THE ROLE OF ARTIFICIAL INTELLIGENCE IN

NUCLEAR MEDICINE

ABSTRACT
In recent years, nuclear medicine has seen a tremendous increase in
the usage of and interest in artificial intelligence (AI). Since the
1950s, AI has existed. Artificial intelligence research in nuclear
medicine has focused particularly well on the application of AI to
more directly imaging-related tasks, such as dosage optimization,
image corrections, and image reconstruction.
In the years ahead, the clinical workflow for nuclear medicine will be
accelerated because to the growing involvement of artificial
intelligence application To determine and classify autoimmune
disorders, radiology is used. This article gives a technical review of
these advancements, classifying the various approaches and
outlining how they were intended to apply and identify current and
future implications in medical, as well as provide an outline of how AI
is being used in nuclear medicine currently .
We give a brief overview to AI design and training processes,next we
examine closely at how they are used in nuclear medicine. The
existing clinical uses of artificial intelligence in nuclear medicine are
presented in this paper.
Because of this, the objective of this article is to give the curious
reader a brief overview.
helping the field of nuclear medicine adopt artificial intelligence and
conduct multidisciplinary research on it.
INTRODUCTION
Prior to anything else, it's crucial to define artificial intelligence (AI),
along with other often used phrases that describe AI, such machine
learning and deep learning. (1)
In a Dartmouth College workshop in 1956, John McCarthy first used
the term "Artificial Intelligence." (2)
Medical imaging is a significant source of patient data. (3)
Over through the last 50 years, the utilization of artificial intelligence
(AI) in nuclear medicine and radiology has expanded (e.g., auto-
contouring) Commentators and AI specialists began predicting that
employment in medicine will become obsolete as soon as 1976.
(4)
Machine learning (ML) is the process by which a computer learns
from previous experiences or data without becoming specifically
programmed to do so. There are primarily three categories of ML
techniques: (i) supervised learning, (ii) unsupervised learning, and
(iii)semi supervised (reinforcement) learning (5)
Under supervised training, An input data set with labels that
correspond to the right output is taught to the model orIn supervised
learning, the most popular method of machine learning, example
data consist of given set of input output pairs, dataset are accessible,
and the model is developed so that its result is as nearest to the
expected label for every input as possible.
Discovering patterns in data for an unknowable conclusion is known
as unsupervised training, which trains data on its own or
Unsupervised learning is the next form of machine learning, when no
response labels are presented. The goal is to uncover latent structure
in the dataset, such as clustering techniques that group related
inputs into smaller groups. (6.58)
Semi-supervised (Reinforcement training) activities investigate each
steps in sequence to determine which is rewarded by getting positive
or negative encouragement, allowing for the studying of interactions'
effects. (7)
Large volumes of data, including clinical pictures, written reports,
and operational data, are frequently produced by clinical nuclear
medicine divisions. (8)
Nuclear medicine as well produces a sizable quantity of
interventional information per day in the form of
predosimetry/postdosimetryresults and medical [Link],
numerous mathematical techniques have been created to obtain a
wide amount of quantitative tumour measurements in the manner
of "radiomic features" from medical pictures. (9)
AI
The term "artificial intelligence" (AI) was firstly used in 1955 to
widely refer to the application of computer methods to missions that
are often related to human intellect (e.g., learning or problem
solving) (10, 11)
In the past, computers that replicate and demonstrate "human"
cognitive abilities related to the human mind, such "learning" and
"problem-solving," were referred to as "artificial intelligence." [12]
In the field of artificial intelligence (AI), it is possible to make
computers that can act like humans and perceive the environment
similarly to us. (13)
AI has the potential to advance nuclear medicine beyond some of
the constraints and assumptions of [Link] the other hand, AI is
prone to certain assumptions that are distinct from those that are
normally engaged with human specialists. (14)
Artificial intelligence may be used to enhance the whole imaging
workflow in the case of medical imaging. (15)
Over the past 50 years, the application of artificial intelligence (AI) in
nuclear medicine and radiology has grown. AI has often been used
to tackle logic and reasoning-related issues.
The end of medical vocations was anticipated by critics and AI
scientists as sooner as 1976. (16)
Since 2015, AI-powered visual surveillance has seen mistake rates
that are lower than those made by humans for the first time. (11,17)
Machine learning that is heavily based in mathematics and statistics
has dominated the subject in the early years of the twenty-first
century. This approach has been very effective in solving many
difficult issues in both business and academics. (18)
In many aspects of human existence, artificial intelligence (AI) is
getting popular rapidly. Since AI has become so broadly adaptable,
daily life has [Link] and AI experts are improving
human life and advancing technology by using AI technologies and
methodologies. (19)
Numerous study programmes that support students in their learning
abilities and enable them to master new material are designed and
developed by AI (20).Students acquire new concepts more rapidly
using simulations and other educational tools.
AI has been used extensively on the Internet, including in search
engines, making it one of the technology's best promising uses.
As long as the challenges and decision-making situations are still
relevant, the information contained within an AI framework may stay
[Link] also facilitates the growth of a learning capacity that can be
used to extend the usefulness and lifespan of the [Link]
attractive aspect of Ai technologies is their capacity to learn from
real-world success and failure. This feature makes AI tools more
reliable as they are used more frequently in applications. (21)
The widespread use of any technology only happens if its
dependability has been demonstrated, and AI has already shown to
be fairly trustworthy in a variety of applications due to its capacity to
replicate procedures that are lacking.
Depending on the computing speed of human intellect throughout a
method of reasoning.
When a cause-and-effect link cannot be shown mathematically, AI
techniques are appropriate. (22)
AI approaches can also handle both qualitative and quantitative data,
a capability that most analytically focused on algorithmic difficulty
and processing power,AI solutions can assist quicker decision-making
by optimizing the decision-making approach. AI may provide quicker
solutions to difficult issues via data collecting, screening, processing,
and decision-making. (23)

EXAMINE

SEGMENTATION

DATA
ANALYSIS

MODELLING

THERAPY,
PROGNOSIS

FIGURE 1. AI Process for examine, segmentation, data analysis,

modelling and therapy
As a result, AI is a helpful auxiliary to radiomics.

LIMITATION OF AI

However, it's important to remember that, like any techniques, AI

techniques have their limitations.
When employing AI-based algorithms to deal with a problem, it is
frequently difficult to acquire a genuine understanding of the issue
and the nature of the result, unlike when using, for instance,
arithmetical programming techniques. A "solution" is still preferable
to "no solution" in tough optimization problems that defy solution
using conventional optimization and arithmetical programming
techniques, according to the supporting argument to the claim that
optimality cannot be guaranteed.
Another limitation, of using AI techniques to tackle a particular
problem is that, for many AI techniques, there is presently little
information available on how to choose the appropriate parameters
to use for a particular method's tuning parameters.
Another restriction or difficulty with automation in general—and AI
in particular—relates to the problem of possible [Link]
example, in the future, if artificial intelligence (AI) techniques are
applied to create partially or completely autonomous automobiles,,
for example, and who should be taken responsible in the event that
an autonomous vehicle is at fault in a collision Despite the fact that
this restriction is not technical in nature, it is a severe problem that
has to be resolved.
A third limitation is that there is now a lot of uncertainty among
transportation professionals about the potential of AI to assist in
solving some of the issues they confront. This is especially true if we
are to explicitly focus on AI applications in
[Link], the use of AI in five main application
areas will be [Link] include (a) traffic management; (b)
forecasting travel demand; (c) safety and security in the
transportation sector; (d) public transportation; and (c)
infrastructure design and development.

MACHINE LEARNING/(DEEP LEARNING)

“Machine learning" (ML), A form of AI known as employs techniques

through data processing without having to be clearly programmed.
(24,25)
A kind of AI known as "machine learning" focuses on creating
computer software that can recognise large amounts of data and
make intelligent forecasts without explicit human programming (26)
While a portion of machine learning (ML) is strongly connected to
computational statistics, which concentrate on computing-assisted
prediction, not all machine learning involves statistical learning. (27 )
First, unlike handmade radiomic characteristics that are
mathematically preset regardless of the data, ML may be used to
find deep radiomic characteristics whose meanings depend entirely
on the data and the job.
In order to find connections, eliminate repetition, develop tractable
depictions in low-dimension spaces, or create prediction models, it is
useful to use machine learning (ML) to mine a huge number of
radiomic characteristics that may be supplemented by other omics
or medical data. (28)
With the use of machine learning, we can make use of imaging data
that is invisible to the human eye and find novel disease patterns and
prediction marker scans. (29)
The benefit of machine learning is that specified patterns are not
constrained. Instead, machine learning can reliably spot recognised
patterns, which might subsequently be used as the foundation for a
diagnostic and prognosis. (30)
The use of machine learning in radiography has increased due to two
recent developments.
First, The amount of medical imaging data being acquired is
increasing. An approximate 3.6 billion radiologic, dental radiographic,
and nuclear medicine exams were carried out per year worldwide
between 2000 and 2007. (31)
Second, A significant increase in the functionality of these
approaches has resulted from recent algorithmic advancements in
the areas of machine learning and new hardware, such as potent
graphics processing units (GPU).
Types of machine learning –
 Supervised learning
 Unsupervised learning
 Semi-supervised learning
 FIGURE 2. Types of Machine Learning

DATA LABELLING

The success of AI relies on correct data labelling as well as the

existence of huge data sets for [Link] accuracy and reliability,
this must be done by readers who have received the necessary
training.
Labels may be produced using a variety of techniques, including
computer modelling and expert judgement. Experts in the subject
should consider the labels to be adequate standards of [Link]
minimise inter- and intraobserver variation when labels are
dependent on expert judgement, it is advised that a complete and in-
depth labelling guide be created and reviewed among labelers.
(32,33)
Labeling data appears to be a straightforward but difficult task that is
necessary for teaching many machine learning systems. We provide
two structured labelling solutions, unique strategies we suggest for
assisting people in defining and improving their ideas in a reliable
way as they label.
NUCLEAR MEDICINE

A branch of medicine known as nuclear medicine focuses on using

radioactive materials for research, diagnosis, and therapy.
Unlike in several European nations (such as France, Germany, and
Italy), where nuclear medicine is a distinct speciality, it is deeply
ingrained in radiology in the United States and, more reportedly, in
The Netherlands. (34 )
One of the main problems, notably in the United States but also
internationally, is the shortage of fresh talent. This is true despite
breakthroughs, the exciting theranostics advancements, and the
rising need for nuclear medicine professionals. (35)
Today's nuclear medical procedures are already greatly impacted by
theranostics, which uses the same receptor for imaging and
treatment. (36)
APPLICATION OF AI IN NUCLEAR MEDICINE

In nuclear medicine imaging, the majority of AI applications have

been concentrated on diagnosis, therapy monitoring, and association
studies with pathology or particular gene mutations.
In order to speed up picture production, lower the amount of tracer
that must be injected, and improve picture quality, it can also be
employed for image production.
An ANN is a great complement to conventional statistical analysis for
specific types of data in research or therapeutic [Link] ANN may
be beneficial in modelling radiation measurements in patients
receiving medication, even if a CNN is necessary for automated
categorization and extraction of information from pictures in
radiation dosimetry.
Nuclear medicine has long used ANNs for many purposes. In 1993, a
single hidden layer with 15 nodes was applied, trained on 100
ventilation-perfusion lung scans using 28 input characteristics, and
then confirmed using 28 fresh instances. (37)
More recently, it was demonstrated that an ANN supervised on
5,685 areas and supported with grounded reality from 6 experienced
nuclear cardiologists, and was outperformed 17-segment defect
assessment in cardiac perfusion images. (38)
Additionally, DL has been applied to pinpoint high-risk individuals
who will gain the most from induction chemotherapy for
nasopharyngeal cancer. (39)
A number of researchers have looked at using CNNs to get over the
limits of maximum-likelihood restoration of activity and attenuation
and offer correct attenuation maps without transfer studies as a
result of advancements in time-of-flight method and DL. (40)
Dosimetry planning for diagnostic and therapeutic purposes
necessitates many imaging sessions, image segmentation, and
mathematical computations, all of which are technically tough and
are frequently seen as difficult clinical tasks for nuclear medicine
specialists. (41)
Scheduling patients, choosing procedures, getting patients ready for
scans, and pre-imaging/treatment dosimetry planning may all be
significantly improved by artificial intelligence systems.(42)
In order to help with the optimization of picture quality, scan
duration, and radiation exposure during diagnostic nuclear medicine
operations, modern generation scanners have started using different
artificial intelligence technologies (43). Artificial intelligence-based
systems can help with pre-therapeutic and post-therapeutic
illumination during therapeutic operations. (44,45 )
The following sections discuss the use of AI in several nuclear
medicine [Link] section (like: CAD, cancer care, potential
benefites, Image interpretation and Reporting, AI and Radiomics as a discovery
tool) describes how AI might be used in the different domains of nuclear
medicine.

FIGURE 3. Succession of AI
 01. Computer-assisted detection (CAD)

Many of the very first medical computer-aided design (CAD) systems,

sometimes known as "expert systems" in the field, were created and
utilised primarily for academic purposes starting in the early 1970s.
(46)
When computer-assisted detection (CAD) was originally introduced
in the 1980s, it was one of the earliest applications of AI in the
medical industry. (47)
In 1998, the Food and Drug Administration (FDA) and the European
Commission (CE) certified the use of CAD for plain chest radiography,
CT chest imaging, and screening and diagnostic mammography. (48)
Before usage, pre-taught techniques in conventional CAD systems
require learning. Their value in the field of medical imaging has
significantly expanded as a result of technological advancement. (49)
CAD in Mammography and Chest radiology

Microcalcifications, lesions, and architectural distortion can be

identified and distinguished using CAD in monitoring and clinical
mammography. (50)
With an emphasis on the early identification of lung cancer, CAD has
been employed in chest imaging for the discovery of lung nodules in
both plain film and CT.(51,52)
The research came to the conclusion that CAD may be used to find
initial lung cancer. [51]
Overall, there is still conflicting evidence about the additional
therapeutic usefulness of CAD in mammography. Some research has
shown that the use of CAD results in a slight increase in recall rate
while increasing cancer detection. (53)
The effectiveness of CAD in identifying additional lung diseases, such
as emphysema, pleural effusion, pneumonia, pneumothorax, and
interstitial lung disease, has been examined in several research. The
results of these have been inconsistent. (54)

02. In cancer care

Throughout the rising use of artificial intelligence and big data

analytics in the field of medicine, initiatives have been taken over the
past few years to increase the level of patient service and quicken
the speed of cancer research. (55)
Large amounts of routine information on cancer diagnosis, treatment
strategy, and results are kept in the form of health-care records and
are either organised or unorganized. (56)
However, there are hazards involved in realising the therapeutic and
economic utility of regular medical big data. Patient privacy must
always be respected in the first place (57,58)
When constructing artificial intelligence systems on such data,
substantial biases, a lack of relevance, and even patient injury might
result from data quality and proven ance difficulties caused by
missing data, inaccurate measurements, and unreported deviations
from clinical procedure. (59)
Screening for breast cancer is another well-researched application of
AI in radiography.(60,61)
Coordinated trials demonstrate a decrease in breast cancer mortality
following mammography mass detection.
The CBIS-DDSM dataset is a sizable freely accessible database for
computer-aided breast cancer detection (Curated Breast Imaging
Subset of the Digital Database for Screening Mammography).(62,63)
03. Potential benefits of artificial intelligence in
nuclear medicine

Numerous retrospective and prospective research define several

strategies that could promote prior cancer detection and more
precise prognosis. (64-66)
Analyses on the possibilities and constraints of the genomes,
proteomics, radiomics, and pathomics signals for the estimation of
the course of cancer are made. (67-69)
The use of artificial intelligence in analytical medical imaging to
diagnose cancer and forecast disease prognosis and clinical
conclusion has shown encouraging results. (69)
04. Image interpretation and reporting

As in radiology, one of the most crucial responsibilities in clinical

nuclear medicine is image interpretation and scan monitoring.
The application of AI in radiology goes well beyond automatically
reporting and interpreting images. (70)
There are three categories of reporting: non-urgent, urgent, and
crucial. While detection with negative findings can be deprioritized,
scans with time-critical discoveries that affect patient treatment can
be emphasized to the radiologist for reporting early.
The radiologist may devote more time to the more complicated
situations if AI successfully processed out routine images.
Radiologists frequently lack quick access to crucial patient data, such
as laboratory findings, histology findings, and patient records that
would aid in the development of document conclusions. (71)
Artificial intelligence techniques may be used to minimize this load,
allowing the expert to focus on other important duties like
interpreting the results, offering his or her professional judgement,
and participating in multimodality therapy groups. (72-74)
05. AI and Radiomics as a Discovery Tool in Nuclear Medicine

AI is anticipated to be essential in supporting medical professionals

and researchers in identifying patterns in sizable biomedical and
imaging databases that are related to patient [Link] ML
techniques have demonstrated promise as effective instruments for
revealing subtle but important correlations among datasets. (75)
Feature redundancy can be caused by technical factors (such as the
mathematical equivalent of radiomics characteristics), by measuring
the same underlying biologic element, or by a biologic causal
connection. (76)
Researchers must utilise cross validation to examine associations
throughout the full dataset with the goal of understanding the
underlying cause. They must then externally evaluate their results to
eliminate false positives and mistaken [Link] ascertain the
baseline false-positive rate for their set of algorithms, for instance,
they may perform the whole AI analysis pipeline on fictitious data
(such as randomised labels) and compare the results to the high
detection rate discovered in the real [Link] summary, radiomics
analysis should eliminate duplicate characteristics, and a multistep
discovery process that is completed by meticulous validation can be
taken into consideration.

Radiologists must identify imaging scenarios that have been shown

to be relevant for diagnosis and/or prognosis in order to diagnose
illness patterns.. One need for this method is that the designs be
described accurately enough for experienced radiologists to detect
them with little interobserver variation.. Radiologists must carefully
analyse pictures in order to reliably identify illness patterns, which is
a difficult undertaking.
06. Image reconstruction
There is a lot of excitement in the field of nuclear medicine over the
advantages that AI may bring to image reconstruction, including
quicker reconstruction, a better signal-to-noise ratio, and lesser
[Link], there are 2 groups of techniques being
investigated in nuclear medicine reconstruction: those that integrate
neural networks into the present physics-based iterative
reconstruction approaches and those that directly rebuild pictures
from projection documents. (77)
Since the beginning of nuclear medicine, departments have had to
strike a balance between radiation doses, longer imaging periods,
low count density in pictures, low signal-to-noise ratio, high partial
size effect, and limited regional pixel density . (78)
There are several computationally costly procedures involved in
nuclear medicine imaging, including attenuation correction, scatter
correction, noise reduction, and partial area correction.(79)
Reconstruction techniques based on AI may have a significant
influence on patient treatment, necessitating adequate validation of
the [Link] should use figures of merit, such as
mean-squared mistakes, structural similarity index, or peak signal-to-
noise ratio, to assess the quality of pictures, but they should also be
aware that these metrics may be deceptive because small,
diagnostically relevant characteristics may be added to or removed
from images without significantly changing summary statistics .(80)
For development studies, computational model observer–based
studies might prove more economic in identifying promising
methods. (81)
A key advancement in medical imaging has been multimodality
fusion imaging with hybrid scanners, and modalities like merged-
PET/CT are currently the preferred option for cancer imaging .(82)
One other hybrid medical scanner with a fast expanding role in
cancer imaging is the combined-PET/MRI. (83)
07. DETECION AND DIAGNOSIS
Although there is a long history of both achievements and losses in
computer-aided diagnosis and detection in radiology, current
developments in AI have brought these technologies closer to being
widely used in nuclear medicine. Automation of medical procedures
in nuclear medicine can be difficult since these procedures must be
extremely resilient to unusual situations, are prone to human error,
and have high stakes (e.g., implants or amputations). (84)
The precision required for the computer-aided diagnosis or
computer-aided detection application should be taken into
consideration when choosing a labelling approach by researchers.
Single-reader labels are inferior than those produced by numerous
readers (adjudication or consensus). (85)
Additionally, developers must be open and honest about the data
sources they use and the features of the learning set population
while developing and evaluating algorithms. No.(86)
08. Histopathology
Microscopic assessment of lymph nodes for metastases after nodal
clearing is a key step in staging breast [Link] on both tumour
grade and slide level receiver-operating feature performance
measures, deep learning based algorithms may identify metastatic
breast tumor in lymph nodes with high accuracy.
According to the study, LYNA (Lymph Node Assistant) outperformed
pathologists in terms of tumour level sensitivity, and it is possible to
employ this technology to more accurately detect tumour [Link]
LYNA system may recognise more lymph node metastatic cancers
even without further changes. (87)
However, a relationship between radiomic data and genomic or
other omic data can offer supporting evidence for histopathologic
conclusions. This is significant since it is thought that up to 23% of
cancer histology results are inaccurate. [88-91]

09. Dermatology

Deep CNN networks identified skin lesions after being developed on

a dataset of 129,450 clinical pictures encompassing 2032 skin
[Link] two binary classifications—malignant melanomas vs
benign naevi and keratinocyte carcinomas versus benign seborrheic
keratosis—the system's efficiency was measured against that of 21
board-certified dermatologists. (92)
10. Ophthalmology
Identifying and supplementary diagnosing are mostly attributed to
medical imaging in the field of ophthalmology, particularly in
disorders that cause blindness.
Machine learning is primarily responsible for how this AI technology
is used. (93)
The most common cause of blindness in people in their working
years is diabetic retinopathy (DR), which primarily disrupts the retinal
microvasculature and causes continuous injury. (94)
Therefore, rapidly requires large-scale DR screening to identify
potentially dangerous alterations at an early stage that would benefit
from treatment and control. (95)
The automated diagnosis of DR has gained a lot of interest, with
research performing microaneurysm, haemorrhage, exudation,
cotton wool spot and neovascularization detection, and even further
categorise phases. Most of them use the fundus images as input. (96)
Conclusion
The importance of artificial intelligence in nuclear medicine is
growing day by day. AI is only a continuation of this and has the
potential to have a huge impact on healthcare. As a result, artificial
intelligence (AI) has the ability to enhance clinical workflow, which
will boost overall efficiency and enable individualised care for
patients. The way humans process the vast amount of images has
altered as an outcome.
1. Noyes J. Artificial intelligence common lisp: fundamentals of symbolic and numeric processing, D.
C. Heath and Company 1992;

2. McCarthy J, Minsky ML, Rochester N, Shannon CE. A proposal for the Dartmouth Summer
Research
Project on Artificial Intelligence, August 31, 1955. AI Mag. 2006;27:12–14.

3. Maes F, Robben D, Vandermeulen D, Suetens P. The role of medical image computing and
machine learning in healthcare. In: Ranschaert ER, Morozov S, Algra PR, editors, Artifcial intelligence
in medical imaging: opportunities, applications and risks. Springer, Berlin; 2019.
[Link]

4. Maxmen JS. The Post-Physician Era: Medicine in the 21st Century. New York, NY: Wiley; 1976.

5. Hussain, Mubashir&Manhas, Dr. (2016). ARTIFICIAL INTELLIGENCE FOR BIG DATA:

POTENTIAL AND RELEVANCE. International Academy of Engineering and Medical Research, 2016.

Volume-1, ISSUE-1.

[Link] SB. Supervised machine learning: a review of classification techniques,Informatica. 2007;

31: 249–68
7. Sutton RS, Barto AG. Introduction toreinforcement learning. 9: MIT Press; [Link]. 1054. doi:
https:// doi. org/ 10. 1109/ TNN.1998. 712192
08. Kharat AT, Singhal S. A peek into the future of radiology using big data applications. Indian J
Radiol Imaging 2017; 27:241–248
09. Gillies RJ, Kinahan PE, Hricak H. Radiomics: images are more than pictures, they are data.
Radiology 2016; 278:563–577.

[Link] A, Tam R, Cadrin-Chenevert A, et al. Canadian Association of Radiologists white paper on

artificial intelligence in radiology, Can Assoc Radiol J. 2018;69: 120–135.

11. McBee MP, Awan OA, Colucci AT, et al. Deep learning in radiology. Acad Radiol. 2018;25:1472–
1480

12. Stuart Russell and Peter Norvig characterize this definition as "thinking humanly" and reject it in
favor of "acting rationally".
13. Russell S, Norvig P. Artificial intelligence a modern approach. 3rd Edn: Pearsons; 2009.

14. Haibe-Kains B, Adam GA, Hosny A, et al. Transparency and reproducibility in artificial intelligence.
Nature. 2020;586:E14–E16.

[Link] I, Bengio Y, Courville A. Deep Learning. MIT Press, 2016.

[Link]

16. Maxmen JS. The Post-Physician Era: Medicine in the 21st Century. New York, NY: Wiley; 1976.

[Link] C, Allen B, Erickson B, et al. A roadmap for foundational research on artificial

intelligence in medical imaging: from the 2018 NIH/RSNA/ACR/The Academy Workshop. Radiology.
2019;291:781–791.

[Link], Jack (2015b). "Why 2015 Was a Breakthrough Year in Artificial

Intelligence". [Link]. Archived from the original on 23 November 2016. Retrieved 23 November 2016..

19. "Total funding of AI startups worldwide 2014-2021 | Statista," 2021.

20. I. Caponetto, J. Earp, and M. Ott, "Gamification and education: A literature review," in European
Conference on Games Based Learning, 2014, vol. 1: Academic Conferences International Limited, p.
50.

21. Russell, S., and P. Norvig. Artificial Intelligence: A Modern Approach, 2nd edition. Pearson
Education, Inc., 2003.

22. Patterson, D. Introduction to Artificial Intelligence and Expert Systems. Prentice Hall, Inc., 1990.

23. Schleiffer, R. Intelligent Agents in Traffic and Transportation. Transportation Research Part C, Vol.
10, 2002, pp. 325–329.
[Link] A, Tam R, Cadrin-Chenevert A, et al. Canadian Association of Radiologists white paper on
artificial intelligence in radiology, Can Assoc Radiol J. 2018;69: 120–135.

[Link] AS, Lundervold A. An overview of deep learning in medical imaging focusing on MRI. Z
Med Phys. 2019;29:102–

26. Mitchell TM. Machine Learning: McGrawHill Higher Education; 1997

[Link], Jerome H (1998). “Data Mining and Statistics: what’s the connection?”. computing
Science and Statistics . 29(1):3-9
[Link] CFW, Ubelh € €or C, Mes SW, et al. Stable prediction with radiomics data. [Link]
website. [Link]

[Link] BJ, Raghunath S, Karwoski RA et al(2013) Quantitative computed tomographyimaging

of interstitial lung diseases. J ThoracImaging 28(5):298–307.
[Link]

[Link] A et al (2013) Computational identification and quantification of trabecular

microarchitecture classes by 3-D texture analysis-based clustering. Bone 54(1):133–140

31. Mettler FA Jr et al (2009) Radiologic and nuclear medicine studies in the United States
andworldwide: frequency, radiation dose, andcom-parisonwith otherradiationsources—1950–
2007.Radiology253(2):520–531
[Link] V, Peng L, Coram M, et al. Development and validation of a deep learning algorithm for
detection of diabetic retinopathy in retinal fundus photographs. JAMA. 2016;316:2402–2410.

33. Krause J, Gulshan V, Rahimy E, et al. Grader variability and the importance of reference
standards for evaluating machine learning models for diabetic retinopathy. Ophthalmology.
2018;125:1264–1272

34. Beyer T, Townsend DW, Brun T, et al. A combined PET/CT scanner for clinical oncology. J Nucl
Med. 2000;41:1369–1379.

[Link] S. Do you want an inspirational career? Choose nuclear medicine. J Nucl Med.
2019;60(suppl 2):25S–28S.

36. Herrmann K, Larson SM, Weber WA. Theranostic concepts: more than just a fashion trend-
introduction and overview. J Nucl Med. 2017;58:1S–2S.

[Link] JA, Palmer E. Neural network analysis of ventilation-perfusion lung scans. Radiology.
1993;186:661–664.

[Link] K, Kudo T, Nakata T, et al. Diagnostic accuracy of an artificial neural network compared
with statistical quantitation of myocardial perfusion images: a Japanese multicenter study. Eur J Nucl
Med Mol Imaging. 2017;44:2280–2289.

[Link] H, Dong D, Fang MJ, et al. Prognostic value of deep learning PET/CTbased radiomics:
potential role for future individual induction chemotherapy in advanced nasopharyngeal carcinoma.
Clin Cancer Res. 2019;25:4271–4279.

[Link] D, Kim KY, Kang SK, et al. Improving the accuracy of simultaneously reconstructed activity
and attenuation maps using deep learning. J Nucl Med. 2018;59:1624–1629.

[Link] G, Hobbs RF. Dosimetry for radiopharmaceutical therapy. Semin Nucl Med 2014; 44:172–
178.

[Link] M, Epelbaum S, Gagliardi G, Colliot O, Dormont D, Dubois B, et al.; Alzheimer’s Disease

Neuroimaging Initiative* and the INSIGHT-preAD study. Reduction of recruitment costs in preclinical
AD trials: validation of automatic pre-screening algorithm for brain amyloidosis. Stat Methods Med
Res 2020; 29:151–164
[Link] B, Liu JZ, Cauley SF, Rosen BR, Rosen MS. Image reconstruction by domain-transform
manifold learning. Nature 2018; 555:487–492.

[Link] T, Akhavanallaf A, Zaidi H. Construction of patient-specific computational models for organ

dose estimation in radiological imaging. Med Phys 2019; 46:2403–2411.

[Link] S, Gafita A, Afshar-Oromieh A, Eiber M, Rominger A, Shi K. Voxel-wise prediction of post-

therapy dosimetry for 177Lu-PSMA I&T therapy using deep learning. J Nucl Med 2020; 61 (Suppl
1):1424.

46. Miller RA, Pople Jr HE, and Myers JD (1982). "Internist-I, an experimental computer-
based diagnostic consultant for general internal medicine". New England Journal of
Medicine. 307 (8): 468–476. doi:10.1056/NEJM198208193070803. PMID 7048091

[Link] K. Computer-Aided diagnosis in medical imaging: historical review, current status and future
potential. Comput Med Imaging Graph 2007; 31(4-5): 198–211. doi: [Link]
10.1016/[Link].2007.02.002

48. Castellino RA. Detection Caided CAD: an overview. Cancer Imaging 2005; 5: 17–19.

[Link] J, Dodelzon K. A review of computer aided detection in mammography. Clin Imaging 2018;
52: 305–9. doi: [Link] 10.1016/[Link].2018.08.014

50. Calas MJG. Cad and mammography: why use this tool? Radiol Bras 2012; 45: 46–52

51. Sahiner B, Chan H-P, Hadjiiski LM, Cascade PN, Kazerooni EA, Chughtai AR, et al. Effect of CAD on
radiologists' detection of lung nodules on thoracic CT scans: analysis of an observer performance
study by nodule size. Acad Radiol 2009; 16: 1518–30. doi: [Link]
2009.08.006

52. Kakeda S, Moriya J, Sato H, Aoki T, Watanabe H, Nakata H, et al. Improved detection of lung
nodules on chest radiographs using a commercial computeraided diagnosis system. AJR Am J
Roentgenol 2004; 182: 505–10. doi: [Link] 2214/ajr.182.2.1820505

[Link] MJ, Whaley DH, Brandt KR, Amrami KK. Screening mammograms: interpretation with
computer-aided detection--prospective evaluation. Radiology 2006; 239: 375–83. doi:
[Link] 1148/radiol.2392042121

[Link] Ginneken B, Hogeweg L, Prokop M. Computer-Aided diagnosis in chest radiography: beyond

nodules. Eur J Radiol 2009; 72: 226–30. doi: [Link] 1016/[Link].2009.05.061
[Link] G, Fan W, Luo H, Zhu X. The emerging roles of artificial intelligence in cancer drug
development and precision therapy. Biomed Pharmacother 2020; 128:110255.

[Link] AL, Gulliford SL, Ebert MA, Orton CG. Point/Counterpoint. Future radiotherapy practice
will be based on evidence from retrospective interrogation of linked clinical data sources rather than
prospective randomized controlled clinical trials. Med Phys 2014; 41:030601.

[Link] M, Ferretti A, Hurst S, Puhan M, Lovis C, Vayena E. Considerations for ethics review of big
data health research: a scoping review. PLoS One 2018; 13:e0204937.
[Link] of Medicine (US) Committee on Regional Health Data Networks; Donaldson MS, Lohr KN,
editors. Health Data in the Information Age: Use, Disclosure, and Privacy. Washington (DC): National
Academies Press (US); 1994. p. 4. Confidentiality and Privacy of Personal Data. [Link]
[Link]/books/NBK236546/

[Link] FB, Hamzah FM, Razali SFM, Jaafar O, Jamil NA. Imputation methods for recovering
streamflow observation: a methodological review. Cogent Environ Sci 2020; 6:1745133

[Link] H, Heindl A, Khara G, Korkinof D, O’Neill M, Yearsley J, Karpati E, Rijken T, Kecskemethy P,

Forrai G. Deep learning in breast cancer screening. In: Ranschaert ER, Morozov S, Algra PR, editors
Artifcial intelligence in medical imaging. Springer (2019). [Link]
2_14.

61. Harvey H, Karpati E, Khara G, Korkinof D, Ng A, Austin C, Rijken T, Kecskemethy P. The role of
deep learning in breast screening. Curr Breast Cancer Rep. 2019;11(1):17–22.
[Link]

62. Lee RS, Gimenez F, Hoogi A, Rubin D. Curated breast imaging subset of DDSM [Dataset]. The
Cancer Imaging Archive 2016.

[Link] K, Vendt B, Smith K, Freymann J, Kirby J, Koppel P, Moore S, Phillips S, Maftt D, Pringle M,
Tarbox L, Prior F. The cancer imaging archive (TCIA): maintaining and operating a public information
repository. J Digit Imaging. 2013;26(6):1045–57. [Link]
64. Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018:
GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA
Cancer J Clin 2018; 68:394–424.

65. Bruni L, Albero G, Serrano B, Mena M, Gómez D, Muñoz J, Bosch FX, de Sanjosé S. ICO/IARC
Information Centre on HPV and Cancer (HPV Information Centre). HumanPapillomavirus and Related
Diseases in India. Summary Report. 2019. [Link]
[Accessed 17 March, 2020].

66. Rose PG, Bundy BN, Watkins EB, Thigpen JT, Deppe G, Maiman MA, et al. Concurrent cisplatin-
based radiotherapy and chemotherapy for locally advanced cervical cancer. N Engl J Med 1999;
340:1144–1153.

67. Thomas GM. Improved treatment for cervical cancer–concurrent chemotherapy and
radiotherapy. N Engl J Med 1999; 340:1198–1200.

68. Yu KH, Snyder M. Omics profiling in precision oncology. Mol Cell Proteomics 2016; 15:2525–
2536.

69. Surasi DS, Bhambhvani P, Baldwin JA, Almodovar SE, O’Malley JP. ¹8 FFDG PET and PET/CT
patient preparation: a review of the literature. J Nucl Med Technol 2014; 42:5–13.

[Link] Technology News. Improving efficiency with an intelligent worklist and rules engine.
Available from: https:// [Link]/content/improvingefficiency-intelligent-worklist-and-
rulesengine.

71. Lakhani P, Prater AB, Hutson RK, Andriole KP, Dreyer KJ, Morey J, et al. Machine learning in
radiology: applications beyond image interpretation. Journal of the American College of Radiology
2018; 15: 350–9. doi: [Link] 2017.09.044
72. Pesapane F, Codari M, Sardanelli F. Artificial intelligence in medical imaging: threat or
opportunity? Radiologists again at the forefront of innovation in medicine. Eur Radiol Exp 2018;
2:35.

73. Nensa F, Demircioglu A, Rischpler C. Artificial intelligence in nuclear medicine. J Nucl Med 2019;
60 (Suppl 2):29S–37S.

74. Seifert R, Weber M, Kocakavuk E, et al. AI and machine learning in nuclear medicine: future
perspectives. Semin Nucl Med 2020. In press. doi: 10.1053/[Link].2020.08.003

[Link] E, Cutler DM, Leskovec J, Mullainathan S, Obermeyer Z. An algorithmic approach to

reducing unexplained pain disparities in underserved populations. Nat Med. 2021;27:136–140.

[Link] E, Mesotten L, Zhovannik I, et al. Plausibility and redundancy analysis to select FDG-PET
textural features in non-small cell lung cancer. Med Phys. 2021; 48:1226–1238

[Link] AJ, Corda G, Mehranian AD, Costa-Luis C, Ellis S, Schnabel JA. Deep learning for PET image
reconstruction. IEEE Trans Radiat Plasma Med Sci. 2021; 5:1–25.

[Link] Human Health Series. Quantitative Nuclear Medicine Imaging: concepts, Requirements And
Methods. Vienna: International Atomic Energy Agency; 2014. p. 24.
[Link] Pub1605_web.pdf. [Accessed December 09,
2020].

[Link] DU, Isa NA, Uzun B, Ozsahin I. Effective analysis of image reconstruction algorithms in
nuclear medicine using fuzzy PROMETHEE. 2018 Advances in Science and Engineering Technology
International Conferences (ASET); Abu Dhabi; 2018. pp. 1–5

[Link] J, Sohn JH, Behr SC, Gullberg GT, Seo Y. CT-less direct correction of attenuation and scatter
in the image space using deep learning for whole-body FDG PET: potential benefits and pitfalls.
Radiol Artif Intell. 2020;3:e200137.

[Link] Z, Rahman MA, Schindler T, Laforest R, Jha AK. A physics and learning-based transmission-less
attenuation compensation method for SPECT. In: Proceedings of SPIE, Medical Imaging 2021: Physics
of Medical Imaging. SPIE; 2021:1159512

[Link] I, Rodríguez-Vigil B, Gómez-León N. Integrated FDG PET/CT: utility and applications in

clinical oncology. Clin Med Oncol 2008; 2:181– 198.

83. Chen Y, An H. Attenuation correction of PET/MR imaging. Magn Reson Imaging Clin N Am 2017;
25:245–255.

[Link] Z, Kitajima K, Hirata K, et al. Preliminary study of AI-assisted diagnosis using FDG-PET/CT for
axillary lymph node metastasis in patients with breast cancer. EJNMMI Res. 2021;11:10

[Link] DA, Moy L, Bredella MA, et al. Assessing radiology research on artificial intelligence: a
brief guide for authors, reviewers, and readers—from the Radiology editorial board. Radiology.
2020;294:487–489.

[Link] B, Quer G, Beaulieu-Jones BK, et al. Minimum information about clinical artificial
intelligence modeling: the MI-CLAIM checklist. Nat Med. 2020;26:1320– 1324.
[Link] Y, Kohlberger T, Norouzi M, Dahl GE, Smith JL, Mohtashamian A, et al. Artificial Intelligence-
Based breast cancer nodal metastasis detection: insights into the black box for pathologists. Arch
Pathol Lab Med 2019; 143: 859–68. doi: [Link] 5858/arpa.2018-0147-OA

[Link] J, Hsieh YC, Acharys S, Rademaker AW, Morrow M. Results of the Lynn Sage
Second-Opinion program for local therapy in patients with breast carcinoma: changes in
management and determinants of where care is delivered. Cancer 2002;94(4):889-894

[Link] PL, Schultz D, Renshaw AA, et al.. The impact of pathology review on treatment
recommendations for patients with adenocarcinoma of the prostate Urol Oncol
2004;22(4):295-299

[Link] Vl, Nessenger KA, Hao N, Wiley EL, Morrow M. changes in brest cancer
therapy because of pathology second opinions. ANN Surg Oncol 2002;9(10):982-987

[Link] JD, Westra WH, Epstein JI. Mandatory second opinion surgical pathology at a
large referral hospital. Cancer 1999;86(11):2426-2435

[Link] A, Kuprel B, Novoa RA, Ko J, Swetter SM, Blau HM, et al. Dermatologist-level classification
of skin cancer with deep neural networks. Nature 2017; 542: 115–8. doi:
[Link]

[Link] A, Taylor P, Kalpathy-Cramer J, Tufail A. Machine learning has arrived. Ophthalmology

2017;124(12):1726-1728

[Link] S, Chowdhury H, Elagouz M, Sivaprasad S. Inravitreal bevacizumab (Avastin) for age-

related maculat degeneration a critical analysis of literature. Eye (Lond) 2010;24(5):816-824

[Link] T, Erlandsen M. Visual prognosis after panretinal photocoagulation for proliferative

diabetic retinopathy. Acta Ophthalmol Scand 2006;84(1):16-20

[Link] E, Karnowski TP, Govindasamy VP, Abdelrahman M, Tobin KW. Automated

diagnosis og retinopathy by content-based image retrieval. Retina 2008;28(10):1463-1477