Artificial intelligence and computer vision in forensic sciences

Abstract

Abstract Background The integration of artificial intelligence (AI) and computer vision (CV) into forensic sciences has transformed the analysis of violence-related evidence, improving precision, objectivity and efficiency across various forensic applications. Objective This systematic review evaluates current AI and CV applications specifically focusing on violence-related forensic evidence analysis, highlighting technological advancements, implementation challenges and future directions. Material and methods We conducted a comprehensive search across PubMed, Scopus and Web of Science (2020–2025) using MeSH terms and keywords related to AI, CV and forensic science. After excluding nonhuman studies, reviews and non-English publications, 206 initial records were screened using the ASReview software. Through dual researcher screening and supplemental expert consultation, we identified 21 eligible studies focusing on AI-driven injury detection and diagnosis in forensic contexts. Results A total of 21 studies demonstrated AI applications across 6 forensic domains: (1) wound/injury classification; (2) head/brain injury; (3) bone fractures; (4) process enhancement and reconstruction; (5) injury degree appraisal; and (6) physical abuse. These areas cover applications such as automated detection of injuries, toolmark analysis and time of injury estimation. Key limitations included reliance on simulated datasets, class imbalances and limited real-world validation. Conclusion The use of AI and CV technologies offers significant advancements in forensic science, particularly in the objective evaluation of trauma-related evidence. Further development of generalizable models, along with standardized datasets and validation protocols, is essential to ensure their integration into routine forensic practice.

Ähnliche Arbeiten