Review Article

Artificial Intelligence Application in Bone Fracture Detection

Ahmed AlGhaithi, Sultan Al Maskari
Division of Orthopedic, Department of Surgery, Sultan Qaboos University Hospital, Muscat, Oman

Abstract
The interest of researchers, clinicians, and industry in artificial intelligence (AI) continues to grow, especially with recent deep‑learning (DL)
advances. Recent published reports have shown the utility of DL for bone fracture diagnosis in the radiological examination. It is important
for practicing physicians to recognize the current scope of DL as it may impact the clinical practices in the near future. This article will give
an insight to the practicing clinician of the current advances in AI fracture diagnosis by reviewing the current literature on this participant.
Electronic databases were searched for relevant articles relating to AI applications in bone fracture detection. We included all published work
in PubMed, Medline, and Cross‑references, which satisfied the inclusion criteria. The search identified 104 references. Of those, 13 articles
were eligible for the analysis. AI advancements in fracture imaging applications can be divided into the categories of fracture detection,
classification, segmentation, and noninterpretive tasks. Despite the potential work presented in the literature, there are many challenges in the
form of clinical translation and its widespread uses. These challenges range from the proof of safety to clearance from the regulatory agencies.

Keywords: Artificial Intelligence, convolutional neural networking, deep learning, fracture imaging, machine learning, musculoskeletal

Introduction levels of AI performance were subhuman, modern versions are

able to match or even surpass humans’ performance.[5] AI has
Bone fractures are among the most common causes of
also shown promising results in complex diagnostics in other
emergency department visits. Diagnostic errors often occur
medical specialties such as ophthalmology, dermatology, and
due to misinterpretation of radiological examination, which
pathology.[6] The aim of this article is to explore the potential
may lead to the delayed treatment and poor outcomes.[1] The
of utilizing AI in fracture diagnosis by reviewing the current
analysis of causes of fracture diagnostic inaccuracies has
literature on this subject.
found them to be multifactorial, including physician factors,
image quality, insufficient clinical information, fracture
type, and polytrauma.[2] Four out of five diagnostic errors Technical Aspects
in an emergency settings are due to physician factors, yet AI, machine learning (ML), DL, and convolutional neural
radiographs are often interpreted by clinicians who lack the networking (CNN) are terminology, which often used
required specialized expertise.[3] Even with an experienced interchangeably [Figure 1]. AI refers to any skill where a
radiologist, physician fatigue and error may increase during machine performs tasks that mimic human intelligence. ML
a long busy day, increasing the risk of missing a subtle is a subfield of AI that enables a machine to learn and improve
fracture.[4] Thus, a model that can offer assistance to physicians from the experience independently of human action. DL is a
presenting second opinions through highlighting concerning more specialized subfield of ML, which can analyze more data
areas in imaging examination may produce more efficient sets transforming the inputs of an algorithm into outputs using
interpretation, standardize quality, and decrease errors. With
recent advances in deep learning (DL) and computer vision, Address for correspondence: Dr. Ahmed AlGhaithi,
artificial intelligence (AI) may play a significant role in this P. O. Box 478 P.C 130, Muscat, Sultanate of Oman.
field. E‑mail: [email protected]

AI is a powerful technology that has demonstrated good Received: 26‑11‑2020 Revised: 18-01-2021
potential at radiographic image interpretation. While earlier Accepted: 26-01-2021 Published Online: 20-02-2021

This is an open access journal, and articles are distributed under the terms of the Creative
Access this article online Commons Attribution‑NonCommercial‑ShareAlike 4.0 License, which allows others to remix,
Quick Response Code: tweak, and build upon the work non‑commercially, as long as appropriate credit is given and
Website: the new creations are licensed under the identical terms.
www.journalmsr.com
For reprints contact: [email protected]

DOI: How to cite this article: AlGhaithi A, Al Maskari S. Artificial intelligence

10.4103/jmsr.jmsr_132_20 application in bone fracture detection. J Musculoskelet Surg Res
2021;5:4-9.

4 © 2021 Journal of Musculoskeletal Surgery and Research | Published by Wolters Kluwer - Medknow
 AI application in fracture detection

the sophisticated computational models such as deep neural work is provided in Table 1. The application of AI in
networks. CNN is evolutional computational technique of DL, fracture imaging can be classified into four major categories:
which can impact the key areas of medicine such as medical Pathology detection (e.g., calcaneus fracture), segmentation
imaging.[7] CNN is built of computational units called nodes, (which means automated segmentation of the region of
which are analogous to biological brain neurons. Each node interest whereby the irrelevant pixels are cropped out and
takes one or more weighted input connections and performs would not influence the training process e.g., cropping out soft
mathematical operations resulting in outputs that can pass to tissue), classification (e.g., calcaneal fracture classification),
other connected nodes. noninterpretive (e.g., image‑quality improvement from
under‑sampled magnetic resonance imaging or low‑dose
Material and Data Source computed tomography [CT]).[5]
Online databases (PubMed and MEDLINE) search was carried
to find the literature related to AI use in fracture diagnosis. Upper Limbs Fractures
The search was carried accordance to preferred reporting The rate of missing a fracture between the upper and lower
items for systematic reviews and meta‑analyses statement. extremity is almost analogous. Upper limb fractures most
Keywords included “artificial intelligence,” “deep learning,” likely to be missed are elbow (6%), hand (5.4%), wrist (4.2%),
“machine learning,” and “fracture.” Searches were conducted and shoulder  (1.9%). [8] Kim and MacKinnon trained a
on April 1, 2020, yielding a total of 104 articles from the model using 1112 images of wrist radiographs, then they
two databases, without applying any restriction on language added additional 100 images for final testing and analysis
or date of publication [Figure 2]. An independent reviewer (comprising 50 fractures and 50 normal). The area under the
performed screening of articles’ titles and abstracts in the curve (AUC) was 0.954, with a diagnostic sensitivity of 90%
first reviewing stage, in addition to the titles and abstracts of and 88% specificity.[9] Lindsey et al. developed another CNN
crossover references. The following inclusion criteria were model for detecting wrist fractures using 135,409 radiographs
used: all levels of evidence and studies on humans. We did not and was able to improve the sensitivity of clinicians’ image
place restrictions on the target population, the outcome of the reading from 88% unaided to 94% aided, and and by doing so,
disease of interest, or the intended context for using the model. misinterpretation improved by 53%.[10] Olczak et al. designed
We excluded from the search nontraumatic musculoskeletal an algorithm for distal radius fractures and tested it on hand and
pathologies and conferences abstracts due to incomplete data wrist radiographs. They compared the network performance
presentation. with two experienced orthopedic surgeons and showed a high

Results Records identified through

Additional records identified
Identification

database searching
The search terms, as described above, identified 216 references (n = 216 PubMed, Medline
through other sources
(n = 0)
[Figure 2]. After duplicate removal, 104 articles titles and Cross-references)
and abstracts were screened. Of these 19 full‑text articles
were assessed independent by both authors for analysis
eligibility, finally 13 studies satisfied all the inclusion and Records after duplicates removed
(n = 104)
exclusion criteria. A  complete list of included published
Screening

Records excluded after

evaluation of titles and
Records screened abstracts
(n = 104) (n = 85)

Full-text articles Full-text articles

assessed for eligibility excluded, with reasons
Eligibility

(n = 19) (n = 5)

Studies satisfied all inclusion

and exclusion criteria
(n = 13)
Included

Studies included in
quantitative synthesis
(n = 13)

Figure  1: Shows the relationships of artificial intelligence, machine Figure  2: Preferred repor ting items for systematic reviews and
learning, deep learning, and convolutional neural network meta‑analyses flow diagram for study selection

Journal of Musculoskeletal Surgery and Research  ¦  Volume 5 ¦ Issue 1 ¦ January-March 2021 5

 AlGhaithi and Al Maskari:

Table  1: Classification of artificial intelligence application in view of body part fracture

Reference Anatomic area Module Modality Compared to human Performance (metric)
purpose expert performance
Kim et al. Wrist Diagnosis Radiographs No Provided proof of concept in fracture detection on
2018 plain radiographs 0.95 (AUC), 90% sensitivity and
88% specificity
Olczak et al. Hand/wrist/ Diagnosis Radiographs Yes Performance in detecting fractures from hand/wrist/
2017 ankle ankles radiograph sensitivity of 90% and specificity
of 88%accuracy of 83% versus. radiologists, 82%
Lindsey et al. Wrist Diagnosis Radiographs Yes Improved clinicians image reading sensitivity from
2018 88% unaided compared to 94% aided
Chung et al. Proximal Diagnosis and Radiographs Yes Diagnosis accuracy of 96%, 99% sensitivity, 97%
2018 humerus classification specificity
(Neer) Classification accuracy range between 65% and
86%, sensitivity 88% to 97%, specificity 83% to
94% (dependent on the type)
Rayan et al. Pediatrics elbow Diagnosis Radiographs No The model accuracy was 88% with sensitivity of
2019 fractures 91% and specificity of 84%
Urakawa Intertrochanteric Diagnosis Radiographs Yes Convolutional neural network outperformed
et al. 2019 hip fractures orthopedic surgeons at detecting, accuracies of 96%
versus 92%, specificities of 97% versus 97%. 57
and sensitivities of 94% versus 88%
Cheng et al. Hip fracture Diagnosis Radiographs No Accuracy of 91%, a sensitivity of 98%, AUC of
2019 0.98
Adams et al. Neck of femur Diagnosis Radiographs Yes Accuracy of 91%, AUC 0.98
2019 Performing junior’s physician increased from
87.6% to 90.5%
Balaji et al. Femur Diagnosis Radiographs No Accuracy of 90.69% with 86.66% sensitivity and
2020 diaphyseal 92.33% specificity
Kitamura Ankle Diagnosis Radiographs No Model with multiple views shown improved
et al. 2019 accuracy in fracture detection of 81% compared
with single view of 76%
Pranata et al. Calcaneus Classification CT No Sanders classification system model accuracy 98%
2019 (Sander)
Rahmaniar Calcaneus Classification CT Yes An accuracy of 86% with computational
et al. 2019 (Sander) performance of 133 frame per second
Burns et al. Spine Diagnosis CT No Model which detect, localize, classify the fractures
2017 and measure bone density vertebral bodies
employing more lumbar and thoracic CT images.
Attained sensitivity was 95.7%
Tomita et al. Spine Diagnosis CT No Model which detect osteoporotic vertebral fractures
2018 achieved an accuracy of 89.2%
Muehlematter Spine CT No Accuracy of classifying of unstable/stable vertebrae
et al. 2019 was low with AUC 0.53
AUC: Area under the curve, CT: Computed tomography

detection rate with a sensitivity of 90% and specificity of Rayan et al. developed a model with a multi‑view approach,
88%.[11] They did not specify the type of fractures or grade of which mimics the human radiologist when reviewing multiple
difficulty of fracture detection. images of acute pediatric elbow fractures. They used 21,456
radiographic studies containing 58,817 elbow radiographs.
Chung et al. trained a CNN model to detect the fractures
The model accuracy was 88%, with a sensitivity of 91% and
of proximal humerus and classify the type of fracture
specificity of 84%.[13]
(four parts Neer’s classification) on a dataset of 1891
anteroposterior shoulder radiographs. The model showed a high
throughput precision of 96% and a mean AUC of 1.00 compared Lower Limbs Fractures
to specialists, with a sensitivity of 99% and a specificity of Hip fractures constitute 20% of patients admitted to orthopedic
97%. However, the task of classifying the fracture was more surgery, while the incidence of occult fractures on radiographs
challenging; the reported accuracy was ranging from 65% ranges from 4% to 9%.[14] Urakawa et al. developed CNN to
to 85%. The model showed superior performance accuracy study intertrochanteric hip fractures in a total of 3346 hip
compared to general physicians and orthopedic surgeons and images (1773 fractured and 1573 nonfractured hip images).
almost similar performance to specialized shoulder surgeons.[12] His model was compared to the performance of five

6 Journal of Musculoskeletal Surgery and Research  ¦  Volume 5  ¦  Issue 1  ¦  January-March 2021

 AI application in fracture detection

orthopedic surgeons and showed accuracy of 96% versus 92%, were included in the study. However, the grading accuracy of
specificities of 97% versus 57% and sensitivities of 94% versus unstable/stable vertebrae was low with AUC of 0.5.[26]
88%.[15] Cheng et al. developed CNN algorithm, which was
pretrained using 25,505 limb radiographs. Achieved algorithm Discussion
accuracy for diagnosing hip fracture is 91%, sensitivity is
98%. The performance achieved has a low false‑negative rate The efficacy of AI compared to human’s intelligence is
of 2%, which make it a good screening tool.[16] Adams et al. emerging as an effective tool to address the current blemishes
developed a model to detect the neck of femur fracture with of human errors. The AI current status of the technology
an accuracy of 91% and AUC 0.98.[17] Balaji et al. developed can be described by Gartner’s hype cycle [Figure 3], which
CNN to diagnose femur diaphyseal fractures. The model defines how a technology, or an innovation progresses through
was developed using 175 radiographs (100 normal and 75 its life cycle from concept to widespread adoption.[27] The
fractured). Then trained to classify the type of diaphyseal cycle consists of five phases: The first phase is a “technology
femur fracture, namely transverse, spiral, and comminuted. trigger” where only technology is envisioned, followed by a
The achieved highest accuracy of 90.7% with 86.6% sensitivity “peak of inflated expectations phase,” where the technology
and 92.3% specificity.[18] profile is raised with successful and unsuccessful trials. Then,
it is followed by the “trough of disillusionment phase” at
Missed ankle and foot fractures are common, especially in which defects in the technology cause disappointment in its
trauma patients. Some reports estimated missed diagnosis effectiveness, followed by the “slope of enlightenment” as
due to different reasons in the initial contact may reach up to companies begin to test it in their own environments. The
44%, of which 66% were due to radiological misdiagnosis.[19] final phase is the “plateau of productivity,” where technology
This is why researchers tried to train models for this purpose. is available in the market.[25] AI in medical applications, and
Kitamura et al. developed CNN of a small number of ankle fracture detection specifically, is still in the early phases of
radiographs (298 normal and 298 fractured ankles). The model this cycle and fall at the peak of the inflated expectation phase
was trained to detect ankle fractures, where ankle fracture was as more reports continue to demonstrate the efficiency of AI
defined as proximal forefoot, midfoot, hind foot, distal tibia, in detecting fractures.[7] Currently, the work published in the
or distal fibula. The model with multiple views has shown
field of orthopedic traumatology to date is small collective
improved accuracy in fracture detection from 76% to 81%.[20]
initiatives, trying to get proof of concept rather than applying
Pranata et al. proposed two types of CNN algorithms for the
technology.
classification of calcaneal fractures using CT images using
the Sanders classification system. The proposed algorithm The objective of integrating AI into the clinical practice is
exhibited 98% accuracy, which makes it a viable tool for future to augment the workflow at clinical environment rather than
use in computer‑assisted diagnosis.[21] Rahmaniar and Wang replacing the workforce. Thus, with the evaluation of new
developed a computer‑aided method for calcaneal fracture computing platforms and the development of new algorithm
detection in CT. Sanders system was also used for fracture models, the new generation of AI is anticipated to advance
classification, where calcaneus fragments were detected and the quality of workflow in several ways namely improving
marked by color segmentation. The achieved performance the experience of care, the diagnoses, minimizing the errors,
accuracy was high (86%), with a computational performance improving time management, and reducing costs.[5] One of the
of 133 frames per second.[22] greatest challenges, which can be improved by AI is accurate

Spine Fractures
The incidence of misdiagnosed spine fractures varies among
studies and ranges from 19.5% to 45%.[23] Burns et al. was
able to detect, localize, classify vertebral spine fractures as
well as measure bone density of vertebral bodies using lumbar
and thoracic CT images. Achieved sensitivity was 95.7%
and a false‑positive rate of 0.29 per patient for compression
fractures detection and localization.[24] Tomita et al. developed
CNN to extract radiological features of osteoporotic vertebral
fractures in CT scan. The model was trained using 1432 CT
scans, comprised of 10,546 sagittal views, and achieved an
accuracy of 89.2%. The product algorithm was then tested on
128 spine CT scans and an accuracy of 90.8% was achieved.[25]
Muehlematter et al. proposed algorithms to detect vertebrae
at risk of fracture using 58 CT scans of patients with acquired
fractures due to vertebral insufficiency. One hundred and Figure  3: Gartner’s hype cycle provides a graphic illustration of the
twenty items (60 stable vertebrae and 60 unstable vertebrae) maturity and deployment of technologies and applications

Journal of Musculoskeletal Surgery and Research  ¦  Volume 5 ¦ Issue 1 ¦ January-March 2021 7

 AlGhaithi and Al Maskari:

radiological diagnosis, especially in an emergency setting by in the way of clinical translation and widespread uses. These
inexperienced or exhausted clinicians. Therefore, the aid of challenges range from proof of safety to clearance from
AI in the fracture detection is more important in augmenting regulatory agencies.
workflow compared to segmentation or classification.[9] For
example, assisting AI in diagnosing difficult fractures such Conclusion
as elbow fracture in children will have a greater impact on
Several AI models demonstrated certain performance at the
the treatment outcomes compared to classifying the type of
expert level. Although the comprehensive interpretation of the
fracture.
image has not been achieved yet, it is too early to consider AI
By integrating AI into the clinical setting, AI is expected operating independently in a clinical setting. However, with
to provide clinicians with better clinical insights needed to the current technology, AI has the potential to be considered
reduce the errors and improve the quality of task interpretation. to augment the efficiency of clinical workflow.
Another important aspect where AI can play a major role is
official reporting systems after office hours. AI should support Ethical approval
a reporting system for an examination performed in hospitals The authors confirm that this review had been prepared in
where the radiologist is not attending in person.[7] accordance to COPE roles and regulation. Given the nature
of the review, IRB review was not required.
The expectation from the latest AI tools is to demonstrate
the state‑of‑the‑art results. It should improve workload and Financial support and sponsorship
increase daily productivity by replacing the manual retrieval This study did not receive any specific grant from funding
of image data from a database to suggest a comparison with agencies in the public, commercial, or not‑for‑profit sectors.
new images, or even for audits and clinical studies. Moreover, Conflicts of interest
AI should drive efficient worklist prioritization in the work There are no conflicts of interest.
environment, communicating the important image analysis
and ensure automatic assignment to the most appropriate Authors contributions
available physician.[4] AAG contributed to developing the project idea, searched
the literature and interpretation of the results and preparation
and revising the manuscript; SAM contributed in developing
Limitations and Challenges of the idea and critically revising the manuscript. All authors
Artificial Intelligence in the have critically reviewed and approved the final draft and
Clinical Setting are responsible for the content and similarity index of the
manuscript.
AI remains far from independently operating in a clinical
setting. In the face of many successful implementations
of AI models, application limitations must be recognized. References
Published works are of an experimental nature and are not 1. Pinto A, Reginelli A, Pinto F, Lo Re G, Midiri F, Muzj C, et al.
Errors in imaging patients in the emergency setting. Br J Radiol
incorporated into daily clinical practice, which may show
2016;89(1061):20150914.
the feasibility and efficacy of proposed diagnostic models. 2. Pinto A, Berritto D, Russo A, Riccitiello F, Caruso M, Belfiore MP, et al.
Added to that even, the published works are challenging to Traumatic fractures in adults: Missed diagnosis on plain radiographs in
be reproduced, because most training data sets and codes the Emergency Department. Acta Biomed 2018;89:111‑23.
3. Hallas P, Ellingsen T. Errors in fracture diagnoses in the emergency
are rarely published. Moreover, the proposed models need to
department–characteristics of patients and diurnal variation. BMC
be integrated within clinical information software as well as Emerg Med 2006;6:4.
Picture Archiving and Communications Systems in order to be 4. Krupinski EA, Schartz KM, Van Tassell MS, Madsen MT, Caldwell
useful. However, until now, very limited data present this type RT, Berbaum KS. Effect of fatigue on reading computed tomography
examination of the multiply injured patient. J Med Imaging [Internet].
of integration.[5] Moreover, the safety demonstration of these
2017 Jul [cited 2020 Aug 10];4(3).
models to regulatory agencies is an important step for clinical 5. Chea P, Mandell JC. Current applications and future directions
translation and widespread uses. However, there is no denying of deep learning in musculoskeletal radiology. Skeletal Radiol
that AI is making rapid progress and great improvements.[4,6] 2020;49:183‑97.
6. Amisha, Malik  P, Pathania  M, Rathaur  VK. Overview of artificial
In general, new generations of DL and in particular CNN have intelligence in medicine. J Family Med Prim Care 2019;8:2328‑31.
successfully demonstrated to be more accurate and rapidly 7. Esteva A, Robicquet A, Ramsundar B, Kuleshov V, DePristo M, Chou K,
et al. A guide to deep learning in healthcare. Nat Med 2019;25:24‑9.
developed with innovative results than earlier generations. 8. Tyson  S, Hatem  SF. Easily missed fractures of the upper extremity.
These approaches are now diagnostically accurate and are Radiol Clin North Am 2015;53:717‑36, viii.
predicted to outperform human experts in the future. It would 9. Kim  DH, MacKinnon  T. Artificial intelligence in fracture detection:
also potentially give a more precise diagnosis to patients. In Transfer learning from deep convolutional neural networks. Clin Radiol
2018;73:439‑45.
general, to be able to interpret and use artificial intelligence 10. Lindsey R, Daluiski A, Chopra S, Lachapelle A, Mozer, Sicular S, et
correctly, physicians must have a clear understanding of the al. Deep neural network improves fracture detection by clinicians.
tools used on AI. Taking in account the challenges standing Proc Natl Acad Sci U S A. 2018;115(45):11591-11596. doi:10.1073/

8 Journal of Musculoskeletal Surgery and Research  ¦  Volume 5  ¦  Issue 1  ¦  January-March 2021

 AI application in fracture detection

pnas.1806905115. fractures in polytrauma patients: A retrospective cohort study. Patient

11. Olczak J, Fahlberg N, Maki A, Razavian AS, Jilert A, Stark A, et al. Saf Surg 2014;8:10.
Artificial intelligence for analyzing orthopedic trauma radiographs. Acta 20. Kitamura G, Chung CY, Moore BE 2nd. Ankle fracture detection
Orthop 2017;88:581‑6. utilizing a convolutional neural network ensemble implemented with
12. Chung SW, Han SS, Lee JW, Oh KS, Kim NR, Yoon JP, et al. Automated a small sample, de novo training, and multiview incorporation. J Digit
detection and classification of the proximal humerus fracture by using Imaging 2019;32:672‑7.
deep learning algorithm. Acta Orthop 2018;89:468‑73. 21. Pranata YD, Wang KC, Wang JC, Idram I, Lai JY, Liu JW, et al. Deep
13. Rayan JC, Reddy N, Kan JH, Zhang W, Annapragada A. Binomial learning and SURF for automated classification and detection of
classification of pediatric elbow fractures using a deep learning calcaneus fractures in CT images. Comput Methods Programs Biomed
multiview approach emulating radiologist decision making. Radiol Artif 2019;171:27‑37.
Intell 2019;1:e180015. 22. Rahmaniar W, Wang WJ. Real‑time automated segmentation and
14. Yu JS. Easily missed fractures in the lower extremity. Radiol Clin North classification of calcaneal fractures in CT images. Appl Sci 2019;9:3011.
Am 2015;53:737‑55, viii.
23. Aso‑Escario J, Sebastián C, Aso‑Vizán A, Martínez‑Quiñones JV,
15. Urakawa  T, Tanaka  Y, Goto  S, Matsuzawa  H, Watanabe  K, Endo  N.
Consolini F, Arregui R. Delay in diagnosis of thoracolumbar fractures.
Detecting intertrochanteric hip fractures with orthopedist‑level accuracy
Orthop Rev (Pavia) 2019;11:7774.
using a deep convolutional neural network. Skeletal Radiol 2019;48:239‑44.
24. Burns JE, Yao J, Summers RM. Vertebral body compression fractures
16. Cheng CT, Ho TY, Lee TY, Chang CC, Chou CC, Chen CC, et al.
Application of a deep learning algorithm for detection and visualization and bone density: Automated detection and classification on CT images.
of hip fractures on plain pelvic radiographs. Eur Radiol 2019;29:5469‑77. Radiology 2017;284:788‑97.
17. Adams M, Chen W, Holcdorf D, McCusker MW, Howe PD, Gaillard F. 25. Tomita N, Cheung YY, Hassanpour S. Deep neural networks for
Computer vs. human: Deep learning versus perceptual training for the automatic detection of osteoporotic vertebral fractures on CT scans.
detection of neck of femur fractures. J Med Imaging Radiat Oncol Comput Biol Med. 2018;98:8-15.
2019;63:27‑32. 26. Muehlematter UJ, Mannil M, Becker AS, Vokinger KN, Finkenstaedt T,
18. Balaji GN, Subashini TS, Madhavi P, Bhavani CH, Manikandarajan A. Osterhoff G, et al. Vertebral body insufficiency fractures: Detection
Computer‑Aided Detection and Diagnosis of Diaphyseal Femur of vertebrae at risk on standard CT images using texture analysis and
Fracture. In: Satapathy SC, Bhateja V, Mohanty JR, Udgata SK, editors. machine learning. Eur Radiol 2019;29:2207‑17.
Smart Intelligent Computing and Applications. Singapore: Springer; 27. Hype Cycle Research Methodology. Gartner. Available from: https://
2020. p. 549‑59. www.gartner.com/en/research/methodologies/gartner. [Last accessed on
19. Ahrberg AB, Leimcke B, Tiemann AH, Josten C, Fakler JK. Missed foot 2020 Apr 25].

Journal of Musculoskeletal Surgery and Research  ¦  Volume 5 ¦ Issue 1 ¦ January-March 2021 9