Pub Date : 2026-03-01Epub Date: 2025-10-09DOI: 10.1016/j.fri.2025.200655
Asan Petrus , Syafruddin Ilyas , Adang Bachtiar , Imam Budi Putra , Ahmad Yudianto
The incidence of disasters has increased significantly in Indonesia over the last two decades, necessitating accurate forensic identification methods to determine victims' identities, including estimating their height from metacarpal bone lengths, a metric for which no official formula yet exists in Indonesia. This study, an analytical correlational cross-sectional research conducted in North Sumatra from March to May 2024, involved 138 subjects and used cluster random sampling. Primary data, namely measurements of the length of the metacarpal bones obtained through computerized X- ray photos of the respondents' right and left palms connected to an X-ray machine, then the data was processed using SPSS software version 26. The statistical analysis methods used include normality tests, Pearson correlation tests, and linear regression tests. Findings showed a significance level (Sig) of 0.001 with a coefficient interval ranging from 0.438 to 0.756, indicating a moderate to strong correlation. A total of 63 new formulas for estimating height from metacarpal bones I-V were discovered, with the best predictors varying by gender and overall best involving multiple variables. This research provides a new set of formulas for estimating height from metacarpal lengths, applicable to the North Sumatran population.
{"title":"Height prediction using metacarpal lengths as measured on radiography in postmortem identification","authors":"Asan Petrus , Syafruddin Ilyas , Adang Bachtiar , Imam Budi Putra , Ahmad Yudianto","doi":"10.1016/j.fri.2025.200655","DOIUrl":"10.1016/j.fri.2025.200655","url":null,"abstract":"<div><div>The incidence of disasters has increased significantly in Indonesia over the last two decades, necessitating accurate forensic identification methods to determine victims' identities, including estimating their height from metacarpal bone lengths, a metric for which no official formula yet exists in Indonesia. This study, an analytical correlational cross-sectional research conducted in North Sumatra from March to May 2024, involved 138 subjects and used cluster random sampling. Primary data, namely measurements of the length of the metacarpal bones obtained through computerized X- ray photos of the respondents' right and left palms connected to an X-ray machine, then the data was processed using SPSS software version 26. The statistical analysis methods used include normality tests, Pearson correlation tests, and linear regression tests. Findings showed a significance level (Sig) of 0.001 with a coefficient interval ranging from 0.438 to 0.756, indicating a moderate to strong correlation. A total of 63 new formulas for estimating height from metacarpal bones I-V were discovered, with the best predictors varying by gender and overall best involving multiple variables. This research provides a new set of formulas for estimating height from metacarpal lengths, applicable to the North Sumatran population.</div></div>","PeriodicalId":40763,"journal":{"name":"Forensic Imaging","volume":"44 ","pages":"Article 200655"},"PeriodicalIF":1.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145884091","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2026-01-15DOI: 10.1016/j.fri.2026.200674
Angi M. Christensen , Mariyam I. Isa , Brian F. Spatola , Timothy P. Gocha
Skeletal trauma analysis typically relies on post-hoc interpretation of fracture features such as crack configuration and fragment shape. Understanding how cracks initiate and propagate is foundational for explaining how such features form, especially in complex trauma where fragments may not be recovered or cannot be fully reconstructed. This study used high-speed video to document crack initiation and propagation in 30 whole human long bones (femora, tibiae, and humeri) from ten cadaveric donors following projectile impact with 9 mm full metal jacket bullets. Two cameras recording at 55,000 frames per second were positioned to capture crack formation at the entrance and exit sites. Fracture sequences were typically complete within 5–8 frames (91–145 μs), with the longest lasting 10 frames (182 μs). A consistent entrance site sequence emerged across all bones and impact locations: initial chipping at the impact site, followed by radial crack propagation, then formation of circumferential cracks between adjacent radial cracks. The entrance sequence mirrors patterns reported in cranial projectile impacts. Crack propagation opposite the entrance was more complex and diverged from cranial patterns. Longitudinal cracks often formed before exit defects and sometimes curved back toward the entrance, intersecting entrance radial cracks. Transverse cracks occasionally developed between longitudinal cracks. The results suggest greater interaction between entrance and exit cracks in long bones than in crania, likely due to their smaller diameter. The fracture sequences captured in these experiments can inform more accurate interpretations of long bone projectile trauma.
{"title":"Crack propagation in projectile impacted human long bones","authors":"Angi M. Christensen , Mariyam I. Isa , Brian F. Spatola , Timothy P. Gocha","doi":"10.1016/j.fri.2026.200674","DOIUrl":"10.1016/j.fri.2026.200674","url":null,"abstract":"<div><div>Skeletal trauma analysis typically relies on post-hoc interpretation of fracture features such as crack configuration and fragment shape. Understanding how cracks initiate and propagate is foundational for explaining how such features form, especially in complex trauma where fragments may not be recovered or cannot be fully reconstructed. This study used high-speed video to document crack initiation and propagation in 30 whole human long bones (femora, tibiae, and humeri) from ten cadaveric donors following projectile impact with 9 mm full metal jacket bullets. Two cameras recording at 55,000 frames per second were positioned to capture crack formation at the entrance and exit sites. Fracture sequences were typically complete within 5–8 frames (91–145 μs), with the longest lasting 10 frames (182 μs). A consistent entrance site sequence emerged across all bones and impact locations: initial chipping at the impact site, followed by radial crack propagation, then formation of circumferential cracks between adjacent radial cracks. The entrance sequence mirrors patterns reported in cranial projectile impacts. Crack propagation opposite the entrance was more complex and diverged from cranial patterns. Longitudinal cracks often formed before exit defects and sometimes curved back toward the entrance, intersecting entrance radial cracks. Transverse cracks occasionally developed between longitudinal cracks. The results suggest greater interaction between entrance and exit cracks in long bones than in crania, likely due to their smaller diameter. The fracture sequences captured in these experiments can inform more accurate interpretations of long bone projectile trauma.</div></div>","PeriodicalId":40763,"journal":{"name":"Forensic Imaging","volume":"44 ","pages":"Article 200674"},"PeriodicalIF":1.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147394497","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2025-11-27DOI: 10.1016/j.fri.2025.200661
G. de Heus , B.S. de Bakker , S.J. Decker , J. Elifritz , H. Hyodoh , F. Marttinen , C. O’Donnell , T.D. Ruder , M.J. Thali , R.R. van Rijn , ISFRI-IAFR congress consortium
Background
The International Society of Forensic Radiology and Imaging (ISFRI) aims to advance forensic radiology and imaging worldwide through its journal Forensic Imaging, formerly Journal of Forensic Radiology and Imaging, and annual congresses. A way to measure success and quality of research presented at these congresses is by assessing the abstract to publication conversion rate.
Objective
To evaluate the percentage of abstracts that were published and identify the journals in which they appeared.
Materials and methods
The publication rate of scientific abstracts from ISFRI and joint ISFRI and International Association of Forensic Radiographers (IAFR) congresses between 2012 and 2022 was determined. Searches were conducted in Pubmed, Google and Google Scholar, followed by manual checks of the table of contents of Journal of Forensic Radiology and Imaging and Forensic Imaging. Results were compared with conversion rates from radiological and nuclear imaging, and forensic medical studies in the Cochrane review by Scherer et al., supplemented with other relevant publications.
Results
Of 464 presented abstracts, 221 (47.6 %) were eventually published as full articles.
Most abstracts were original research 279 (60.1 %), followed by 99 (21.3 %) case reports, and 79 (17.0 %) reviews. Abstracts were published in a wide variety of journals, but the majority 52 (23.5 %) were published in the society’s own journal.
Conclusion
The conversion rate of the annual ISFRI and joint ISFRI and IAFR congresses was higher than other radiological societies, yet over 50 % of abstracts remains unpublished. Future research should explore reasons for non-publication to address these issues and increase the conversion rate.
国际法医放射学和成像学会（ISFRI）旨在通过其期刊《法医成像》（前身为《法医放射学和成像杂志》）和年度大会在全球范围内推进法医放射学和成像。衡量在这些大会上发表的研究的成功和质量的一种方法是评估摘要到出版物的转化率。目的评估摘要发表的百分比，并确定其发表的期刊。材料和方法确定2012 - 2022年ISFRI及ISFRI与国际法医放射技师协会（IAFR）联合大会的科学摘要发表率。在Pubmed、谷歌和谷歌Scholar中进行检索，然后人工查阅Journal of Forensic Radiology and Imaging和Forensic Imaging的目录。将结果与Scherer等人在Cochrane综述中的放射学和核成像以及法医研究的转化率进行比较，并辅以其他相关出版物。结果464篇摘要中，最终发表全文221篇（47.6%）。摘要以原创性研究279篇（60.1%）为主，其次是病例报告99篇（21.3%），综述79篇（17.0%）。摘要发表在各种各样的期刊上，但大多数（23.5%）发表在学会自己的期刊上。结论ISFRI年会及ISFRI与IAFR联合大会的转换率高于其他放射学会，但仍有超过50%的摘要未发表。未来的研究应该探索不发表的原因，以解决这些问题，提高转化率。
{"title":"Conversion to publication rate of abstracts presented at annual congresses of the International Society of Forensic Radiology and Imaging","authors":"G. de Heus , B.S. de Bakker , S.J. Decker , J. Elifritz , H. Hyodoh , F. Marttinen , C. O’Donnell , T.D. Ruder , M.J. Thali , R.R. van Rijn , ISFRI-IAFR congress consortium","doi":"10.1016/j.fri.2025.200661","DOIUrl":"10.1016/j.fri.2025.200661","url":null,"abstract":"<div><h3>Background</h3><div>The International Society of Forensic Radiology and Imaging (ISFRI) aims to advance forensic radiology and imaging worldwide through its journal <em>Forensic Imaging</em>, formerly <em>Journal of Forensic Radiology and Imaging</em>, and annual congresses. A way to measure success and quality of research presented at these congresses is by assessing the abstract to publication conversion rate.</div></div><div><h3>Objective</h3><div>To evaluate the percentage of abstracts that were published and identify the journals in which they appeared.</div></div><div><h3>Materials and methods</h3><div>The publication rate of scientific abstracts from ISFRI and joint ISFRI and International Association of Forensic Radiographers (IAFR) congresses between 2012 and 2022 was determined. Searches were conducted in Pubmed, Google and Google Scholar, followed by manual checks of the table of contents of <em>Journal of Forensic Radiology and Imaging</em> and <em>Forensic Imaging</em>. Results were compared with conversion rates from radiological and nuclear imaging, and forensic medical studies in the Cochrane review by Scherer et al., supplemented with other relevant publications.</div></div><div><h3>Results</h3><div>Of 464 presented abstracts, 221 (47.6 %) were eventually published as full articles.</div><div>Most abstracts were original research 279 (60.1 %), followed by 99 (21.3 %) case reports, and 79 (17.0 %) reviews. Abstracts were published in a wide variety of journals, but the majority 52 (23.5 %) were published in the society’s own journal.</div></div><div><h3>Conclusion</h3><div>The conversion rate of the annual ISFRI and joint ISFRI and IAFR congresses was higher than other radiological societies, yet over 50 % of abstracts remains unpublished. Future research should explore reasons for non-publication to address these issues and increase the conversion rate.</div></div>","PeriodicalId":40763,"journal":{"name":"Forensic Imaging","volume":"44 ","pages":"Article 200661"},"PeriodicalIF":1.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145748808","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2026-02-27DOI: 10.1016/j.fri.2026.200678
Annalisa N. Pedroni , Dominic Gascho , Andrea E. Steuer , Stephan A. Bolliger , Michael J. Thali
Introduction
Postmortem computed tomography (PMCT) has emerged as an increasingly important diagnostic tool in forensic medicine and has been used since 2015 to triage decedents delivered to the Zurich Institute of Forensic Medicine in order to more efficiently handle increasing caseloads. This retrospective study evaluated the use of quick toxicological screening methods (QT) in addition to PMCT for advanced postmortem triage.
Methods
All 128 triage and autopsy cases of the Zurich Institute of Forensic Medicine from the years 2019, 2021, and 2023 that received both PMCT and at least partial QT were analyzed using forensic radiology and toxicology reports as well as the final forensic report. QT included immunoassay for 10 substances, screening by untargeted liquid chromatography tandem mass spectrometry (LC-MS/MS), and blood alcohol analysis by headspace gas chromatography mass spectrometry (HD-GC-MS).
Results
Of all cases externally examined at the scene in 2023, 24% were further investigated and resolved with triage CT while 38.8% received an autopsy. Since 2015, this is a decline in autopsy numbers by 21.3% and a rise in triage CT by 88.1%, with an increase in external examinations by 16.2%. Of the cases receiving triage with CT and QT, 29.4% were further investigated with autopsy and 18.1% received standard toxicological analysis (STA). QT was able to identify intake of substances in 82.8% of cases and was positive in all 33 cases of intoxication. Cardiac, metabolic or infectious diseases were difficult to detect via triage with CT and QT.
Conclusion
QT is an important tool to confirm or exclude drug intake via triage, especially if there are signs of drug use at the scene or radiological signs indicating intoxication in PMCT. The combination of CT triage and QT may support workflow efficiency and informed decision-making within a forensic institution, particularly in high-caseload settings, by focusing investigations on the key forensic questions, ultimately benefitting the judicial authorities as well as the bereaved families.
{"title":"Virtopsy and Quick-Tox: Establishing a multimodal workflow for postmortem forensic examination","authors":"Annalisa N. Pedroni , Dominic Gascho , Andrea E. Steuer , Stephan A. Bolliger , Michael J. Thali","doi":"10.1016/j.fri.2026.200678","DOIUrl":"10.1016/j.fri.2026.200678","url":null,"abstract":"<div><h3>Introduction</h3><div>Postmortem computed tomography (PMCT) has emerged as an increasingly important diagnostic tool in forensic medicine and has been used since 2015 to triage decedents delivered to the Zurich Institute of Forensic Medicine in order to more efficiently handle increasing caseloads. This retrospective study evaluated the use of quick toxicological screening methods (QT) in addition to PMCT for advanced postmortem triage.</div></div><div><h3>Methods</h3><div>All 128 triage and autopsy cases of the Zurich Institute of Forensic Medicine from the years 2019, 2021, and 2023 that received both PMCT and at least partial QT were analyzed using forensic radiology and toxicology reports as well as the final forensic report. QT included immunoassay for 10 substances, screening by untargeted liquid chromatography tandem mass spectrometry (LC-MS/MS), and blood alcohol analysis by headspace gas chromatography mass spectrometry (HD-GC-MS).</div></div><div><h3>Results</h3><div>Of all cases externally examined at the scene in 2023, 24% were further investigated and resolved with triage CT while 38.8% received an autopsy. Since 2015, this is a decline in autopsy numbers by 21.3% and a rise in triage CT by 88.1%, with an increase in external examinations by 16.2%. Of the cases receiving triage with CT and QT, 29.4% were further investigated with autopsy and 18.1% received standard toxicological analysis (STA). QT was able to identify intake of substances in 82.8% of cases and was positive in all 33 cases of intoxication. Cardiac, metabolic or infectious diseases were difficult to detect via triage with CT and QT.</div></div><div><h3>Conclusion</h3><div>QT is an important tool to confirm or exclude drug intake via triage, especially if there are signs of drug use at the scene or radiological signs indicating intoxication in PMCT. The combination of CT triage and QT may support workflow efficiency and informed decision-making within a forensic institution, particularly in high-caseload settings, by focusing investigations on the key forensic questions, ultimately benefitting the judicial authorities as well as the bereaved families.</div></div>","PeriodicalId":40763,"journal":{"name":"Forensic Imaging","volume":"44 ","pages":"Article 200678"},"PeriodicalIF":1.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147394502","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Frontal sinuses are widely used in forensic human identification, and some studies have observed their contribution to sex estimation. This study aims to reproduce the methodology and validate in a new sample the logistic regression models developed by a previous study with a Brazilian subpopulation for evaluation of sexual dimorphism based on volumetric characteristics and parameters from 3D images of the frontal sinuses. The current study sample consisted of 130 CBCT scans of patients from two different areas of Brazil, of both sexes and aged 20 to 78 years. A trained examiner performed semi-automatic segmentation using ITK-SNAP program to obtain volume and analysis of 13 variables based on 3D images of sinuses in frontal, basal, and lateral views in Image J software. The area and volume variables presented significantly higher values for males. According to the receiver operating characteristic and area under the curve, the use of the significant variables presented in the logits of the previous study obtained accuracy of 65.39% for frontal and basal views, while accuracy reached 72.31% for the final model in this study. In the development of specific equations for the sample of our study, the highest accuracy in an isolated view was 74.62% for the frontal view equation and the lowest was 68.46% for the basal view, while the final model reached a 79.23% accuracy. The equations developed specifically for this sample achieved better accuracy values at estimating sex than the application of the equations with the significant variables of the previous study.
{"title":"Three-dimensional reconstruction of frontal sinus for sex estimation through cone beam computed tomography: validation and development of logits among Brazilians","authors":"Vanessa Moreira Andrade , Carlos Eduardo Raymundo , Deborah Queiroz Freitas , Luiz Francesquini Júnior","doi":"10.1016/j.fri.2026.200671","DOIUrl":"10.1016/j.fri.2026.200671","url":null,"abstract":"<div><div>Frontal sinuses are widely used in forensic human identification, and some studies have observed their contribution to sex estimation. This study aims to reproduce the methodology and validate in a new sample the logistic regression models developed by a previous study with a Brazilian subpopulation for evaluation of sexual dimorphism based on volumetric characteristics and parameters from 3D images of the frontal sinuses. The current study sample consisted of 130 CBCT scans of patients from two different areas of Brazil, of both sexes and aged 20 to 78 years. A trained examiner performed semi-automatic segmentation using ITK-SNAP program to obtain volume and analysis of 13 variables based on 3D images of sinuses in frontal, basal, and lateral views in Image J software. The area and volume variables presented significantly higher values for males. According to the receiver operating characteristic and area under the curve, the use of the significant variables presented in the logits of the previous study obtained accuracy of 65.39% for frontal and basal views, while accuracy reached 72.31% for the final model in this study. In the development of specific equations for the sample of our study, the highest accuracy in an isolated view was 74.62% for the frontal view equation and the lowest was 68.46% for the basal view, while the final model reached a 79.23% accuracy. The equations developed specifically for this sample achieved better accuracy values at estimating sex than the application of the equations with the significant variables of the previous study.</div></div>","PeriodicalId":40763,"journal":{"name":"Forensic Imaging","volume":"44 ","pages":"Article 200671"},"PeriodicalIF":1.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146037136","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2025-12-31DOI: 10.1016/j.fri.2025.200670
Ranim Bejaoui, Rim Mahouachi, Hela Mahersia
In the medical field, diagnostic accuracy increasingly depends on advanced imaging modalities such as X-ray, CT, MRI, ultrasound, and OCT for detecting both cancerous and non-cancerous anomalies. The integration of Artificial Intelligence (AI) has significantly enhanced the capabilities of Medical Image Analysis (MIA) in automating segmentation and classification tasks.
However, the wide variety of anomaly types and imaging modalities poses significant challenges in developing AI approaches that are both accurate and adaptable to real-world clinical conditions. Despite a growing body of research, there is a lack of a unified, comparative review that systematically examines how AI methods are applied across different anomaly types and processing stages.
In particular, the absence of a structured framework distinguishing workflows for cancerous versus non-cancerous diseases, across key stages such as data preprocessing, segmentation, and classification, limits the ability of researchers and practitioners to design optimized, disease-aware AI pipelines. Preprocessing steps such as denoising, contrast enhancement, and data cleaning are applied inconsistently, while segmentation is sometimes omitted or treated independently of classification, resulting in fragmented methodologies. These inconsistencies hinder performance comparisons and impede the development of robust, interpretable models.
To address these gaps, this work proposes a taxonomy to classify AI-based medical imaging approaches by anomaly type, imaging modality, and processing strategy. This taxonomy highlights methodological strengths and limitations and is complemented by a comparative analysis that examines workflow variations and the impact of each pipeline stage on diagnostic outcomes. The resulting framework offers a structured foundation to guide researchers, clinicians, and developers in creating scalable, interpretable, and clinically effective AI models for medical image analysis.
{"title":"A five-year systematic review of AI-based medical image analysis: From preprocessing to classification","authors":"Ranim Bejaoui, Rim Mahouachi, Hela Mahersia","doi":"10.1016/j.fri.2025.200670","DOIUrl":"10.1016/j.fri.2025.200670","url":null,"abstract":"<div><div>In the medical field, diagnostic accuracy increasingly depends on advanced imaging modalities such as X-ray, CT, MRI, ultrasound, and OCT for detecting both cancerous and non-cancerous anomalies. The integration of Artificial Intelligence (AI) has significantly enhanced the capabilities of Medical Image Analysis (MIA) in automating segmentation and classification tasks.</div><div>However, the wide variety of anomaly types and imaging modalities poses significant challenges in developing AI approaches that are both accurate and adaptable to real-world clinical conditions. Despite a growing body of research, there is a lack of a unified, comparative review that systematically examines how AI methods are applied across different anomaly types and processing stages.</div><div>In particular, the absence of a structured framework distinguishing workflows for cancerous versus non-cancerous diseases, across key stages such as data preprocessing, segmentation, and classification, limits the ability of researchers and practitioners to design optimized, disease-aware AI pipelines. Preprocessing steps such as denoising, contrast enhancement, and data cleaning are applied inconsistently, while segmentation is sometimes omitted or treated independently of classification, resulting in fragmented methodologies. These inconsistencies hinder performance comparisons and impede the development of robust, interpretable models.</div><div>To address these gaps, this work proposes a taxonomy to classify AI-based medical imaging approaches by anomaly type, imaging modality, and processing strategy. This taxonomy highlights methodological strengths and limitations and is complemented by a comparative analysis that examines workflow variations and the impact of each pipeline stage on diagnostic outcomes. The resulting framework offers a structured foundation to guide researchers, clinicians, and developers in creating scalable, interpretable, and clinically effective AI models for medical image analysis.</div></div>","PeriodicalId":40763,"journal":{"name":"Forensic Imaging","volume":"44 ","pages":"Article 200670"},"PeriodicalIF":1.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145938688","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Distinguishing between live births and stillbirths is substantially important in forensic autopsies. Lung-gastric flotation tests and histopathological examinations have been widely used for determining the diagnosis. In addition, postmortem computed tomography (PMCT) has recently been investigated as a potential diagnostic tool. In this report, we investigated the characteristic PMCT findings of stillbirths in three forensic autopsy cases. PMCT was performed in three male newborns (estimated gestational age: 7–9 months) with postmortem intervals of 10 hours (Case 2), 2 days (Case 1), and 1 week (Case 3). All cases comprised out-of-hospital deliveries: two infants were transported to the hospital after birth, while one was found deceased. In each case, the lungs and gastrointestinal tract sank during the flotation test, and no microscopic alveolar dilation was observed, leading to a diagnosis of stillbirth. The mean CT value of the lung field ranged between 42.1 and 49.0 Hounsfield units, indicating soft tissue-like density. In all cases, the ductus arteriosus appeared to be more dilated than did the left and right pulmonary arteries. No gas or fluid was detected in the main or lobar bronchi among all cases. This study is limited by the small number of cases with all being stillbirths. However, the imaging findings described in these cases of stillbirth suggest that PMCT might contribute to the differentiation between live births and stillbirths. In conclusion, the combination of high lung field density, ductus arteriosus patency, and absence of gas in the main and lobar bronchi on PMCT suggests a characteristic pattern of stillbirths.
{"title":"Characteristic findings of chest post-mortem computed tomography in stillbirths: A report of three cases","authors":"Hikaru Kuninaka , Momoka Tanabe , Noriko Ogawa , Moe Mukai , Ayako Nasu , Kazuho Maeda , Chiaki Fuke , Tsuneo Yamashiro , Yoko Ihama","doi":"10.1016/j.fri.2025.200665","DOIUrl":"10.1016/j.fri.2025.200665","url":null,"abstract":"<div><div>Distinguishing between live births and stillbirths is substantially important in forensic autopsies. Lung-gastric flotation tests and histopathological examinations have been widely used for determining the diagnosis. In addition, postmortem computed tomography (PMCT) has recently been investigated as a potential diagnostic tool. In this report, we investigated the characteristic PMCT findings of stillbirths in three forensic autopsy cases. PMCT was performed in three male newborns (estimated gestational age: 7–9 months) with postmortem intervals of 10 hours (Case 2), 2 days (Case 1), and 1 week (Case 3). All cases comprised out-of-hospital deliveries: two infants were transported to the hospital after birth, while one was found deceased. In each case, the lungs and gastrointestinal tract sank during the flotation test, and no microscopic alveolar dilation was observed, leading to a diagnosis of stillbirth. The mean CT value of the lung field ranged between 42.1 and 49.0 Hounsfield units, indicating soft tissue-like density. In all cases, the ductus arteriosus appeared to be more dilated than did the left and right pulmonary arteries. No gas or fluid was detected in the main or lobar bronchi among all cases. This study is limited by the small number of cases with all being stillbirths. However, the imaging findings described in these cases of stillbirth suggest that PMCT might contribute to the differentiation between live births and stillbirths. In conclusion, the combination of high lung field density, ductus arteriosus patency, and absence of gas in the main and lobar bronchi on PMCT suggests a characteristic pattern of stillbirths.</div></div>","PeriodicalId":40763,"journal":{"name":"Forensic Imaging","volume":"44 ","pages":"Article 200665"},"PeriodicalIF":1.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145748821","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2026-02-05DOI: 10.1016/j.fri.2026.200677
Samson Apostolakis , Till Sieberth , Sabine Franckenberg , Thomas Frauenfelder , Gary Hatch , Akos Dobay , Lars C. Ebert
Purpose
Computed tomography (CT) has transformed medical diagnostics by providing detailed two- and three-dimensional images of internal structures. However, interpreting CT data remains challenging and time-consuming due to the need for repeated image evaluations in various window settings. Adjusting window and level parameters not only consumes time but also complicates the analysis of pathologies involving multiple tissue types.
Method
This study introduces a novel AI-driven multi-tissue windowing (M-Win) technique for CT data visualization. The approach consists of two main steps: automatic segmentation using a convolutional neural network based on the U-Net architecture, followed by mapping the segmented structures to their respective window settings. The U-Net model was trained and validated on a dataset of postmortem CT scans covering the chest, abdomen, and pelvis. The algorithm assigns each voxel to a specific window based on standard Hounsfield unit (HU) ranges, translating HU values within a window to corresponding grayscale levels. Values below the window range appear black, while those above appear white, enabling simultaneous visualization of multiple tissue windows within a single image.
Discussion
M-Win holds promise for enhancing CT applications across various medical fields. Compared to existing techniques, the U-Net-based method offers advantages such as reduced window border artifacts. Although current limitations include segmentation artifacts and processing time, the potential benefits of streamlined and comprehensive CT analysis make M-Win a valuable addition to medical imaging. Future research will focus on clinical integration, real-time windowing capabilities, and performance comparisons with standard and alternative image review methods.
{"title":"All in one (and one for all) – U-Net driven multiwindow tissue visualization of computed tomography data","authors":"Samson Apostolakis , Till Sieberth , Sabine Franckenberg , Thomas Frauenfelder , Gary Hatch , Akos Dobay , Lars C. Ebert","doi":"10.1016/j.fri.2026.200677","DOIUrl":"10.1016/j.fri.2026.200677","url":null,"abstract":"<div><h3>Purpose</h3><div>Computed tomography (CT) has transformed medical diagnostics by providing detailed two- and three-dimensional images of internal structures. However, interpreting CT data remains challenging and time-consuming due to the need for repeated image evaluations in various window settings. Adjusting window and level parameters not only consumes time but also complicates the analysis of pathologies involving multiple tissue types.</div></div><div><h3>Method</h3><div>This study introduces a novel AI-driven multi-tissue windowing (M-Win) technique for CT data visualization. The approach consists of two main steps: automatic segmentation using a convolutional neural network based on the U-Net architecture, followed by mapping the segmented structures to their respective window settings. The U-Net model was trained and validated on a dataset of postmortem CT scans covering the chest, abdomen, and pelvis. The algorithm assigns each voxel to a specific window based on standard Hounsfield unit (HU) ranges, translating HU values within a window to corresponding grayscale levels. Values below the window range appear black, while those above appear white, enabling simultaneous visualization of multiple tissue windows within a single image.</div></div><div><h3>Discussion</h3><div>M-Win holds promise for enhancing CT applications across various medical fields. Compared to existing techniques, the U-Net-based method offers advantages such as reduced window border artifacts. Although current limitations include segmentation artifacts and processing time, the potential benefits of streamlined and comprehensive CT analysis make M-Win a valuable addition to medical imaging. Future research will focus on clinical integration, real-time windowing capabilities, and performance comparisons with standard and alternative image review methods.</div></div>","PeriodicalId":40763,"journal":{"name":"Forensic Imaging","volume":"44 ","pages":"Article 200677"},"PeriodicalIF":1.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147394491","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Forensic anthropology has become crucial for global cases involving firearm-related injuries. Although skeletal evidence is valuable, its presentation in court may re-traumatise families or bias reactions, influencing the verdict. To mitigate these risks, the adoption of 3D printing technologies in court has increased, allowing the handling of human remains without the drawbacks of presenting biological evidence. This study aimed to validate 3D-printed skeletal technologies as alternatives for courtroom evidence, focusing on cranial bullet defects through 3D topographical analysis and investigating the accuracy of 3D-printed skeletal models. Samples were scanned using micro-focus X-ray computed tomography; their 3D meshes were generated, 3D printed using powder Selective Laser Sintering, resin Stereolithography, and polylactic acid (PLA) filament Fused Deposition Modelling technologies, and scanned again. The reference and 3D print meshes were aligned, and a colour map allowed visual inspection of morphological discrepancies of up to 1 mm (0 mm shown in blue, 1 mm in red). Powder-based prints exhibited the highest accuracy for representing crania, predominantly coloured dark blue (0 mm). PLA filament prints were accurate for examining smaller cranial surfaces (mostly 0 mm), whereas resin prints were the least accurate for crania (mostly 0.5-1 mm). 3D-printed skeletal material can be utilised in legal settings with a colour map elucidating discrepancies. While powder-based prints are preferred, other materials may better suit specific applications. Further research should evaluate the impact of 3D prints on judicial decision-making and refine 3D printing techniques for forensic anthropology.
{"title":"Validating the morphology of 3D-printed cranial projectile trauma as a skeletal alternative for utilisation in a court of law","authors":"Claudia Ibáñez Martín , Ericka Noelle L'Abbé , Pieter Daniël de Wet , Alison Fany Ridel","doi":"10.1016/j.fri.2025.200667","DOIUrl":"10.1016/j.fri.2025.200667","url":null,"abstract":"<div><div>Forensic anthropology has become crucial for global cases involving firearm-related injuries. Although skeletal evidence is valuable, its presentation in court may re-traumatise families or bias reactions, influencing the verdict. To mitigate these risks, the adoption of 3D printing technologies in court has increased, allowing the handling of human remains without the drawbacks of presenting biological evidence. This study aimed to validate 3D-printed skeletal technologies as alternatives for courtroom evidence, focusing on cranial bullet defects through 3D topographical analysis and investigating the accuracy of 3D-printed skeletal models. Samples were scanned using micro-focus X-ray computed tomography; their 3D meshes were generated, 3D printed using powder Selective Laser Sintering, resin Stereolithography, and polylactic acid (PLA) filament Fused Deposition Modelling technologies, and scanned again. The reference and 3D print meshes were aligned, and a colour map allowed visual inspection of morphological discrepancies of up to 1 mm (0 mm shown in blue, 1 mm in red). Powder-based prints exhibited the highest accuracy for representing crania, predominantly coloured dark blue (0 mm). PLA filament prints were accurate for examining smaller cranial surfaces (mostly 0 mm), whereas resin prints were the least accurate for crania (mostly 0.5-1 mm). 3D-printed skeletal material can be utilised in legal settings with a colour map elucidating discrepancies. While powder-based prints are preferred, other materials may better suit specific applications. Further research should evaluate the impact of 3D prints on judicial decision-making and refine 3D printing techniques for forensic anthropology.</div></div>","PeriodicalId":40763,"journal":{"name":"Forensic Imaging","volume":"44 ","pages":"Article 200667"},"PeriodicalIF":1.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145840241","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-03-01Epub Date: 2025-12-19DOI: 10.1016/j.fri.2025.200664
Shafiya Qadeer Memon , James Martin Brown , Muhammad Moazzam Jawaid
Background:
The high mortality rate linked to coronary heart disease has driven intensive research for non-invasive diagnostic techniques. However, quantifying non-calcified plaques remains a significant challenge due to their low attenuation in computed tomography (CT) images, often resembling adjacent blood and myocardial tissues. Clinically, early detection and quantification of such plaques can aid physicians in preventing or delaying serious cardiac events by enabling timely intervention. Furthermore, accurate plaque assessment may help reduce unnecessary angioplasty procedures, thereby improving healthcare resource utilization. Accordingly, we propose a support vector machine (SVM)-based model for precise quantification of non-calcified plaques within the coronary vasculature.
Methods:
The primary indicator of non-calcified plaques in Computed Tomography Angiography is their relatively lower attenuation; therefore, we developed efficient discriminative features to capture these variations in a relative context. Initially, we calculated the vessel-wall thickness in normal arterial segments. Next, the vessel wall was excluded from the segmented coronary tree, and a Gaussian Mixture Model was applied to derive attenuation-based features such as posterior probabilities and fuzzy labels. In the final step, handcrafted features were employed to classify each voxel as either lumen or plaque.
Results:
The dataset includes 20 clinical CT scans from the Rotterdam database. For a volume-specific SVM model, the proposed method achieved mean sensitivity, specificity, and accuracy values of 92.40%, 83.70%, and 91.16%, respectively, using 10-fold cross-validation. These results demonstrate that the proposed quantification approach provides human-level accuracy in detecting non-calcified plaques.
{"title":"Non-calcified coronary plaque quantification in CT images using voxel-based descriptive features","authors":"Shafiya Qadeer Memon , James Martin Brown , Muhammad Moazzam Jawaid","doi":"10.1016/j.fri.2025.200664","DOIUrl":"10.1016/j.fri.2025.200664","url":null,"abstract":"<div><h3>Background:</h3><div>The high mortality rate linked to coronary heart disease has driven intensive research for non-invasive diagnostic techniques. However, quantifying non-calcified plaques remains a significant challenge due to their low attenuation in computed tomography (CT) images, often resembling adjacent blood and myocardial tissues. Clinically, early detection and quantification of such plaques can aid physicians in preventing or delaying serious cardiac events by enabling timely intervention. Furthermore, accurate plaque assessment may help reduce unnecessary angioplasty procedures, thereby improving healthcare resource utilization. Accordingly, we propose a support vector machine (SVM)-based model for precise quantification of non-calcified plaques within the coronary vasculature.</div></div><div><h3>Methods:</h3><div>The primary indicator of non-calcified plaques in Computed Tomography Angiography is their relatively lower attenuation; therefore, we developed efficient discriminative features to capture these variations in a relative context. Initially, we calculated the vessel-wall thickness in normal arterial segments. Next, the vessel wall was excluded from the segmented coronary tree, and a Gaussian Mixture Model was applied to derive attenuation-based features such as posterior probabilities and fuzzy labels. In the final step, handcrafted features were employed to classify each voxel as either lumen or plaque.</div></div><div><h3>Results:</h3><div>The dataset includes 20 clinical CT scans from the Rotterdam database. For a volume-specific SVM model, the proposed method achieved mean sensitivity, specificity, and accuracy values of 92.40%, 83.70%, and 91.16%, respectively, using 10-fold cross-validation. These results demonstrate that the proposed quantification approach provides human-level accuracy in detecting non-calcified plaques.</div></div>","PeriodicalId":40763,"journal":{"name":"Forensic Imaging","volume":"44 ","pages":"Article 200664"},"PeriodicalIF":1.0,"publicationDate":"2026-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145840240","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号