Comparative radiologic identification with standardized single CT images of the paranasal sinuses-Evaluation of inter-rater reliability

Fully automated radiologic identification focusing on the sternal bone

A crucial task in forensic investigations is the identification of unknown deceased. In general, secure identification methods rely on a comparison of ante mortem (AM) with post mortem (PM) data. However, available morphologic approaches are often dependent on the expertise and experience of the examiner, and often lack standardisation and statistical evidence. The objective of this study was therefore to overcome the current challenges via developing a fully automated radiologic identification (autoRADid) method based on the sternal bone. An anonymised AM data set consisting of 91 chest computed tomography (CT) scans, as well as an anonymised PM data set of 42 chest CT scans were included in this work. Out of the 91 available AM CT data sets, 42 AM scans corresponded to the 42 PM CT scans. For the fully automated identification analysis, a custom-made python pipeline was developed, which automatically registers AM data to the PM data in question using a two-step registration method. To evaluate the registration procedure and subsequent identification success, the image similarity measures Jaccard Coefficient, Dice Coefficient, and Mutual Information were computed. The highest value for each metric was retrieved in order to analyse the correspondence between AM and PM data. For all three similarity measures, 38 out of the 42 cases were matched correctly. This corresponds to an accuracy of 91.2%. The four unsuccessful cases incorporated surgical interventions taking place between the AM and the PM CT acquisition or poor CT scan quality preventing robust registration results. To conclude, the presented autoRADid method seems to be a promising fully automated tool for a reliable and facile identification of unknown deceased. A final pipeline combining all three similarity measures is open source and publicly available for efficient future identifications of unknown deceased.

Implementation of a personal identification system using alveolar bone images

OBJECTIVE

In recent years, personal identification has been performed using antemortem panoramic X-ray images and postmortem-CT images. Using these, we have developed a personal identification method that focuses on the alveolar bone. This study examined the effectiveness of this method and aimed to implement a reproducible system.

MATERIALS AND METHODS

For personal identification, a total of 633 CT images and panoramic X-ray images belonging to three groups with different conditions were used.　These images were 160 sets in the same person group and 96,820 in the other groups. The similarity of alveolar bone images was calculated using the landmark method of Procrustes analysis. The processes were system implemented and the methodology was validated.

RESULTS

The ability to identify between the same person group and other person groups showed 0.9769 as the area under the curve (AUC: ROC curve). At the cutoff value of 4.978, there was no false rejection rate, but false acceptance rate was slightly higher.

CONCLUSION

This method was useful as a screening method for personal identification. In addition, system implementation was efficient and reduced human error. In the future, we aim to realize a more efficient personal identification method using distortion-corrected images and including auto-detective landmarks using deep learning.

Frontal Sinus Morphological and Dimensional Variation as Seen on Computed Tomography Scans

Frontal sinus variation has been used in forensic anthropology to aid in positive identification since the 1920s. As imaging technology has evolved, so has the quality and quantity of data that practitioners can collect. This study examined frontal sinus morphological and dimensional variation on computed tomography (CT) scans in 325 individuals for assigned sex females and males from African-, Asian-, European-, and Latin American-derived groups. Full coronal sinus outlines from medically derived CT images were transferred into SHAPE v1.3 for elliptical Fourier analysis (EFA). The dimensional data were measured directly from the images using the MicroDicom viewer. Statistical analyses—Pearson’s chi-square, ANOVA, and Tukey post hoc tests—were run in R Studio. Results indicated that 3.7% lacked a frontal sinus and 12.0% had a unilateral sinus, usually on the left (74.3%). Additionally, no statistically significant morphological clustering using EFA was found based on assigned sex and/or population affinity. However, there were statistically significant differences dimensionally (height and depth) when tested against assigned sex and population affinity, indicating that the interactive effects of sexual dimorphism and adaptive population histories influence the dimensions but not the shape of the frontal sinus.

Validation of a post-mortem computed tomography method for age estimation based on the 4th rib in a French population

Age estimation is a key factor for identification procedure in forensic context. Based on anthropological findings, degenerative changes of the sternal extremity of the 4th rib are currently used for age estimation. These have been adapted to post-mortem computed tomography (PMCT). The aim of this study was to validate a post-mortem computed tomography method based on a revision of the Iscan's method on a French sample. A total of 250 PMCT (aged from 18-98 years (IQR 36-68 years, median 51 years); 68 (27%) females) from the Medicolegal Institute of Paris (MLIP) were analyzed by two radiologists. The sternal extremity of 4th right rib was scored using method adapted from Iscan et al. Weighted κ was used to evaluate intra- and inter-observer reliability and Spearman correlation was performed to evaluate relationship between age and score. Confidence intervals for individual prediction of age based on 4th rib score and sex were computed with bootstrapping. The intra-observer reliability and inter-observer reliability were almost perfect (weighted κ = 0.85 [95%CI: 0.78-0.93] and 0.82 [95%CI 0.70-0.96] respectively). We confirmed a high correlation between the 4th rib score and subject age (rho = 0.72, p < 0.001), although the confidence intervals for individual age prediction were large, spanning over several decades. This study confirms the high reliability of Iscan method applied to PMCT for age estimation, although future multimodal age prediction techniques may help reducing the span of confidence intervals for individual age estimation.Trial registration: INDS 0,509,211,020, October 2020, retrospectively registered.

Human identification performed with skull's sphenoid sinus based on deep learning

Human identification plays a significant role in the investigations of disasters and criminal cases. Human identification could be achieved quickly and efficiently via 3D sphenoid sinus models by customized convolutional neural networks. In this retrospective study, a deep learning neural network was proposed to achieve human identification of 1475 noncontrast thin-slice CT scans. A total of 732 patients were retrieved and studied (82% for model training and 18% for testing). By establishing an individual recognition framework, the anonymous sphenoid sinus model was matched and cross-tested, and the performance of the framework also was evaluated on the test set using the recognition rate, ROC curve and identification speed. Finally, manual matching was performed based on the framework results in the test set. Out of a total of 732 subjects (mean age 46.45 years ± 14.92 (SD); 349 women), 600 subjects were trained, and 132 subjects were tested. The present automatic human identification has achieved Rank 1 and Rank 5 accuracy values of 93.94% and 99.24%, respectively, in the test set. In addition, all the identifications were completed within 55 s, which manifested the inference speed of the test set. We used the comparison results of the MVSS-Net to exclude sphenoid sinus models with low similarity and carried out traditional visual comparisons of the CT anatomical aspects of the sphenoid sinus of 132 individuals with an accuracy of 100%. The customized deep learning framework achieves reliable and fast human identification based on a 3D sphenoid sinus and can assist forensic radiologists in human identification accuracy.

The Effects of Cranial Orientation on Forensic Frontal Sinus Identification as Assessed by Outline Analyses

Simple Summary

Frontal sinus patterns are unique amongst individuals. When faced with an unknown decedent, investigators can compare the frontal sinus pattern observed in postmortem radiographs to antemortem radiographs of the suspected individual to make a positive identification. Ideally, the antemortem and postmortem radiographs are oriented in the same exact position, but this can be challenging. This study investigates how slight variations in radiographic orientation affect sinus outlines and potentially impact identification. Frontal sinus models were created from CT scans (21 individuals) and digitally oriented across three clinically relevant views. From each standard orientation (looking straight ahead), eight 5° deviations were obtained in horizontal (left/right), vertical (up/down), and diagonal (e.g., left-up vs. right-down) directions. Within and between individual differences in sinus size and outline shape were assessed. Sinus breadth remained relatively stable across deviations, while sinus height was affected by small vertical deviations. Although radiographic vertical deviations resulted in statistical differences, impacts on outline matches were minimal. However, practitioners need to take particular care in matching radiographic orientation for smaller and/or discontinuous (right and left sides separated) sinuses, which are more likely to lose part of the sinus in more inferiorly oriented views and, thus, could affect various methods of sinus identification.

Abstract

The utility of frontal sinuses for personal identification is widely recognized, but potential factors affecting its reliability remain uncertain. Deviations in cranial position between antemortem and postmortem radiographs may affect sinus appearance. This study investigates how slight deviations in orientations affect sinus size and outline shape and potentially impact identification. Frontal sinus models were created from CT scans of 21 individuals and digitally oriented to represent three clinically relevant radiographic views. From each standard view, model orientations were deviated at 5° intervals in horizontal, vertical, and diagonal (e.g., left-up) directions (27 orientations per individual). For each orientation, sinus dimensions were obtained, and outline shape was assessed by elliptical Fourier analyses and principal component (PC) analyses. Wilcoxon sign rank tests indicated that sinus breadth remained relatively stable (p > 0.05), while sinus height was significantly affected with vertical deviations (p < 0.006). Mann–Whitney U tests on Euclidean distances from the PC scores indicated consistently lower intra- versus inter-individual distances (p < 0.05). Two of the three orientations maintained perfect (100%) outline identification matches, while the third had a 98% match rate. Smaller and/or discontinuous sinuses were most problematic, and although match rates are high, practitioners should be aware of possible alterations in sinus variables when conducting frontal sinus identifications.

Computer-aided superimposition of the frontal sinus via 3D reconstruction for comparative forensic identification

The anatomical uniqueness of the frontal sinus morphology has been widely used for comparative forensic identification using various techniques, mostly including 2D X-rays or one fixed slice of an axial computed tomography (CT) scan image. However, computer-aided 3D automatic graphical comparison techniques can provide accurate comparisons between two 3D models that allow users to comply with even the strictest deviation standards, avoiding error-prone identification of frontal sinuses with similar morphologies. The study proposes the use of a computer-aided comparative paradigm based on the 3D-3D frontal sinus model superimposition process and further assesses the anatomical uniqueness of frontal sinuses using a large Chinese Han sample. Three hundred thirty-six patients older than 20 years with two multi-slice CT scans were collected. Frontal sinus 3D models were semi-automatically segmented through Dolphin Imaging software. Automatic pairwise comparisons of 336 matched pairs from the same person and 340 mismatched pairs from different individuals with an analysis of average root mean square (RMS) point-to-point distance were performed using Geomagic Studio Qualify software. RMS ranged between 0.005 and 1.032 (mean RMS 0.390 ± 0.25 mm) in the group of matches and between 1.107 and 19.363 (mean RMS 4.49 ± 2.69 mm) in the group of mismatches. On average, the RMS value was over ten-fold greater in mismatches than in matches. Statistically significant differences in RMS between the group of matches and mismatches were assessed using the Mann-Whitney U test (p < 0.05). This study supports the value of the frontal sinus with a 3D computer-aided superimposition method for human identification with large samples when DNA, fingerprints, and dental materials are not accessible.

Evaluation of anatomical landmark position differences for head computed tomography: A reliability study among technologists

INTRODUCTION

In computed tomography (CT) imaging protocols, lack of practice standards and variability in head positioning may all yield substantial inter-study image variance in the clinical setting which may limit the diagnostic and comparative value of subsequent scans. We aimed to evaluate repeatability of multiplanar reformatting of head CT based on the tuberculum sella (TS) to internal occipital protuberance (IOP) reference line and reduce variance.

METHODS

Reference lines that correspond to the TS-IOP plane on high-resolution CT scans were reviewed by technologists manually to calculate Yaw (z-rotation, rotation along the superoinferior direction), Pitch (x-rotation, rotation along the left-right direction), and Roll (y-rotation, rotation along the anteroposterior direction) angles in this pre-post design intervention study. The Yaw, Pitch, and Roll angles deviating from the reference TS-IOP in the head CT images before and after technologist training were measured with the technologists' actual graphical prescriptions, and their differences were calculated with t-tests. The intra-rater agreement was calculated using the intraclass correlation coefficient (ICC).

RESULTS

Mean pitch, yaw, and roll before technologist training was 6.7° ± 5.4°, 0.9° ± 1.5°, and 1.1° ± 1.2° and after training were 3.2° ± 2.6°, 0.6° ± 1.1°, and 0.6° ± 1.1°, respectively. Technologist training resulted in a significant decrease in pitch (p < 0.001) and roll (p = 0.001) inter-subject variability with respect to the TS-IOP line, however no significant difference for the yaw correction (p = 0.065) was noted. Intra-rater agreement regarding the reproducibility of TS-IOP reformation was excellent (ICC>0.950).

CONCLUSION

TS-IOP reference line corrected for direct roll, yaw, and pitch can be readily achieved by trained technologists.

IMPLICATIONS FOR PRACTICE

Adoption of the TS-IOP reference line should facilitate intra- and intermodality comparisons, leading to more reproducible and readily interpretable CT images.

A new reference line for coronal head CT to align with MRI: development of a standardised approach

BACKGROUND AND PURPOSE

There are great variations in how different technologists create the different imaging planes that can make a precise comparison of computed tomography and magnetic resonance imaging difficult. We aimed to identify a reference line for the coronal images on a computed tomography topography parallel to the posterior borderline of the brainstem (PB), matching standard coronal magnetic resonance imaging planes.

METHODS

We retrospectively reviewed computed tomography topography images of 80 consecutive patients to determine a computed tomography plane to match the PB on magnetic resonance imaging. These included the tuberculum sella (TS)-anterior arch of the C1 vertebra (C1), TS-tip of dens axis (D), dorsum sellae (DS)-C1 and DS-D. We compared these methods of prescribing the coronal computed tomography plane to coronal magnetic resonance imaging planes by measuring the angles between TS-C1 and PB, TS-M and PB, DS-C1 and PB, DS-D and PB on midsagittal brain magnetic resonance images. Bland-Altman plots were created to assess intra-observer reliability.

RESULTS

The angles between the PB line and each topogram-determined line are as follows: TS-C1, 10.40° ± 4.86°; TS-D, 22.46° ± 5.23°; DS-C1, 3.01° ± 3.16°; and DS-D, 11.53° ± 4.10°. The mean angles between the DS-C1 and the PB lines were significantly smaller than the mean angle between any other line (DS-D, TS-C1, or TS-D, all P < 0.001). Intra-observer agreement regarding the angular position of the reformatted coronal images on the lateral scout image was excellent (intraclass correlation coefficient >0.900, P < 0.05).

CONCLUSIONS

The DS-C1 is almost parallel to the PB and easily identifiable on the lateral scout topography of brain computed tomography. Utilising the DS-C1 line as the baseline for brain computed tomography could allow better corroboration with coronal magnetic resonance imaging angulation.

Morphological analysis of three-dimensionally reconstructed frontal sinuses from Chinese Han population using computed tomography

The uniqueness and reliability of frontal sinuses for personal identification have gained wide recognition in forensics. However, few studies have assessed the usefulness of a three-dimensional (3D) model of the frontal sinus for human identification. This study aimed to develop standardized techniques to classify the frontal sinus according to its 3D morphological metrics and discover the usefulness of the 3D frontal sinus model in identification of Chinese Han population. One hundred and ninety-six computed tomography (CT) scans of patients older than 20 years (84 males and 112 females) were collected. A 3D frontal sinus digital model was segmented using Dolphin Imaging software. The following morphological metrics of the 3D frontal sinus were used to develop the coding system: bilateral or unilateral, spatial relationships of the two sides, number of septations, superior volume side, the shape of the 3D model of each side, shape of the medial surface and frontal ostium on each side, number of accessory septations on each side, number of supra-orbital cells of the medial surface and lateral surface on each side, and number of the arcades on each side. The new coding system accurately identified all of our research individuals. This study discovered a number of individual variations in the 3D frontal sinus morphology patterns. A coding system, which is based on these morphological patterns, exposes the morphological variants of frontal sinuses and presents the usefulness of 3D frontal sinus model for human identification.

Virtual reconstruction of paranasal sinuses from CT data: A feasibility study for forensic application

PURPOSE

The purpose of this study was to report the feasibility of computed modelization and reconstitution of the paranasal sinuses, before and after trauma, from CT data.

MATERIALS AND METHODS

We modeled and reconstructed the paranasal sinuses of two patients (A and B), before and after trauma, using two different softwares (3DSlicer^® and Blender^®). Both patients had different numbers and locations of fractures. The 3DSlicer^® software was used to create a 3D model from CT data. We then imported the 3D data into the Blender^® software, to reconstruct and compare the dimensions of the paranasal sinuses before and after trauma.

RESULTS

The 3 fragments of patient A and the 7 fragments of patient B could be repositioned in the pre-traumatic configuration. Distance measurements proved to be similar between pre- and post-traumatic 3D volumes.

CONCLUSION

After simple trauma, bone facial anatomy reconstruction is manually feasible. The whole procedure could benefit from automatization through machine learning. However, this feasibility must be confirmed on more severely fractured paranasal sinuses, to consider an application in forensic identification.