Andreas Heinrich
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Andreas Heinrich has a total of 98 co-authors. Here are the top co-authors by number of shared publications:
- Ulf Teichgräber
Department of Radiology, Jena University Hospital, Friedrich Schiller University Jena, Jena, Germany
16 shared publications - Felix Güttler
Institute of Diagnostic and Interventional Radiology, Jena University Hospital, Friedrich Schiller University Jena, Jena, GERMANY
9 shared publications Gita Mall
Institute of Forensic Medicine, Jena University Hospital – Friedrich Schiller University, Jena, Thuringia, Germany
7 shared publicationsSebastian Schenkl
Institute of Forensic Medicine, Jena University Hospital – Friedrich Schiller University, Jena, Germany
7 shared publicationsFelix Güttler
Institute for Diagnostic and Interventional Radiology, Jena University Hospital, Jena, Germany
6 shared publicationsFelix V. Güttler
Department of Radiology, Jena University Hospital – Friedrich Schiller University, Jena, Germany
6 shared publications
Latest Publications
- Open Access
- 14 July 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
AI-based organ weight estimation from postmortem computed tomography.
Accurate organ weight determination is essential in forensic autopsy. Postmortem computed tomography (CT) combined with Artificial Intelligence (AI)-based segmentation allows automated, non-invasive organ delineation. This study evaluated the accuracy of CT-derived organ weights against conventional autopsy measurements. A total of 100 postmortem CT examinations (mean age 54 ± 21 years; 35 females, 65 males) were analyzed. Brain, heart, liver, kidneys, and spleen were segmented automatically in two model resolutions, and organ weights were derived from voxel volumes and Hounsfield unit-based tissue densities. Across all organs, mean differences between CT- and autopsy-derived organ weights were - 12.69 ± 107.96 g (median - 2.95 g) for the 1.5 mm model and - 18.42 ± 116.67 g (median - 1.95 g) for the 3 mm model. Mean absolute differences were 70.76 ± 83.07 g (median 29.05 g) and 77.43 ± 90.60 g (median 30.50 g), respectively. Agreement was highest for the brain and liver, followed by the kidneys and spleen, whereas the heart showed the largest deviations. Segmentation failures were uncommon and were mainly associated with decomposition, trauma, or incomplete organ coverage. These findings indicate that AI-based segmentation enables reliable, largely non-invasive organ weight estimation in postmortem CT and may support forensic investigations and personal identification workflows based on antemortem imaging.
- 14 July 2026
- Published by Springer Nature in Scientific Reports
- Vol. 16 (1)
Open AccessAI-based organ weight estimation from postmortem computed tomography.
Accurate organ weight determination is essential in forensic autopsy. Postmortem computed tomography (CT) combined with Artificial Intelligence (AI)-based segmentation allows automated, non-invasive organ delineation. This study evaluated the accuracy of CT-derived organ weights against conventional autopsy measurements. A total of 100 postmortem CT examinations (mean age 54 ± 21 years; 35 females, 65 males) were analyzed. Brain, heart, liver, kidneys, and spleen were segmented automatically in two model resolutions, and organ weights were derived from voxel volumes and Hounsfield unit-based tissue densities. Across all organs, mean differences between CT- and autopsy-derived organ weights were - 12.69 ± 107.96 g (median - 2.95 g) for the 1.5 mm model and - 18.42 ± 116.67 g (median - 1.95 g) for the 3 mm model. Mean absolute differences were 70.76 ± 83.07 g (median 29.05 g) and 77.43 ± 90.60 g (median 30.50 g), respectively. Agreement was highest for the brain and liver, followed by the kidneys and spleen, whereas the heart showed the largest deviations. Segmentation failures were uncommon and were mainly associated with decomposition, trauma, or incomplete organ coverage. These findings indicate that AI-based segmentation enables reliable, largely non-invasive organ weight estimation in postmortem CT and may support forensic investigations and personal identification workflows based on antemortem imaging.
- Open Access
- 24 April 2026
- Published by MDPI AG in Bioengineering
- Vol. 13 (5) , 494
Automatic Personal Identification Using a Single MRI Slice
Identification of unknown individuals is challenging, and radiological imaging databases provide rich anatomical information for automated recognition. This study evaluated whether a single routine magnetic resonance imaging (MRI) slice contains sufficient person-specific features to identify individuals in large databases. It analyzed 11,078 head MRI examinations from 5770 individuals (age 52 ± 18 years, 2714 men) acquired between 2002 and 2025. For identification, 112 individuals were randomly selected across eight 10-year age groups, and one slice from four anatomical regions was extracted. The remaining 10,966 MRI examinations with 247,804 slices formed the reference database. Distinctive anatomical features were automatically extracted using computer vision (CV), and the identification rate was evaluated by rank. Using a single MRI slice, the identification rate at rank 1 reached 96% (107/112) for the best-performing region, the maxillary sinus, among 5770 potential identities. Across all regions, the rank 1 identification rate ranged from 91% to 96%; combining them increased rank 1 and 10 identification rates to 98% (110/112) and 99% (111/112). Identification rate remained stable over several years, with only two cases showing reduced rank 1 performance, likely due to age-related morphological changes. A single MRI slice contains stable, individualized features sufficient for reliable identification in large databases, supporting automated CV-based personal identification across years.
- 24 April 2026
- Published by MDPI AG in Bioengineering
- Vol. 13 (5) , 494
Open AccessAutomatic Personal Identification Using a Single MRI Slice
Identification of unknown individuals is challenging, and radiological imaging databases provide rich anatomical information for automated recognition. This study evaluated whether a single routine magnetic resonance imaging (MRI) slice contains sufficient person-specific features to identify individuals in large databases. It analyzed 11,078 head MRI examinations from 5770 individuals (age 52 ± 18 years, 2714 men) acquired between 2002 and 2025. For identification, 112 individuals were randomly selected across eight 10-year age groups, and one slice from four anatomical regions was extracted. The remaining 10,966 MRI examinations with 247,804 slices formed the reference database. Distinctive anatomical features were automatically extracted using computer vision (CV), and the identification rate was evaluated by rank. Using a single MRI slice, the identification rate at rank 1 reached 96% (107/112) for the best-performing region, the maxillary sinus, among 5770 potential identities. Across all regions, the rank 1 identification rate ranged from 91% to 96%; combining them increased rank 1 and 10 identification rates to 98% (110/112) and 99% (111/112). Identification rate remained stable over several years, with only two cases showing reduced rank 1 performance, likely due to age-related morphological changes. A single MRI slice contains stable, individualized features sufficient for reliable identification in large databases, supporting automated CV-based personal identification across years.
- Open Access
- 6 April 2026
- Published by MDPI AG in Bioengineering
- Vol. 13 (4) , 426
AI-Based Angle Map Analysis of Facial Asymmetry in Peripheral Facial Palsy
Peripheral facial palsy (PFP) causes pronounced facial asymmetry and functional impairment, highlighting the need for reliable, objective assessment. This study presents a novel, fully automated, reference-free method for quantifying facial symmetry using artificial intelligence (AI)-based facial landmark detection. A total of 405 datasets from 198 PFP patients were analyzed, each including nine standardized facial expressions covering both resting and dynamic movements. AI detected 478 landmarks per image, from which 225 paired landmarks were used to compute local asymmetry angles. Systematic evaluation identified 91 highly informative landmark pairs, primarily around the eyes, nose and mouth, which simplified the analysis and enhanced discriminatory power, while also enabling region-specific assessment of asymmetry. Statistical evaluation included Kruskal–Wallis H-tests across clinical scores and Spearman correlations, showing moderate to strong associations (0.32–0.73, p < 0.001). The fully automated pipeline produced reproducible results and demonstrated robustness to head rotation. Intuitive full-face angle maps allowed direct assessment of asymmetry without a reference image. This AI-driven approach provides a robust, objective, and visually interpretable framework for clinical monitoring, severity classification, and treatment evaluation in PFP, combining quantitative precision with practical applicability.
- 6 April 2026
- Published by MDPI AG in Bioengineering
- Vol. 13 (4) , 426
Open AccessAI-Based Angle Map Analysis of Facial Asymmetry in Peripheral Facial Palsy
Peripheral facial palsy (PFP) causes pronounced facial asymmetry and functional impairment, highlighting the need for reliable, objective assessment. This study presents a novel, fully automated, reference-free method for quantifying facial symmetry using artificial intelligence (AI)-based facial landmark detection. A total of 405 datasets from 198 PFP patients were analyzed, each including nine standardized facial expressions covering both resting and dynamic movements. AI detected 478 landmarks per image, from which 225 paired landmarks were used to compute local asymmetry angles. Systematic evaluation identified 91 highly informative landmark pairs, primarily around the eyes, nose and mouth, which simplified the analysis and enhanced discriminatory power, while also enabling region-specific assessment of asymmetry. Statistical evaluation included Kruskal–Wallis H-tests across clinical scores and Spearman correlations, showing moderate to strong associations (0.32–0.73, p < 0.001). The fully automated pipeline produced reproducible results and demonstrated robustness to head rotation. Intuitive full-face angle maps allowed direct assessment of asymmetry without a reference image. This AI-driven approach provides a robust, objective, and visually interpretable framework for clinical monitoring, severity classification, and treatment evaluation in PFP, combining quantitative precision with practical applicability.
- Open Access
- 4 March 2026
- Published by Georg Thieme Verlag KG in RöFo - Fortschritte auf dem Gebiet der Röntgenstrahlen und der bildgebenden Verfahren
- Vol. 198 (11) , 1426-1436
A Review on Automatic Personal Identification Using Panoramic Radiographs and Computed Tomography
Identifying completely unknown individuals is a major challenge in forensic and emergency medicine. Radiology offers a promising solution by using unique anatomical features on medical images to identify both living and deceased persons. Although emergency or postmortem images could be matched against large clinical databases, such applications remain largely experimental. This review examines current methods in automatic radiology-based personal identification, evaluates their performance, and highlights potential applications in forensic and clinical settings. A narrative review of studies published from 2018 onwards was conducted using PubMed and Google Scholar. Included studies applied automated or semi-automated personal identification to panoramic radiographs (PR) or computed tomography (CT) using reference datasets. A narrative approach was used to synthesize results descriptively due to heterogeneity in study design, dataset size, and methodology. Of the 32 included studies, 15 focused on PR-to-PR, 8 on head CT-to-CT, 7 on body CT-to-CT, and 2 on CT-to-PR identification. The most commonly applied approach was descriptor-based computer vision (CV), used in 9 studies. Deep learning was applied in 8 studies for feature extraction, and in 2 studies each for classification and bone segmentation. Several methods perform well in controlled settings. Descriptor-based CV provides the most flexibility and strongest evidence, especially for large-database comparisons and postmortem applications. Deep learning approaches, including feature extraction, classification, and automatic bone segmentation, also show promise for cross-individual matching but require further validation. Automatic radiology-based personal identification holds significant potential for forensic and clinical use, yet the development of standardized large-scale reference databases and robust automated pipelines remains a key challenge.
- 4 March 2026
- Published by Georg Thieme Verlag KG in RöFo - Fortschritte auf dem Gebiet der Röntgenstrahlen und der bildgebenden Verfahren
- Vol. 198 (11) , 1426-1436
Open AccessA Review on Automatic Personal Identification Using Panoramic Radiographs and Computed Tomography
Identifying completely unknown individuals is a major challenge in forensic and emergency medicine. Radiology offers a promising solution by using unique anatomical features on medical images to identify both living and deceased persons. Although emergency or postmortem images could be matched against large clinical databases, such applications remain largely experimental. This review examines current methods in automatic radiology-based personal identification, evaluates their performance, and highlights potential applications in forensic and clinical settings. A narrative review of studies published from 2018 onwards was conducted using PubMed and Google Scholar. Included studies applied automated or semi-automated personal identification to panoramic radiographs (PR) or computed tomography (CT) using reference datasets. A narrative approach was used to synthesize results descriptively due to heterogeneity in study design, dataset size, and methodology. Of the 32 included studies, 15 focused on PR-to-PR, 8 on head CT-to-CT, 7 on body CT-to-CT, and 2 on CT-to-PR identification. The most commonly applied approach was descriptor-based computer vision (CV), used in 9 studies. Deep learning was applied in 8 studies for feature extraction, and in 2 studies each for classification and bone segmentation. Several methods perform well in controlled settings. Descriptor-based CV provides the most flexibility and strongest evidence, especially for large-database comparisons and postmortem applications. Deep learning approaches, including feature extraction, classification, and automatic bone segmentation, also show promise for cross-individual matching but require further validation. Automatic radiology-based personal identification holds significant potential for forensic and clinical use, yet the development of standardized large-scale reference databases and robust automated pipelines remains a key challenge.
- Open Access
- 4 December 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (12) , e0337989
Personal identification via matching of curved multiplanar computed tomography reconstructions and panoramic radiographs
Computer vision (CV)-based personal identification enables automated matching of recent radiological images with clinical databases to identify unknown individuals. This study aimed to assess whether a panoramic radiograph (PR)-like image reconstructed from computed tomography (CT) data using curved multiplanar reconstruction could enable CV-based personal identification using a PR database. A method was developed to automatically generate PR-like images with adjustable parameters, based on 50 CT examinations including the jaw region (38.64 ± 16.72 years; 17 females, 33 males), allowing for variations such as tooth rotations. Systematic modification of parameters enabled the generation of different representations to determine optimal settings for a large number of individuals. Multiple PR-like images per identity were tested against a PR database containing 82,036 PRs from 43,379 individuals. Utilizing the most effective individual parameter settings, 72% (36/50) of individuals were correctly identified at rank 1, 82% (41/50) at rank 10, and 96% (48/50) at rank 100 – out of 43,379 possible individuals. The rank describes the position of the matched image in a list sorted after a descending similarity score. When the optimal parameters were applied to a larger number of individuals, the identification rates were 50% (25/50) at rank 1, 64% (32/50) at rank 10, and 78% (39/50) at rank 100. In conclusion, CV demonstrates potential for personal identification by comparing automatically generated PR-like images with a large PR database.
- 4 December 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (12) , e0337989
Open AccessPersonal identification via matching of curved multiplanar computed tomography reconstructions and panoramic radiographs
Computer vision (CV)-based personal identification enables automated matching of recent radiological images with clinical databases to identify unknown individuals. This study aimed to assess whether a panoramic radiograph (PR)-like image reconstructed from computed tomography (CT) data using curved multiplanar reconstruction could enable CV-based personal identification using a PR database. A method was developed to automatically generate PR-like images with adjustable parameters, based on 50 CT examinations including the jaw region (38.64 ± 16.72 years; 17 females, 33 males), allowing for variations such as tooth rotations. Systematic modification of parameters enabled the generation of different representations to determine optimal settings for a large number of individuals. Multiple PR-like images per identity were tested against a PR database containing 82,036 PRs from 43,379 individuals. Utilizing the most effective individual parameter settings, 72% (36/50) of individuals were correctly identified at rank 1, 82% (41/50) at rank 10, and 96% (48/50) at rank 100 – out of 43,379 possible individuals. The rank describes the position of the matched image in a list sorted after a descending similarity score. When the optimal parameters were applied to a larger number of individuals, the identification rates were 50% (25/50) at rank 1, 64% (32/50) at rank 10, and 78% (39/50) at rank 100. In conclusion, CV demonstrates potential for personal identification by comparing automatically generated PR-like images with a large PR database.
- Open Access
- 15 October 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-11
Automated body weight measurement using computed tomography
Computed tomography (CT) images provide detailed information on tissue composition and density. Body weight estimation from CT can be highly relevant in emergency and forensic contexts where conventional weighing is impractical or unavailable—for example, due to unconscious or immobilized patients, lack of appropriate equipment, or the necessity to preserve forensic trace evidence. An automated method for body weight estimation using antemortem and postmortem CT data from clinical scanners was evaluated. Two retrospective cohorts were included: 30 forensic postmortem whole-body CT scans and 66 clinical torso CT scans. Tissue segmentation applied predefined Hounsfield Unit thresholds for fat, muscle, and bone, with voxel volumes multiplied by corresponding tissue densities to calculate weight. In forensic cases, CT-derived weights differed by − 1.26 ± 2.58 kg (median: − 1.15 kg) from measured weights. In clinical cases, torso weight was extrapolated to total body weight using a corrected scaling factor that accounted for body and scan length, resulting in an average difference of 0.53 ± 4.62 kg (median: 0.23 kg). Linear regression demonstrated strong correlations (R2 = 0.97/0.96 forensic/clinical). These findings indicate that routinely acquired CT scans enable accurate, automated, and retrospective body weight estimation, providing a non-invasive alternative when direct measurement is not feasible while simultaneously allowing body composition analysis.
- 15 October 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-11
Open AccessAutomated body weight measurement using computed tomography
Computed tomography (CT) images provide detailed information on tissue composition and density. Body weight estimation from CT can be highly relevant in emergency and forensic contexts where conventional weighing is impractical or unavailable—for example, due to unconscious or immobilized patients, lack of appropriate equipment, or the necessity to preserve forensic trace evidence. An automated method for body weight estimation using antemortem and postmortem CT data from clinical scanners was evaluated. Two retrospective cohorts were included: 30 forensic postmortem whole-body CT scans and 66 clinical torso CT scans. Tissue segmentation applied predefined Hounsfield Unit thresholds for fat, muscle, and bone, with voxel volumes multiplied by corresponding tissue densities to calculate weight. In forensic cases, CT-derived weights differed by − 1.26 ± 2.58 kg (median: − 1.15 kg) from measured weights. In clinical cases, torso weight was extrapolated to total body weight using a corrected scaling factor that accounted for body and scan length, resulting in an average difference of 0.53 ± 4.62 kg (median: 0.23 kg). Linear regression demonstrated strong correlations (R2 = 0.97/0.96 forensic/clinical). These findings indicate that routinely acquired CT scans enable accurate, automated, and retrospective body weight estimation, providing a non-invasive alternative when direct measurement is not feasible while simultaneously allowing body composition analysis.
- Open Access
- 27 August 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-11
Deep learning-based automatic facial symmetry scoring in peripheral facial palsy
Unilateral peripheral facial palsy (PFP) results in facial asymmetry and functional impairment, reducing quality of life. Accurate, objective assessment is vital for monitoring and rehabilitation. This study presents an automated method utilizes standardized 2D photographs to visualize facial dynamics using heatmaps and calculates an objective symmetry score, quantifying movement symmetry. Retrospective analysis included 405 facial datasets from 198 PFP patients. Images were processed using a deep learning-based facial landmark detection and an affine alignment algorithm. Heatmaps were generated from grayscale difference images, and symmetry scores calculated by comparing mirrored facial halves within a defined mask. Stennert movement scores were correlated with symmetry scores using Spearman’s rank correlation. The method was applied successfully to all datasets, with symmetry scores ranging from 0 to 0.99 (mean 0.85 ± 0.12), varying by expression level. Heatmaps highlighted asymmetries matching clinical findings. In 85% of cases, Stennert trends aligned with symmetry scores; 9% showed stable Stennert scores but changing symmetry scores, suggesting higher sensitivity. Significant negative correlations (r = − 0.32 to − 0.66, p < 0.0001) confirmed greater clinical severity corresponds to lower symmetry scores. In conclusion, the automated method provides an objective, reliable, and accessible tool for assessing facial symmetry in PFP, thereby improving clinical evaluation and facilitating precise rehabilitation monitoring.
- 27 August 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-11
Open AccessDeep learning-based automatic facial symmetry scoring in peripheral facial palsy
Unilateral peripheral facial palsy (PFP) results in facial asymmetry and functional impairment, reducing quality of life. Accurate, objective assessment is vital for monitoring and rehabilitation. This study presents an automated method utilizes standardized 2D photographs to visualize facial dynamics using heatmaps and calculates an objective symmetry score, quantifying movement symmetry. Retrospective analysis included 405 facial datasets from 198 PFP patients. Images were processed using a deep learning-based facial landmark detection and an affine alignment algorithm. Heatmaps were generated from grayscale difference images, and symmetry scores calculated by comparing mirrored facial halves within a defined mask. Stennert movement scores were correlated with symmetry scores using Spearman’s rank correlation. The method was applied successfully to all datasets, with symmetry scores ranging from 0 to 0.99 (mean 0.85 ± 0.12), varying by expression level. Heatmaps highlighted asymmetries matching clinical findings. In 85% of cases, Stennert trends aligned with symmetry scores; 9% showed stable Stennert scores but changing symmetry scores, suggesting higher sensitivity. Significant negative correlations (r = − 0.32 to − 0.66, p < 0.0001) confirmed greater clinical severity corresponds to lower symmetry scores. In conclusion, the automated method provides an objective, reliable, and accessible tool for assessing facial symmetry in PFP, thereby improving clinical evaluation and facilitating precise rehabilitation monitoring.
- Open Access
- 24 May 2025
- Published by Springer Nature in International journal of legal medicine
- Vol. 139 (5) , 1-12
Automated identification of unknown decedents: matching postmortem CT images with clinical databases
The maxillary sinus plays an important role in the forensic identification of unknown deceased individuals. This study aimed to evaluate whether matching postmortem computed tomography (CT) images from virtual autopsies with antemortem CT examinations from a clinical database enables reliable identification using computer vision (CV) techniques. From ten virtual autopsies, CT images of the maxillary sinus were selected for comparison against 853 antemortem CT examinations from 738 individuals. A total of 60,255 antemortem CT slices underwent image processing, CV feature extraction, and were stored in an antemortem CV database. The number of matching points between CV features of the postmortem image and the antemortem reference image served as an indicator of identification accuracy. The identification rate was 50% (5/10) at rank 1 (with the sought identity having the highest number of matching points), 80% (8/10) at rank 2, and 100% (10/10) at rank 7 among the 738 potential identities. Challenges were observed when the antemortem reference CT examination depicted only parts of the maxillary sinus or when injuries were present. Additionally, postmortem imaging should closely replicate antemortem imaging standards to maximize the number of CV matching points. In conclusion, the findings suggest that it is feasible to identify individuals using postmortem CT images from virtual autopsies in combination with a clinical database. However, postmortem imaging should ideally adhere to clinical imaging standards to achieve more CV matching points for the sought identity with the antemortem reference.
- 24 May 2025
- Published by Springer Nature in International journal of legal medicine
- Vol. 139 (5) , 1-12
Open AccessAutomated identification of unknown decedents: matching postmortem CT images with clinical databases
The maxillary sinus plays an important role in the forensic identification of unknown deceased individuals. This study aimed to evaluate whether matching postmortem computed tomography (CT) images from virtual autopsies with antemortem CT examinations from a clinical database enables reliable identification using computer vision (CV) techniques. From ten virtual autopsies, CT images of the maxillary sinus were selected for comparison against 853 antemortem CT examinations from 738 individuals. A total of 60,255 antemortem CT slices underwent image processing, CV feature extraction, and were stored in an antemortem CV database. The number of matching points between CV features of the postmortem image and the antemortem reference image served as an indicator of identification accuracy. The identification rate was 50% (5/10) at rank 1 (with the sought identity having the highest number of matching points), 80% (8/10) at rank 2, and 100% (10/10) at rank 7 among the 738 potential identities. Challenges were observed when the antemortem reference CT examination depicted only parts of the maxillary sinus or when injuries were present. Additionally, postmortem imaging should closely replicate antemortem imaging standards to maximize the number of CV matching points. In conclusion, the findings suggest that it is feasible to identify individuals using postmortem CT images from virtual autopsies in combination with a clinical database. However, postmortem imaging should ideally adhere to clinical imaging standards to achieve more CV matching points for the sought identity with the antemortem reference.
- Open Access
- 27 April 2025
- Published by Springer Nature in European Radiology
- Vol. 35 (10) , 5900-5910
Computer vision-based personal identification using 2D maximum intensity projection CT images
Objectives: Computer vision (CV) mimics human vision, enabling the automatic comparison of radiological images from recent examinations with a vast image database for unique identification. This method offers significant potential in emergencies involving unknown individuals. This study assesses whether maximum intensity projection (MIP) images from thoracic computed tomography (CT) examinations are suitable for automated CV-based personal identification. Methods: The study analyzed 12,465 native CT examinations of the thorax from 8177 individuals, focusing on MIP images to assess their potential for CV-based personal identification in 300 cases. CV automatically identifies and describes features in images, which are then matched to reference images. The number of matching points was used as an indicator of identification accuracy. Results: The identification rate was 98.67% (296/300) at rank 1 and 99.67% (299/300) at rank 10, among over 8177 potential identities. Matching points were higher for images of the same individual (7.43 ± 5.83%) compared to different individuals (0.16 ± 0.14%), with 100% representing the maximum possible matching points. Reliable matching points were mainly found in the thoracic skeleton, sternum, and spine. Challenges arose when the patient was curved on the table or when medical equipment was present in the image. Conclusion: Unambiguous identification based on MIP images from thoracic CT examinations is highly reliable, even for large CV databases. This method is applicable to various 2D reconstructions, provided anatomical structures are comparably represented. Radiology offers extensive reference images for CV databases, enhancing automated personal identification in emergencies. Key Points: Question Computer vision-based personal identification holds great potential, but it remains unclear whether maximum intensity projection images from thoracic-CT scans are suitable for this purpose. Findings Maximum intensity projection images of the thorax are highly individual, with computer vision-based identification achieving nearly 100% rank-1 accuracy across a potential 8177 identities. Clinical relevance Radiology holds a vast collection of reference images for a computer vision database, enabling automated personal identification in emergency examinations. This improves patient care and communication with relatives by providing access to medical history. Graphical
- 27 April 2025
- Published by Springer Nature in European Radiology
- Vol. 35 (10) , 5900-5910
Open AccessComputer vision-based personal identification using 2D maximum intensity projection CT images
Objectives: Computer vision (CV) mimics human vision, enabling the automatic comparison of radiological images from recent examinations with a vast image database for unique identification. This method offers significant potential in emergencies involving unknown individuals. This study assesses whether maximum intensity projection (MIP) images from thoracic computed tomography (CT) examinations are suitable for automated CV-based personal identification. Methods: The study analyzed 12,465 native CT examinations of the thorax from 8177 individuals, focusing on MIP images to assess their potential for CV-based personal identification in 300 cases. CV automatically identifies and describes features in images, which are then matched to reference images. The number of matching points was used as an indicator of identification accuracy. Results: The identification rate was 98.67% (296/300) at rank 1 and 99.67% (299/300) at rank 10, among over 8177 potential identities. Matching points were higher for images of the same individual (7.43 ± 5.83%) compared to different individuals (0.16 ± 0.14%), with 100% representing the maximum possible matching points. Reliable matching points were mainly found in the thoracic skeleton, sternum, and spine. Challenges arose when the patient was curved on the table or when medical equipment was present in the image. Conclusion: Unambiguous identification based on MIP images from thoracic CT examinations is highly reliable, even for large CV databases. This method is applicable to various 2D reconstructions, provided anatomical structures are comparably represented. Radiology offers extensive reference images for CV databases, enhancing automated personal identification in emergencies. Key Points: Question Computer vision-based personal identification holds great potential, but it remains unclear whether maximum intensity projection images from thoracic-CT scans are suitable for this purpose. Findings Maximum intensity projection images of the thorax are highly individual, with computer vision-based identification achieving nearly 100% rank-1 accuracy across a potential 8177 identities. Clinical relevance Radiology holds a vast collection of reference images for a computer vision database, enabling automated personal identification in emergency examinations. This improves patient care and communication with relatives by providing access to medical history. Graphical
- 27 February 2025
- Published by Springer Nature in Rechtsmedizin
- Vol. 35 (2) , 55-64
Künstliche Intelligenz in der forensisch-radiologischen Altersdiagnostik
- 27 February 2025
- Published by Springer Nature in Rechtsmedizin
- Vol. 35 (2) , 55-64
Künstliche Intelligenz in der forensisch-radiologischen Altersdiagnostik