Machine learning (ML)

It is a statistical technique for fitting models to data and learning by training models using data. It is a range of powerful computational algorithms capable of generating predictive models via intelligent autonomous analysis of relatively large and often unstructured data. This technology enables DNA analysis faster and more accurate.

Deep learning (DL)

Deep Learning, a subcategory of machine learning, is a deep neural network that has Specific configuration in which neurons are organized in several successive layers, which can independently learn representations of the data and progressively extract complex features to perform tasks such as computer vision and natural language processing (NPL) and is used in medicine to detect diseases from medical imaging.7

Natural language processing (NPL):-

It is a theoretically motivated range of computational techniques for analyzing and representing naturally occurring texts at one or more levels of linguistic analysis to achieve human-like language processing for a range of tasks or applications, being used in medicine to structure information in healthcare systems and extract relevant information from narrative texts to provide data for decision-making.8

Robotics

The Robot Institute of America defined it as “a reprogrammable multi functional manipulator designed to move material, parts, tools, or specialized devices through variable programmed motions for various tasks”. Virtobot, a Robot System for Optical 3D Scanning in Forensic Medicine, is a new technique for non-invasive autopsy.9

Artificial neural network (ANN)

This new term, “Artificial Neural Network,” is derived from biological neural networks that develop the structure of a human brain. Similar to the human brain, which has neurons interconnected to one another, artificial neural networks also have neurons that are interconnected to one another in various layers of the networks. It mimics the natural human neuron system.  It is a new data processing pattern described as a mathematic simulation system of the structure and function of the human brain.10

Convolutional neural networks (CNNs)

It is a regularized type of feed-forward neural network that learns feature engineering by itself via filter optimization. Using regularized weights over fewer connections prevents vanishing gradients and exploding gradients seen during backpropagation in earlier neural networks. CNNs make the explicit assumption that inputs have specific structures like images. This allows for the encoding of this property into the architecture by sharing the weights for each location in the image and having neurons respond only locally.11

Applications of AI

Many AI applications have been developed in the field of thanatology, especially for post-mortem identification, the calculation of the post-mortem interval.12, 13, and the establishment of the cause of death.14, 15, 16 In some clinical forensic medicine practices, AI applications are also helpful, such as age estimation.17, 18, 19, gender determination, 19, 20, and age estimation of bruises.21, 22, 23, 24 Bruises undergo color changes due to the breaking down of hemoglobin during the healing process. initially reddish, then violet-blue; on the third day, it becomes greenish due to biliverdin, then after 4–5 days, it turns brown due to bilirubin and later turns yellow due to the accumulation of hemosiderin pigment, disappearing in 10–14 days. These color changes were initially studied by naked eye examination, which is inaccurate. Now, the machines can detect such color changes more accurately, which is less time-consuming. It was also used in the assessment and management of violent behaviors among prisoners.24

Now, AI technologies are also applied in various fields of autopsies.25 This is very important because autopsy examination plays a vital role in criminal investigation. Virtopsy, which is also known as Virtual Autopsy, is noteworthy to be cited here. This non-invasive technology uses various imaging techniques such as CT scans, MRI, and 3D surface scanning to examine the body.26 These technologies enable thorough study of internal injury or pathological conditions without invasive dissection of the body.27

In cases of unknown or mutilated bodies, identification may be challenging. In the traditional methods for identification, we depend on different parameters like anthropology, facial descriptions, tattoos, scars, body marks, etc. In the newer era, we use fingerprints, iris scans, DNA analysis, etc. However, in this modern era, different AI technologies are introduced. Super human-like computers can be utilized for identification. These machines can process and store data such as facial features, gait, voice, retinal patterns, fingerprint patterns, etc.

In crime scene investigations too, AI is also very useful. Conventional methods, such as naked eye examination, microscopy, infrared examination, and various chemical tests, will take longer and sometimes may give erroneous reports, which may be due to inadequate samples. However, machines can detect such trace evidence more accurately within a shorter duration. In cases of sexual offenses, it may be challenging to detect the presence of sperm, especially when there is a very small sample. In such cases, the use of Convolutional neural networks trained by the VGG19 network and a variation of VGG19 with 1942 can fulfill this task. They can reduce the scanning time by locating the sperm on the microscope images. Detection and analysis of many trace evidence can be carried out by using AI. It is also very convenient to reconstruct the crime scene by creating video animation.28

In forensic ballistics, AI technologies are also now applied. When a bullet leaves a barrel, it carries microscopic evidence, which can be analyzed to establish the type of weapon used. It guides experts to the place where they need to look for gunpowder and cartridge tubes and compare the traces with a database through image processing without actual human involvement. Some algorithms allow for the highlighting of the residues resulting from firing with firearms, allowing for the detection of the explosion inside the barrel changes due to shock waves, as well as the provision of data that allow for the establishment of the class and caliber.

AI has future prospects in identifying various poisonous drugs and substances in the field of Forensic Toxicology. Helma C. et al. revealed in a scientific paper that there can be human errors by using the spectrophotometer, neutron, and high-performance liquid chromatography (HPLC), and in this sense, AI can play an essential role by providing a data set as a sample which will increase the precision of the method, the efficiency, and even the reduction in the costs of investigations.29

AI technologies are also applied to evaluate medical malpractice cases. This may be useful to a certain extent in the future investigation of medical negligence cases.