A systematic review on Artificial Intelligence for

Forensic sex and age estimation with diagnostic

imaging tools.
shama patyal
Lovely Professional University
Dr Tejasvi Pandey (  [email protected] )
Lovely Professional University

Research Article

Keywords: Artificial Intelligence, Machine Learning, Radiography, Magnetic resonance imaging,

Computerized tomography, Forensic Identification

Posted Date: February 2nd, 2023

DOI: https://doi.org/10.21203/rs.3.rs-2496015/v1

License:   This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License

Page 1/20
Abstract
Background: The article aims to study artificial intelligence and compare the results of different AI
algorithms in collaboration with radio diagnostic devices for age and sex estimation for forensic benefits.

Methods: Articles published between January 2012 and April 2022 were searched using different
databases. Twenty-six articles were selected based on inclusion and exclusion criteria. Prisma guidelines
were followed in the synthesis of this article.

Conclusions: Artificial intelligence (AI) is a technology that involves computerized algorithms to

dichotomize complex data. AI is widely used in diagnostic imaging to detect and quantify a clinical
condition. This systematic review aimed to explain the role of AI in the diagnostic imaging modality of
radiology in forensic Identification. AI technology is now widely used for age and sex estimation. Most of
the AI models are based on machine learning (ML) programs, artificial neural networks (ANN), and
convolutional neural networks (CNN). The results of the studies are promising, providing great accuracy
and decision-making. These AI-based models will act as identification tools in mass disaster and
medicolegal cases. In cooperation with ML algorithms can increase the identification of unknown
skeleton remains. Further improvement in AI programs and diagnostic tools is needed for better accuracy
and specificity in Forensic investigations. Realistic applications of these models are needed, and the
accuracy rate can be enhanced by comparing these models to different populations with wide sample
sizes. Most of the studies in this review paper were conducted on healthy subjects, studies on subjects
having developmental disorders should also be conducted for validation of these algorithms so that they
can be used in any scenario.

1. Background
Recent advancement in technologies changes the capabilities of different equipment. New introductions
of 3D, 4D and 5D technologies have transformed radiodiagnosis for a better approach in terms of more
detailed and clear images resulting in better patient care management. Radiography is a way of
generating optical images for scientific and medical research and is concerned with the inner structures
of the body which gives an observable look at the functions body’s interior tissues. It helps in the
recognition and treatment of disorders c[1]. To make it easy to recognize the abnormalities, this procedure
uses a reference data bank of the normal anatomy and functioning of the organs. This approach involves
conventional as well radioscopic imaging. It uses electromagnetic radiation like X-rays and gamma rays,
ultrasound, magnetic imaging, and thermal & isotope imaging. Other instruments are also used for the
documentation of knowledge regarding the location and the functions of the organs of the body. Medical
imaging, without insidious techniques, creates images of the human body’s internal structures. These
images which are produced are generated by arithmetically, logically, and fast processors. Those signals
are translated into digital images later. The various kinds of tissues inside the body are represented by
those signals [2]. Digital images, regularly, have a necessary function to play. Digital imaging processing
in the medical field refers to the managing of pictures with the help of a computer. There are several kinds

Page 2/20
of methods practiced for this processing, like image selection, storage, demonstration, and
communication. An image is a function that means the measurement of the characteristics of a viewed
sight, such as illumination or colour. The use of computers to manipulate a digital image is called the
image-processing technique [3]. This method has a lot of advantages, e.g., elasticity, adaptability, storage
of data and communication. Film-based radiology which includes computerized radiography, digital
radiography (flat-panel), computed tomography (CT), magnetic resonance imaging, digital
mammography, and digital fluoroscopy is now redundant and now has developed into various digital
imaging modalities (MRI), nuclear medicine, and surgical sonography for diagnosis. Thus, in radiology,
digital image processing is now a routine skill for radiologists. Moreover, in radiation treatment
preparation, the use of Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) scanners is
now an integral module of imaging [4]. With the help of various advanced techniques of image resizing,
the images can now be held efficiently. This approach has several sets of rules that can be synchronously
applied to the images. Two-dimensional and three-dimensional images can now be processed in different
dimensions. The development of techniques related to Image processing dates to the 1960s. These
techniques soon carved their niche in various fields, such as astronomy, medical purposes, and television
image enhancement. The cost of image processing came down and it became faster in the 1970s when
the advent of computer systems happened. The processing of images was much faster and cheaper in
the 2000s [5] . At a high stage, it may be claimed that artificial intelligence has been an intrinsic part of
imaging technology for several years. The Automatic Exposure System (AED), established in the 1980s,
was probably the first case in conventional radiographic method [6]. The radiographer might choose the
kV value for X-ray imaging, but the computer calculated when sufficiently quanta had entered the film to
create a diagnostic image, and hence the finishing mass of each coverage. Although this would not
diagnose or analyses photos, it did move an aspect of decision-making from the radiographer to the
computer, in the acceptance that the machine might render this decision more effectively than the
radiographer, helping both the company and the patient through removing duplicate images attributable
to inaccurate exposure and maximizing test dosing. Radiographers instantly embraced this technology
into their training because they saw the advantages it offered to patient care and image attainment
practice, mainly when patient body physiques obstructed the quality of the image. Though, owing to
structural variations and mistakes, there was also a need for human supervision. Digital image
manipulation is the benefit used by the radiologists and radiographers operating diagnostic modalities
[7]. Pre-processing relies on modality and correct devices anomalies such as the detection of differential
light, performance, pixels that are dead or dark noise. Millions of imaging procedures are used all over the
world. Medical imaging or radiography is rapidly developing due to recent advancements in techniques
for image processing. It includes image detection, analysis, enhancement etc. By image processing,
volume and percentage of the tissues detected are enhanced [8]. Radiology has come long way now
since its invention in the late 19th century and will continue to advance even further with upcoming
artificial technology (AI) which is in evolutionary phase in diagnostic imaging. In past different studies
have shown the specificity and accuracy of AI based models in detection of tumors and other several
lesions. Models based on AI and Machine Learning (ML) help in management of patient data and
detection of small radiographic abnormalities.
Page 3/20
Perception and Reasoning

Most of the functions give functionalities that help the reasoning and perception tasks. we can classify
these functionalities into seven types.

Segmentation: Many apps include a segmentation feature that allows them to designate a particular
organ. The segmentation not only relieves radiologists of this mission, but it also optimizes the
scarce attentional capital that radiologists have throughout their work and theoretically eliminates
both false-negative errors and false-positive by assisting them in concentrating on the most
important component and picture [9]. This is often accomplished by removing fewer interesting
facets of the picture to minimize knowledge saturation and thereby concentrate radiologists'
attention on the most significant aspects of the image (Figure 1).
Quantification and extraction of features: Numerous of the functions calculate specific
characteristics of the image, assess specific organ characteristics, or derive measurable information
through the image. The result is frequently portrayed in the form of facts and maps.
Highlighting the suspicious areas and Detecting: This functionality group concentrates on a certain
disease or deformity and searches for symptoms and underlines them. These functions are
frequently prepared for a specific (common) disease and seek to ensure correct picture analysis and
to assist radiologists in detecting and deciding on such problems.
Longitudinal analysis, cross-referencing and comparison: Some apps go a step further and offer
diagnostic insights by comparing various photographs of the same patient over time to identify
differences in those aspects. This role is often used to identify any related cases that have recently
been diagnosed, giving information into future appointments and evaluations by the radiologist.
Classifying abnormalities and Diagnosis: Diagnosis is a standard feature that draws on earlier
features but adds a conclusion about the probability of specific issues. This judgement can be made
by comparing to common/healthy expectations (e.g., normal brain size) and finding troublesome
areas. These applications differ in how often they build their outputs as "the real diagnosis" or "pre-
diagnosis" to be more investigated by physicians.
Prognosis: Few programmed have the potential to forecast the risk of specific illnesses or
complications based on an analysis of recent examinations. This prognosis is often based on a
specific issue and can provide relevant clinical details in addition to the information derived from the
picture.
Case prioritization, Patients profiling and synopsis: A collection of software does not perform
explicitly through images, but instead collects additional details regarding the patient from past
documentation and electronic medical documents and shows it next to each patient's image. They
are willing to do that because of the empirical insights they have. Radiologists to provide a broader
understanding of the patient's experience and symptoms, and hence to be more efficient. Precisely
analyses their photographs (Figure 2).

Radiology in Forensic
Page 4/20
The field of forensic science is vast, and it includes application of different other fields for forensic
benefits. The field of diagnostic radiology can be useful in Forensic Identification (age, sex, stature and
ethnicity) [10], identification of cause of death, type of injury, fatal or non-fatal wounds, to locate other
forensic evidence inside body, foreign bodies, criminal and civil court rooms trial and research [11]. Direct
comparison of antemortem and postmortem images plays a significant role in human identification.
Comparison of injuries, evaluation and detection of pattern depends upon the experience and knowledge
of radiological expert. The directions and locations of fractures from the digital images give the
information about type and shape of weapon. Postmortem CT-Scan as compare to conventional autopsy
is useful in advanced decomposed deceased bodies for detection of air embolism, pulmonary aspiration
of blood, traumas in case of children and detailed documentation of injuries and tissue damages [12].
With technology progress the diagnosis become fast and exact location of disease can be located so it
also provides lots of potential in forensic radiology. As forensic radiology is an emerging field it needs
extensive research, standardization with specific trainings to the radiologists. But with use of different
machine learning algorithms the output of forensic radiology and other medicolegal procedures can be
enhanced. In this systematic review article, we are going to study different automatic algorithms for age
and sex estimation using diagnostic images obtained from diagnostic radiology.

2. Main Text
Source

The PRISMA (preferred reporting items for systematic reviews) guidelines were followed for conducting
this systematic review article. The data for the present article is gathered from the databases which are
available free and do not require any institutional access. The article published in PubMed, Google
Scholar are mainly collected and from the other research engines like Open Science Directory, Free
Medical Journals, Directory of Open Access Journals and OpenMD.com that have been published
between (January 2012 to April 2022). The Boolean searches (AND OR NOT) were used while searching
the articles. The strategy of search mainly focusses on the articles that use the keywords like forensic
identification, artificial intelligence, machine learning, diagnostic radiology, forensic radiology, age
estimation, sex estimation, forensic medicine.

Study selection and data extraction

The articles were chosen based on their titles, and then a preliminary search was conducted using the
articles' abstracts. 186 items were chosen in the first round of selection. Using Mendeley software, 60
publications were eliminated based on duplication, and 55 articles connected to conference articles and
reviews that were unrelated to the topic of our study were also eliminated. In accordance with the
inclusion and exclusion criteria, 38 more records were deleted. After reading the article's complete text
and abstract, seven records were eliminated during the screening process.

Ethics reviews are not necessary for this review article and a procedure was not registered. The first two
authors discussed selection and extraction, and the second author took the lead whenever there was a
Page 5/20
discrepancy.

Criteria for Studies' Inclusion

• The papers must leverage technologies like AI and ML and must focus on forensic identification using
diagnostic imaging modalities.

• The articles that are fully searchable.

• The study's conclusion should be reached without any contention among the authors.

Exclusion the research' criteria

• Articles containing partial texts; articles available in languages other than English.

• Articles that do not concentrate on radiological imaging techniques.

Applying the aforementioned criteria further decreased the number of papers for our investigation to 26.
The papers were carefully examined, and for the qualitative extraction, a PRISMA flow chart (Figure 3)
was used. These publications were carefully examined for a quantitative assessment with the publication
year as a reference to understand the current trend of AI.

3. Results
A total 26 studies were selected for this review paper based on exclusion and inclusion criteria. The
articles mainly focus on the machine learning approach using diagnostic images as input material. Most
of the models were based on NN (neural networks), ANN (Artificial Neural Network) and CNN
(convolutional neural network). Bone age assessment was the main objective of majority of studies.
Digital radiographs scan was the commonly used modality of these studies (Fig. 4). MRI scans were also
used in 7 studies. Hand and wrist were the most common area of interest for age estimation whereas for
sex determination skull was the choice of most of the authors. Highly standardized data was used as
input in these AI models.

4. Discussions
AI approach for Age and Sex estimation

Sex and age estimation are the basic step of forensic identification. Sex estimation excludes half of the
population [13]. Similarly, estimation of age at time of causalities, earthquakes, genocide, mass-disasters
are of great importance. Although very limited studies confirm the use of AI and ML in forensic, the
selected studies in this review shown the efficiency and accuracy in predicting sex and age faster than
radiologists hence decreasing the reporting time [14].

Page 6/20
For this systematic review article, we have analyzed the studies that use AI based algorithms in forensic
perspective. Different training and test modules were generated with help of radiology modalities in these
studies which proved to be superior to human eyes and free from human biasness.

4.1 Age estimation

Age estimation for forensic and legal matters can be done by using established methods of radiology
which are based on visual examination of bone ossifications, skull sutures, pubic symphysis,
calcification, skeletal maturity and degenerative changes [15]. Bone age (BA) is a measure of skeletal
maturity. BA is also use for detecting abnormalities in development, monitoring growth hormone,
estimating height, detecting endocrinal disorders [16]. Hand radiographs are mostly used for assessing
BA as the hand radiographs are easy to perform and have multiple bones in a single radiograph and
exposure dose is also less as compared to other body parts. Two highly used method for age estimations
are Greulich- Pyle (GP) method and Tanner-Whitehouse (TW) method which are based on radiographs of
hand of children [17]. GP method is based on matching to reference images from the standard atlas.
Whereas TW method is based on scoring method. GP method is older than TW method and easy for use.
GP method was developed in 1917-1942 using x- ray images of Caucasian children. Whereas TW method
developed in 1950 using British children and further revised in 2001 using radiographs from different
ethnic populations and eliminating 20 bone score to 13 bone score. In adults age determination can be
done by studying developing dentition, degenerative changes of skeleton, fusion of epiphyses, deposition,
remodelling and bone resorptions. But all these methods are influenced by subjectivity of the examiner to
overcome this biasness more accurate and reliable methods are needed. Automated image analyses
methods were developed by different developers in past.

D. Giordano in 2015 examined 360 radiographs of hand [18]. For age range 0-6 years with equal number
of male and female radiographs. Hidden Markov Model (HMM) with a modified TW-2 method was used.
The mean error was reported 0.41±0.33 years with high rate of accuracy. The main drawback of the study
was limited age range where it can be applied.

A study conducted by Darko Stern and Martin Urschler in 2016 took the 3D images of hand from a
database of 132 MRI images and generate a regression algorithm using fusion development information
from different bones of hand and giving results in line with radiologists with 0.82±0.56 years of absolute
deviation from chronological age [19]. They choose Random Regression Forests (RRF) algorithms. RRF is
a non-linear approach that provide better overview features during trees construction. The method used
different fusion strategies and assumed hand bones to be localized, aligned and cropped.

In 2018 Darko Stern et al. conducted another study trained on 328 MRI images using Random Forests
and Deep Convolutional Methods (ML) using 3D Hand MRI images and obtaining mean absolute error of
0.37±0.51 years for the age range less than or equal to 18 years [20]. All the MRI examinations were done
on 3.0 T scanner with age range 13-25 years. The selected 141 subjects were equal or under 18 years old.
This multifactorial method was further tested to 2D images which were publicly available datasets of

Page 7/20
radiographs, and the results were like other automatic methods that are developed for x-ray images. The
positive attribute of this study was use of non-ionizing radiation.

One more article by Darko Stern et al. in 2019 based on deep convolutional neural network was based on
three input anatomical areas (hand, clavicle bone and tooth) [21]. The age range was between 13-25
years using MRI multi factorial nonionizing technique as taken in above study. They provide the idea of
fusing three different anatomical sites for age estimation. A total of 322 Caucasian male subjects’ data
was collected, 134 subjects were below 18 years. As per the authors this study was the first approach for
automatic evaluation of information of different anatomical areas on a wide range of age for age
estimation. The prediction error of 1.01±0.74 years was noted. The results also showed that regression-
based methods were more suitable for the study. But special attention is needed because of biological
variations within the same age range and resulting in wrong prediction of minor as adult and adult as
minor.

J.R.Kim et al. in 2017 used a program based on GP and deep learning for BA evaluation for age range 3-
17 years [22]. A total 200 radiographs of left hand were studied using three combination first was with
software alone, second was with software and two radiologist third combination was of radiologists and
GP atlas. The result of the study confirmed the increase in accuracy with the software. The increase in
efficiency and decrease in reading time was observed. The concordance range of software was 57.5%
and 72.5% for radiologists and 69.5% software alone whereas with GP it was 49.5% and 63% only. This
automatic system gave three estimated bone age values in order of percentage based on probability.
Certain limitations were observed in this study first was small sample size, second estimated bone age
differ from the actual chronological age in some cases. Third children below 2 years were not studied as
this software was based on GP atlas method.

J. H. Lee et al. used a deep convolutional neural network learning technique in 2018 where a regression
model was trained with help of training set of hand radiographs [23]. A deep learning tool named Caffe
was demonstrated. The mean absolute difference was 18.9 months and concordance correlation
coefficient were 0.78. The data for training was small due to which training and test performance value
differs.

T.D. Bui in 2019 gave a model based on TW-3 and deep convolutional networks to enhance the bone age
accuracy [24]. The method based on 6 regions of interests (ROI) from TW-3. The methodology was based
on two phases first was ROI detection and second was ROI classification using regression networks. The
method showed that TW-3 based deep CNN overcomes the limitations of GP based deep CNN. For better
future study implication of the whole image analysis should be included.

A study conducted in 2021 by C V Pham et al. estimated the age of adults, they investigated 814 adults
within the age range of 20-70 years. It was a fully automated approach using PMCT scans as input and
predicting age as output by studying two bones mandible and femur with using a model based on 3D
CNN [25]. The result showed that femur alone was better as age indicator than mandible alone. On
combining mandible and femur the accuracy further improved and mean absolute error was decreased.
Page 8/20
Most of the models discussed above were for children but this model estimates the age in adults and
modality of choice was also different from above studies. Further validation of this model on different
ethnicity is required and segmentation of other age indicator can be segmented together to get more
precise and accurate BA.

David B Larson et al. concluded that DL CNN model can estimate age with accuracy similar to the
radiologist with very less reading time [26]. The model used hand x-ray images as input and the
performance was measured in terms of root mean square (RMS) and mean absolute difference which
was 0.63 and 0.50 years respectively and RMS for Digital Hand Atlas data set was 0.73 years.

Another study published in 2017 by Hyunkwang Lee also used CNN to develop fully automated program
to segment ROI and pre-process input radiographs and perform BA [27]. Hands radiograph of age group
5-18 years were selected. Interpretation time was less than or equal to 2 seconds and gave accuracy of
61.4% in females and 57.32% in males. Although results are promising for future but accuracy rate in this
study was less as compared to above studies. An article published in 2018 was also CNN based by S.H.
Tajmir age range 5-18 years was selected for the study. Accuracy rate was 68.2% root mean square error
was 0.548 years and 98.6% within 1 year [28]. The combined accuracy rate of AI and radiologist was high
as compared to AI alone, radiologist alone.

A fully automatic 2-D knee segmentation based on MRI 3-D images and CNN was developed by Paul-
Louis Prove et al. in 2019. Multiple pre-processing steps were involved for correcting image intensity to
reduce size of image [29]. Initial results gave MAE (mean absolute error) of 0.48±0.32 years for test set of
14 subjects. The proposed method will provide reproducible and faster BA in future.

In 2020 an article published by author Markus et al. used 3-D MRI knee scans and CNN tree-based ML
program for age estimation in the age group 13-21 years old in Caucasian population, 404 sagittal and
185 coronal MRI images were studied [30]. The methodology was based on three main steps- Image pre-
processing, bone segmentation and final Age estimation. MAE was 0.67 ± 0.49 years with 90.9%
accuracy was observed. 88.6% was the sensitivity and 94.2% was the specificity. The results of the study
are promising for future implications, but a large training data set is required.

Another study conducted by Fuk Hay Tang et al. based on MRI scans was published in 2018. ANN and
TW3 program were used on 79 subjects having age range 12-17 years on Chinese population [31]. MRI of
left hand and wrist were used as input material the sections were in coronal view. Independent indicators
like height and weight of subjects, intensity and composition of bone marrow quantified by MRI were also
used as input. This ML approach was 10-fold more accurate than TW3 method alone.

A study demonstrated the accuracy and efficiency of BoneXpert 2.1 in BA irrespective of gender. Less
reading time was also its quality as compared to GP method. 514 patients in the age range 3-17 years
were selected among the German population. Left hand and wrist radiograph were studied using AI by
Christian Booz et al. the correlation between AI and reference was significantly higher (r=0.99) as
compare to other method (r=0.90) and mean reading time was also reduced to 87% [32].
Page 9/20
Spampinato studied age range 0-18 years age group for automatic BA. A CNN network BoNet was used
to assess BA on publicly available dataset comprising different races, gender and age range [33]. DL
techniques were applied, and results showed an average difference of 0.8 years between manual and
automatic evaluation. The authors tested existing CNN like Google Net, Oxford Net and Over Feat which
were already pretrained on a dataset of 1400 radiographs. The authors stated that many convolutional
layers do not mean high performance always. In their study only 5 layers gave the desired performance.

4.2 Sex determination

Skeletal bones play a key role in gender estimation. Gender identification is very crucial whenever skeletal
remains are found for medicolegal and courtroom purpose. The skeleton bones play the major role in sex
determination and with help of radiology modalities the sex identification become easier. Skull and pelvis
play a vital role in gender identification, the shape and size of the bones are different in male and females
that is the common criteria to differentiate male and female skeleton [34]. Sex prediction is a crucial part
for a forensic expert in making biological profile. Different biomarkers are studied by the anthropologist in
estimating sex [35]. In literature almost all the bones were studied, and accuracy was compared among
different populations. Femur, patella, wrist, clavicle, sternum, mandible, metatarsals are used in sex
predictions [15, 36] . Costal cartilage calcification was also studied in past for estimating sex [37, 38].
With advancement in AI technology the rate of accuracy and specificity in predicting sex is increased and
that is shown by number of studies discussed here.

Mumtaz A. Kaloi in 2018 proposed a network based on CNN using radiographs of left hand of children in
age range 1 month to 18 years old for gender determination [39]. The accuracy rate was 98%, Class
Activation Mapping determined that carpals, ulna, and radius were more specific for gender determination
as compared to upper hand bones.

Wen Yang et al. in 2019 proposed a BPNN (backpropagation neural network) which was an improved
version by using skull [40]. A total of Chinese 267 skulls from whole skull database of CT scan were
studied out of which 153 were of females with age range 18-88 years. Six parameters were used as input
to get the desired result, for improving generalization ability Adaboost algorithm were used. The accuracy
rate was 96.76% with 0.01 mean square error.

Yongjie Cao et al. developed CNN models based on pelvic regions and finally compared these models
with two forensic anthropologists using morphological methods [41]. Dataset of 862 subjects were
divided for training, validation and testing. CNN model was developed for ventral pubis, dorsal pubis,
greater sciatic notch, pelvic inlet, ischium and acetabulum. Except last two prediction metrics were over
0.9. rate of accuracy was higher than the two experts. The accuracy rate was 98.2% for Pelvic Inlet.

A deep learning ANN was trained on 900 virtual skulls constructed from CT images [42]. The proposed
model accuracy rate was 95% the author James Bewes et al. stated that once this program is trained it is
rapid to use and abolish the human biasness. 500 subjects of each sex were selected in the age range of
18-60 years. The CT images were collected from the PACS database. The study population mainly
Page 10/20
consisted of European population and three non-European population was also observed (Chinese,
Indian, Vietnamese). Distinct ethnicity was a positive aspect as compared to other previous studies.
GoogleNet a CNN network was adapted for the study and transfer learning was performed with help of
MATLAB. The final three layers of neural network was modified to classify image as ‘male or female’ at
the input level.

A unique study based on grey matter of brain was conducted by Nathaniel E. Anderson et al. was
published in 2018. MRI scans were used of 1300 subjects in the age range of 12-66 years for training and
testing the model based on ML [43]. The accuracy rate was greater than 93%. Orbitofrontal and
frontopolar regions were seen to be larger in females and anterior medial temporal portion larger in
males. They also replicate the same results of accuracy rate in a healthy control sample. MRI based
methods are being developed but require more research with different skeleton structures.

A retrospective study based on Iranian population was conducted in the age group of 18-70 years
(Maryam Farhadian et al.). Cone beam CT images of 190 subjects were collected of mastoid region [44].
The measurements of mastoid process were conducted by using different data mining algorithms. Total
9 landmarks were measured. Different models were compared using cross validation. But the random
forest model was the best giving 97% accuracy rate and ANN model was the least giving 84% accuracy,
for rest 5 models all showed accuracy greater than 90%. The inter mastoid distance and the distance
between the most prominent convex mastoid point played the largest role in sex estimation. The author
concluded that finding best method for different dataset is challenging and require widespread
investigation and further studies on different populations are required.

A research paper was published in year 2015 by M.F. Darmawan et al. This paper analysed the bones of
left hand of Asian children for sex estimation with three different models- Discriminant Function Analysis
(DFA), Support Vector Machine (SVM) and ANN [45]. 333 x-ray images were taken of children below 19
years. Nineteen hand bones were studied using Free Image Software and MATLAB. The age group 16-19
(96.6% accuracy) and 7-9 (80.49%accuracy) were considered best for sex estimation as the average
accuracy rate for these groups was more than 80%. The ANN model was superior to other two models in
terms of accuracy rate in the (above mentioned age group) findings of this study. The results of the study
showed that each model has best accuracy rate on each different age group.

Similarly, another ANN-based study where sex determine from the shape of calvarium was conducted
with help of 1700 lateral CT scans with age range 25-92 years [46]. The sample was consisted of
Caucasian male and female (850 each). The result of the study yields an accuracy rate higher than 80%.
The author (Fabio Cavalli et al.) suggested the use of such model on other bones to improve the
susceptibility of the proposed methodology.

ML based study using different algorithms like ANN, SVM and logistic regression (LR) was conducted to
generate models for sex estimation based on measurement of 47 landmarks on cranium [47]. 393
Bulgarian adult’s 3D CT images were selected. The accuracy of models was evaluated by 10×10-fold
cross validation and the accuracy rate was greater than 95%. The best accuracy result was of SVM the
Page 11/20
authors Diana Toneva et al. suggested the use of such model on other skeleton bones of the body for
forensic scope.

Sehyo Yune et al. in 2018 developed a CNN based model for predicting sex on the basis of radiographs of
hand and wrist in a sample size of 1531 with age range 5-70 years [48] . The results were than compare
with the radiologists, but the accuracy rate of the model was quite high as compared to the radiologists.
Hence showing that DL can be used to identify patterns where human perception fails. The accuracy rate
was 95.9% for the CNN model and 58% for the two radiologists. Heat maps were generated with help of
class activation maps (CAM) and showed that model focused mainly on the 2nd and 3rd metacarpal
base or thumb sesamoid in women, and distal radioulnar joint, distal radial physis and epiphysis, or 3rd
metacarpophalangeal joint in men. The model predicted 77.8% of accuracy when done on the sample
which was sexually disorder/transgender.

We observed four major limitations in the above discussed studies: 1. Selected gender and age group
were evaluated in above studies that was also with in very narrow range. 2. Training of medical experts is
needed for visualizing and interpreting results of different deep learning techniques. 3. A realistic
applications of these models in forensic setup is required for its futuristic implication in court room trials.
4. The output of the models depend upon the input image scans and if the quality of scans is not good
enough it might not be accepted by the model.

5. Conclusions
In this review paper we analysed AI applications using different image modalities for age and sex
estimation in forensic context for human Identification. AI technology is successfully applied in many
challenging tasks of health care and diagnostic imaging. One of the benefits of AI models is that they are
free from human biasness and fast in giving results with accuracy rate similar or greater than the human
experts. The artificial neural networks provide the benefit in discovering the useful features of an image
and discarding unhelpful image’s part automatically. Neural Networks are very fast once trained and
require only milliseconds to identify skeleton structures. Machine Learning have the potential to improve
themselves by increasing more data there is not any upper limit for the volume of input data hence
millions of images of skeletal structure can be used to train a neural network for better analyzation. The
radiographic techniques which are used in alive individuals have the drawback of radiation hazards, but it
can be overcome by using non-ionizing radiation modalities (like MRI, USG). Realistic applications of
these models are needed, and accuracy rate can be enhanced by comparing these models on different
populations with wide sample sizes. Most of the studies in this review paper were conducted on healthy
subjects, studies on subjects having developmental disorders should also be conducted for validation of
these algorithms so that they can be use in any scenario. In unidentified cases radiographic techniques
act as cheap and fast mode of identification. Utilization of AI models can increase the speed as well as
performance of forensic anthropologists and radiologists.

Abbreviations
Page 12/20
AI: Artificial intelligence

BA: Bone age

CT: Computed Tomography

ML: Machine Learning

MRI: Magnetic Resonance Imaging

ANN: Artificial neural networks

CNN: Convolutional neural networks

LR: logistic regression

SVM: Support Vector Machine

TW: Tanner-Whitehouse

Declarations
Ethics approval and consent to participate– “Not applicable”
Consent for publication- “Not applicable”
Availability of data and material- “The datasets used and/or analysed during the current study
available from the corresponding author on reasonable request.”
Competing interests- The authors declare that they have no competing interests.
Funding- “Not applicable”
Authors' contributions- "SP collected the data, compiling and formatting for the article. TP and SP
analyzed and interpreted the articles. TP participated in its design and coordination and helped to
draft the manuscript. SP contributed to writing the manuscript under the guidance of TP. All authors
read and approved the final manuscript."
Acknowledgements- "Not applicable"

References
1. MacMahon H, Doi K (1991) Digital Chest Radiography. Clin Chest Med 12:19–32.
https://doi.org/10.1016/S0272-5231(21)00724-3
2. Audette MA, Ferrie FP, Peters TM (2000) An algorithmic overview of surface registration techniques
for medical imaging. Med Image Anal 4:201–217. https://doi.org/10.1016/S1361-8415(00)00014-1
3. Doi K (2005) Current status and future potential of computer-aided diagnosis in medical imaging. Br
J Radiol 78:s3–s19. https://doi.org/10.1259/bjr/82933343

Page 13/20
 4. Higaki T, Nakamura Y, Tatsugami F, et al (2019) Improvement of image quality at CT and MRI using
deep learning. Jpn J Radiol 37:73–80. https://doi.org/10.1007/s11604-018-0796-2
5. Aggarwal A, Mittal M, Battineni G (2021) Generative adversarial network: An overview of theory and
applications. Int J Inf Manag Data Insights 1:100004. https://doi.org/10.1016/j.jjimei.2020.100004
6. Roguin A, Schwitter J, Vahlhaus C, et al (2008) Magnetic resonance imaging in individuals with
cardiovascular implantable electronic devices. Europace 10:336–346.
https://doi.org/10.1093/europace/eun021
7. Nair MK, Nair UP (2007) Digital and Advanced Imaging in Endodontics: A Review. J Endod 33:1–6.
https://doi.org/10.1016/j.joen.2006.08.013
8. Kumar S, Bhandari AK, Raj A, Swaraj K (2021) Triple Clipped Histogram-Based Medical Image
Enhancement Using Spatial Frequency. IEEE Trans Nanobioscience 20:278–286.
https://doi.org/10.1109/TNB.2021.3064077
9. Doi K (2007) Computer-aided diagnosis in medical imaging: Historical review, current status and
future potential. Comput Med Imaging Graph 31:198–211.
https://doi.org/10.1016/j.compmedimag.2007.02.002
10. Widek T, De Tobel J, Ehammer T, Genet P (2022) Forensic age estimation in males by MRI based on
the medial epiphysis of the clavicle. Int J Legal Med. https://doi.org/10.1007/s00414-022-02924-9
11. Kira K, Chiba F, Makino Y, et al (2022) Stature estimation by semi-automatic measurements of 3D CT
images of the femur. Int J Legal Med. https://doi.org/10.1007/s00414-022-02921-y
12. Decker SJ, Braileanu M, Dey C, et al (2019) Forensic Radiology: A Primer. Acad Radiol 26:820–830.
https://doi.org/10.1016/j.acra.2019.03.006
13. Akhlaghi M, Sheikhazadi A, Naghsh A, Dorvashi G (2010) Identification of sex in Iranian population
using patella dimensions. J Forensic Leg Med 17:150–155.
https://doi.org/10.1016/j.jflm.2009.11.005
14. Patyal Shama, Tejasvi B (2021) Artificial Intelligence with Radio-Diagnostic Modalities in Forensic
Science-A Systematic Review. CEUR Workshop Proc 3058:
15. Scheuer L (2002) Application of osteology to forensic medicine. Clin Anat 15:297–312.
https://doi.org/10.1002/ca.10028
16. Haymond M, Kappelgaard A-M, Czernichow P, et al (2013) Early recognition of growth abnormalities
permitting early intervention. Acta Paediatr 102:787–796. https://doi.org/10.1111/apa.12266
17. Urschler M, Krauskopf A, Widek T, et al (2016) Applicability of Greulich–Pyle and Tanner–Whitehouse
grading methods to MRI when assessing hand bone age in forensic age estimation: A pilot study.
Forensic Sci Int 266:281–288. https://doi.org/10.1016/j.forsciint.2016.06.016
18. Giordano D, Kavasidis I, Spampinato C (2016) Modeling skeletal bone development with hidden
Markov models. Comput Methods Programs Biomed 124:138–147.
https://doi.org/10.1016/j.cmpb.2015.10.012

Page 14/20
19. Stern D, Ursekter M (2016) From individual hand bone age estimates to fully automated age
estimation via learning-based information fusion. Proc - Int Symp Biomed Imaging 2016-June:150–
154. https://doi.org/10.1109/ISBI.2016.7493232
20. Štern D, Payer C, Urschler M (2019) Automated age estimation from MRI volumes of the hand. Med
Image Anal 58:. https://doi.org/10.1016/j.media.2019.101538
21. Štern D, Payer C, Giuliani N, Urschler M (2019) Automatic Age Estimation and Majority Age
Classification from Multi-Factorial MRI Data. IEEE J Biomed Heal Informatics 23:1392–1403.
https://doi.org/10.1109/JBHI.2018.2869606
22. Kim JR, Shim WH, Yoon HM, et al (2017) Computerized bone age estimation using deep learning-
based program: Evaluation of the accuracy and efficiency. Am J Roentgenol 209:1374–1380.
https://doi.org/10.2214/AJR.17.18224
23. Lee JH, Kim KG (2018) Applying deep learning in medical images: The case of bone age estimation.
Healthc Inform Res 24:86–92. https://doi.org/10.4258/hir.2018.24.1.86
24. Bui TD, Lee JJ, Shin J (2019) Incorporated region detection and classification using deep
convolutional networks for bone age assessment. Artif Intell Med 97:1–8.
https://doi.org/10.1016/j.artmed.2019.04.005
25. Pham C Van, Lee SJ, Kim SY, et al (2021) Age estimation based on 3D post-mortem computed
tomography images of mandible and femur using convolutional neural networks. PLoS One 16:.
https://doi.org/10.1371/JOURNAL.PONE.0251388
26. Larson DB, Chen MC, Lungren MP, et al (2018) Performance of a deep-learning neural network model
in assessing skeletal maturity on pediatric hand radiographs. Radiology 287:313–322.
https://doi.org/10.1148/radiol.2017170236
27. Lee H, Tajmir S, Lee J, et al (2017) Fully Automated Deep Learning System for Bone Age
Assessment. J Digit Imaging 30:427–441. https://doi.org/10.1007/s10278-017-9955-8
28. Tajmir SH, Lee H, Shailam R, et al (2019) Artificial intelligence-assisted interpretation of bone age
radiographs improves accuracy and decreases variability. Skeletal Radiol 48:275–283.
https://doi.org/10.1007/s00256-018-3033-2
29. Pröve PL, Jopp-van Well E, Stanczus B, et al (2019) Automated segmentation of the knee for age
assessment in 3D MR images using convolutional neural networks. Int J Legal Med 133:1191–1205.
https://doi.org/10.1007/s00414-018-1953-y
30. Mauer MA der, Well EJ van, Herrmann J, et al (2021) Automated age estimation of young individuals
based on 3D knee MRI using deep learning. Int J Legal Med 135:649–663.
https://doi.org/10.1007/s00414-020-02465-z
31. Tang FH, Chan JLC, Chan BKL (2019) Accurate Age Determination for Adolescents Using Magnetic
Resonance Imaging of the Hand and Wrist with an Artificial Neural Network-Based Approach. J Digit
Imaging 32:283–289. https://doi.org/10.1007/s10278-018-0135-2
32. Booz C, Yel I, Wichmann JL, et al (2020) Artificial intelligence in bone age assessment: accuracy and
efficiency of a novel fully automated algorithm compared to the Greulich-Pyle method. Eur Radiol
Page 15/20
 Exp 4:. https://doi.org/10.1186/s41747-019-0139-9
33. Spampinato C, Palazzo S, Giordano D, et al (2017) Deep learning for automated skeletal bone age
assessment in X-ray images. Med Image Anal 36:41–51.
https://doi.org/10.1016/j.media.2016.10.010
34. Ramamoorthy B, Pai MM, Prabhu L V., et al (2016) Assessment of craniometric traits in South Indian
dry skulls for sex determination. J Forensic Leg Med 37:8–14.
https://doi.org/10.1016/j.jflm.2015.10.001
35. Drusini AG, Toso O, Ranzato C (1997) The coronal pulp cavity index: A biomarker for age
determination in human adults. Am J Phys Anthropol 103:353–363.
https://doi.org/10.1002/(SICI)1096-8644(199707)103:3<353::AID-AJPA5>3.0.CO;2-R
36. Mastrangelo P, De Luca S, Sánchez-Mejorada G (2011) Sex assessment from carpals bones:
Discriminant function analysis in a contemporary Mexican sample. Forensic Sci Int 209:196.e1-
196.e15. https://doi.org/10.1016/j.forsciint.2011.04.019
37. Middleham HP, Boyd LE, McDonald SW (2015) Sex determination from calcification of costal
cartilages in a Scottish sample. Clin Anat 28:888–895. https://doi.org/10.1002/ca.22491
38. Patyal S, Bhatia T (2022) Evaluating costal cartilage for sex and age estimation from PA chest
radiographs of North Indian population: a retrospective study. Egypt J Forensic Sci 12:41.
https://doi.org/10.1186/s41935-022-00298-y
39. Kaloi MA, He K (2018) Child Gender Determination with Convolutional Neural Networks on Hand
Radio-Graphs. 1–8
40. Yang W, Liu X, Wang K, et al (2019) Sex determination of three-dimensional skull based on improved
backpropagation neural network. Comput Math Methods Med 2019:.
https://doi.org/10.1155/2019/9163547
41. Cao Y, Ma Y, Yang X, et al (2021) Use of deep learning in forensic sex estimation of virtual pelvic
models from the Han population. Forensic Sci Res 0:1–10.
https://doi.org/10.1080/20961790.2021.2024369
42. Bewes J, Low A, Morphett A, et al (2019) Artificial intelligence for sex determination of skeletal
remains: Application of a deep learning artificial neural network to human skulls. J Forensic Leg Med
62:40–43. https://doi.org/10.1016/j.jflm.2019.01.004
43. Anderson NE, Harenski KA, Harenski CL, et al (2019) Machine learning of brain gray matter
differentiates sex in a large forensic sample. Hum Brain Mapp 40:1496–1506.
https://doi.org/10.1002/hbm.24462
44. Farhadian M, Salemi F, Shokri A, et al (2020) Comparison of data mining algorithms for sex
determination based on mastoid process measurements using cone-beam computed tomography.
Imaging Sci Dent 50:323. https://doi.org/10.5624/isd.2020.50.4.323
45. Darmawan MF, Yusuf SM, Abdul Kadir MR, Haron H (2015) Comparison on three classification
techniques for sex estimation from the bone length of Asian children below 19 years old: An analysis

Page 16/20
 using different group of ages. Forensic Sci Int 247:130.e1-130.e11.
https://doi.org/10.1016/j.forsciint.2014.11.007
46. Cavalli F, Lusnig L, Trentin E (2017) Use of pattern recognition and neural networks for non-metric
sex diagnosis from lateral shape of calvarium: an innovative model for computer-aided diagnosis in
forensic and physical anthropology. Int J Legal Med 131:823–833. https://doi.org/10.1007/s00414-
016-1439-8
47. Toneva D, Nikolova S, Agre G, et al (2021) Machine learning approaches for sex estimation using
cranial measurements. Int J Legal Med 135:951–966. https://doi.org/10.1007/s00414-020-02460-4
48. Yune S, Lee H, Kim M, et al (2019) Beyond Human Perception: Sexual Dimorphism in Hand and Wrist
Radiographs Is Discernible by a Deep Learning Model. J Digit Imaging 32:665–671.
https://doi.org/10.1007/s10278-018-0148-x

Figures

Figure 1

AI functionalities related to perception and reasoning.

Page 17/20
Figure 2

Share of AI Functionalities.

Page 18/20
Figure 3

Flow chart for Screening and selection of articles.

Page 19/20
Figure 4

Choice of modality in selected studies

Page 20/20