were created from the PM computed tomography (CT) volumetric data using ray-summation processing for training U-Net. One was a projection of all body tissues, and the other was a projection of only bones. The performance of the U-Net for bone extraction was evaluated using Intersection over Union, Dice coefficient, and the area under the receiver operating characteristic curve. Additionally, AM chest radiographs were used to evaluate its performance with real 2D images. Our results indicated that bones could be extracted visually and accurately from both AM and PM images using U-Net. The extracted bone images could provide useful information for personal identification in forensic pathology.