A first step towards a machine learning-based framework for bloodstain classification in forensic science

IF 2.2 3区医学 Q1 MEDICINE, LEGAL

Forensic science international Pub Date : 2024-10-31 DOI:10.1016/j.forsciint.2024.112278

Hyeonah Jung , Yeon-Soo Jo , Yoseop (Joseph) Ahn , Jaehoon (Paul) Jeong , Si-Keun Lim

{"title":"A first step towards a machine learning-based framework for bloodstain classification in forensic science","authors":"Hyeonah Jung , Yeon-Soo Jo , Yoseop (Joseph) Ahn , Jaehoon (Paul) Jeong , Si-Keun Lim","doi":"10.1016/j.forsciint.2024.112278","DOIUrl":null,"url":null,"abstract":"<div><div>Bloodstains found at a crime scene can help estimate the events that occurred during the crime. Reconstructing the crime scene by analyzing the bloodstain pattern contributes to understanding the bloody event. Therefore, it is essential to classify bloodstains through bloodstain pattern analysis (BPA) and accurately estimate the actions that took place at that time. In this study, we investigate the potential of using machine learning and deep learning to determine an action related to bloodstain data through the accessment of the corresponding bloodstain type by creating a prototype classification model. There are 14 types of bloodstain according to the classification system based on appearance. In this study, we test the classification potential of each bloodstain data for three bloodstain patterns such as Swing, Cessation, and Impact. Through experiments, it is shown that our prototype classification model for the selected bloodstains is developed and the accuracy of the resulting model is evaluated to be 80 %.</div></div>","PeriodicalId":12341,"journal":{"name":"Forensic science international","volume":"365 ","pages":"Article 112278"},"PeriodicalIF":2.2000,"publicationDate":"2024-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Forensic science international","FirstCategoryId":"3","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0379073824003608","RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"MEDICINE, LEGAL","Score":null,"Total":0}

引用次数: 0

Abstract

Bloodstains found at a crime scene can help estimate the events that occurred during the crime. Reconstructing the crime scene by analyzing the bloodstain pattern contributes to understanding the bloody event. Therefore, it is essential to classify bloodstains through bloodstain pattern analysis (BPA) and accurately estimate the actions that took place at that time. In this study, we investigate the potential of using machine learning and deep learning to determine an action related to bloodstain data through the accessment of the corresponding bloodstain type by creating a prototype classification model. There are 14 types of bloodstain according to the classification system based on appearance. In this study, we test the classification potential of each bloodstain data for three bloodstain patterns such as Swing, Cessation, and Impact. Through experiments, it is shown that our prototype classification model for the selected bloodstains is developed and the accuracy of the resulting model is evaluated to be 80 %.

查看原文本刊更多论文

基于机器学习的法医学血迹分类框架的第一步

在犯罪现场发现的血迹有助于估计犯罪过程中发生的事件。通过分析血迹模式重建犯罪现场有助于了解血腥事件。因此，必须通过血迹模式分析（BPA）对血迹进行分类，并准确估计当时发生的行为。在本研究中，我们通过创建一个原型分类模型，研究了使用机器学习和深度学习通过评估相应的血迹类型来确定与血迹数据相关的行动的潜力。根据基于外观的分类系统，血迹共有 14 种类型。在本研究中，我们测试了每种血迹数据对 Swing、Cessation 和 Impact 等三种血迹模式的分类潜力。实验结果表明，我们为所选血迹建立了原型分类模型，模型的准确率达到 80%。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Forensic science international 医学-医学：法

CiteScore

5.00

自引率

9.10%

发文量

285

审稿时长

49 days

期刊介绍： Forensic Science International is the flagship journal in the prestigious Forensic Science International family, publishing the most innovative, cutting-edge, and influential contributions across the forensic sciences. Fields include: forensic pathology and histochemistry, chemistry, biochemistry and toxicology, biology, serology, odontology, psychiatry, anthropology, digital forensics, the physical sciences, firearms, and document examination, as well as investigations of value to public health in its broadest sense, and the important marginal area where science and medicine interact with the law. The journal publishes: Case Reports Commentaries Letters to the Editor Original Research Papers (Regular Papers) Rapid Communications Review Articles Technical Notes.