Maedica 17 420

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

MAEDICA – a Journal of Clinical Medicine

Mædica - a Journal of Clinical Medicine 2022; 17(2): 420-426


https://doi.org/10.26574/maedica.2022.17.2.420
R eview

Integrating Artificial Intelligence for


Clinical and Laboratory Diagnosis –
a Review
Taran Rishit UNDRUa, Utkarsha UDAYb, Jyothi Tadi LAKSHMIc, Ariyanachi KALIAPPANd,
Saranya MALLAMGUNTAe, Shalam Sheerin NIKHATf, V SAKTHIVADIVELg, Archana GAURh
a
Systems Engineer at TCS, Backend developer at TCS-Apple, Hyd, India
b
West Bengal University of Health Sciences, Kolkata, India
c
Department of Microbiology, All India Institute of Medical Sciences, Bibinagar, India
d
Department of Anatomy, AIIMS Bibinagar, Bibinagar, Yadadri-Bhuvanagiri dist., India
Department of Microbiology, ESIC Medical College and Hospital Sanath Nagar,
e

Hyderabad, India
Department of Microbiology, AIIMS Bibinagar, Bibinagar, Yadadri-Bhuvanagiri dist., India
f

g
Department of General Medicine, AIIMS Bibinagar, Bibinagar, Yadadri-Bhuvanagiri dist.,
India
h
Department of Physiology, AIIMS Bibinagar, Bibinagar, Yadadri-Bhuvanagiri dist., India

ABSTRACT
Introduction: The development of medical artificial intelligence (AI) is related to programs intended
to help clinicians formulate diagnoses, make therapeutic decisions and predict outcomes. It is bringing a
paradigm shift to healthcare, powered by the increasing availability of healthcare data and rapid progress in
analytical techniques (1). Artificial intelligence techniques include machine learning methods for structured
data, such as classical support vector machines and neural networks, modern deep learning (DL), and
natural language processing for unstructured data.
Methodology: More than 50 articles were reviewed and 41 of them were shortlisted. The review was
based on a literature search in PubMed, Embase, Google Scholar, and Scopus databases.
Review: Laboratory medicine incorporates new technologies to aid in clinical decision-making, disease
monitoring, and patient safety. Clinical microbiology informatics is progressively using AI. Genomic
information from isolated bacteria, metagenomic microbial results from original specimens, mass spectra
recorded from grown bacterial isolates and large digital photographs are examples of enormous datasets in
clinical microbiology that may be used to build AI diagnoses.
Conclusion: Technological innovation in healthcare is accelerating and has become increasingly interwoven
with our daily lives and medical practices such as smart health trackers and diagnostic algorithms.
Keywords: artificial intelligence, AI, laboratory medicine, infectious diseases, clinical diagnosis.

Address for correspondence:


Dr. Lakshmi Jyothi Tadi, Additional Professor
Department of Microbiology, All India Institute of Medical Sciences, Bibinagar, India
Email: dr.tljyothi@gmail.com

Article received on the 11th of May 2022 and accepted for publication on the 18th of June 2022

420 Maedica A Journal of Clinical Medicine, Volume 17, No. 2, 2022


Integration of Artificial Intelligence (AI)

BACKGROUND (CT) scans, for example, onto the body, to see

I
both at the same time. As long as those images
ntegrating artificial intelligence (AI) in clini- are lined up just right, it is as if surgeons have an
cal diagnosis and treatment recommenda- X-ray vision.
tions is now being widely implemented. Artificial intelligence aims to mimic human
Artificial intelligence is a branch of sci- cognitive functions. It is bringing a paradigm shift
ence that deals with helping machines find to healthcare, powered by the increasing avai­
a solution to complex problems in a more human- lability of healthcare data and rapid progress of
like fashion. This generally involves borro­wing analytical techniques (1). Artificial intelligence
characteristics from human intelligence and ap- techniques include machine learning methods
plying them as algorithms in a com­puter-friendly for structured data, such as the classical support
way. q vector machine and neural network, and mo­
dern deep learning, as well as natural language
processing for unstructured data. Major areas
that use AI tools include cancer, neurology, me­
dicine and cardiology (1).
Machine learning is the type of AI which is
being predominantly used in the field of medi-
cine. Neural networks and deep learning are the
most complex kinds of machine learning, with
many levels of features or variables that predict
outcomes.
FIGURE. Goals of AI
The detection of possibly malignant tumours
in radiography pictures is a common use of deep
learning in healthcare.
INTRODUCTION Deep learning is increasingly being used in
radiomics or the discovery of clinically significant
T he development of medical artificial intelli-
gence is related to the development of AI
programs that are intended to help the clinician
patterns in imaging data that are beyond the de-
tection ability of the human eye (2).
Yet, another type, known as natural language
in the formulation of a diagnosis, making of the­
processing (NLP), includes applications such as
ra­peutic decisions, and prediction of the out- speech recognition, text analysis, translation and
come. Such systems include artificial neural net- other goals related to language. There are two ba-
works (ANN), fuzzy expert systems, hybrid sic approaches to it: statistical and semantic NLP.
intel­ligent systems, and evolutionary computa- In healthcare, the dominant applications of
tion. The advancement of intelligent medical NLP involve the creation, understanding and
technologies has paved the way for the emer- classification of clinical documentation and pu­
gence of a new medical field: augmented medi- blished research. Natural language processing
cine. Other digital tools, such as surgical naviga- systems can analyse unstructured clinical notes
tion systems for computer-assisted surgery, and on patients, prepare reports (eg, on radiology exa­
virtuality-reality continuum tools for surgery, minations), transcribe patient interactions and
pain management, and psychotic diseases, are conduct conversational AI.
also enabling augmented medicine. Robotic process automation employed in
AccuVein is one example. The handheld de- healthcare for repetitive operations such as prior
vice uses laser-based technology to “see” through authorization, updating patient information, and
your skin and into your veins. It is intended to billing.
make it easier for doctors, nurses or others to Since the 1970s, when MYCIN was created at
find a vein to draw blood or place an IV device. Stanford to diagnose blood-borne bacterial infec-
The augmented reality technology included a tions, AI has been focused on disease diagnosis
headset with a display the doctors could see and therapy. Artificial intelligence is a process of
through to the person. It allowed them to project acquiring data followed by its interpretation and
images from X-rays or computed tomography learning to achieve the desired outcome (3). q

Maedica A Journal of Clinical Medicine, Volume 17, No. 2, 2022 421


Integration of Artificial Intelligence (AI)

METHODOLOGY shown that human-driven control methods had


a major influence on screening, analysis, predic-

M ore than 50 articles were reviewed, of


which 41 were shortlisted. The review was
based on a literature search in PubMed, Embase,
tion, and tracking of infected persons. They also
compared AI and non-AI based ways to identify
COVID-19 symptoms and had clearly highligh­
Google Scholar, and Scopus databases. q ted AI as a crucial instrument in infectious di­
sease management as well as its current deploy-
REVIEW ment in the COVID-19 pandemic, with the goal
of reducing time, cost and human effort, toge­
L aboratory medicine is always incorporating
new technology to aid clinical decision-ma­
king, disease monitoring and patient safety. In-
ther with providing efficient and dependable so-
lutions in the pandemic (4).
Lin et al conducted a retrospective study in
novation has the ability to alter healthcare sys-
which they used a deep learning model –
tems and laboratory medicine by providing
COVNet is a COVID-19 detection neural net-
healthcare workers with the knowledge and
work that was created to extract visual informa-
tools they need to give better care to more pa-
tion from volumetric chest CT scans for
tients while using less resources. Artificial intelli-
COVID-19 detection; CT scans of communi­ -
gence has the ability to transform present diag-
ty-acquired pneumonia (CAP) and other non-
nostic, disease preventive and control techniques,
pneumonia abnormalities were included to test
dramatically improving patient safety and treat-
the robustness of the model. The authors disco­
ment quality. To enhance workflow and person-
vered that in the independent test set, the
nel utilisation, labs now employ software to au-
per-scan sensitivity and specificity for identifying
tomate sample, operation, and outcome
CAP were 87% (152 of 175 scans) and 92% (239
management. Rule-based autoverification, for of 259 scans), respectively, with an area under
example, compares patient outcomes to many the receiver operating characteristic curve of
factors in order to validate and expedite repor­ 0.95 (95% CI: 0.93-0.97). They demonstrated
ting or reactive actions. Simultaneously, sophisti- that a deep learning algorithm could detect coro-
cated systems monitor activities to identify bo­ navirus 2019 and distinguish it from commu­ -
ttlenecks and warn of possible problems, such as nity-acquired pneumonia and other lung di­
STAT sample delays or reagent expiry. Health- seases (5).
care systems rely on digitalization to manage Another implementation of AI in the
hundreds or thousands of point-of-care (POC) COVID-19 pandemic is treatment surveillance,
testing devices and their data outside of the cen- which allows for automatic prediction of virus
tral lab. These rule-based programmes conduct spread, diseased individuals and keeping the
activities and calculations directly as they are public informed about the pandemic situation as
programmed, following predetermined logic. Ar- well as contact tracing of individuals by identi­
tificial intelligence is the next phase in the evolu- fying "hot spots" to trace the infection and pre-
tion of laboratory software. Awareness of AI and dict the future course and chances of remission.
its applications is still an ongoing debate amongst Airport testing is one of the numerous precau-
the treating physicians, but it has shown accura- tions used to prevent the spread of a contagious
cies in the radiological and laboratory diagnosis disease. Various machine learning parameters,
of infectious diseases. including Matlab, nested one versus one (OVO)
Infectious diseases are characterised by rapid support vector machine (SVM), leave one out
transmission and harmful effects, so researchers cross-validation (LOOCV) and SVM learning
must explore and foresee the location and future method, have been used in combination in dia­
intensity of an epidemic. Mathematicians have gnostics by separating genetic sequences from
been able to use machine learning algorithms bacteria (6).
not only to estimate the size and location of the Also, AI assists in the creation of vaccines and
epidemic, but also to investigate the infection re- medications by speeding up drug development
lationship. The COVID-19 pandemic has recent- methodologies, diagnosing processes and clini-
ly served as a catalyst for AI and innovation. Kaur cal trial management (7). Chatbots have been
et al discussed the results of AI models and also developed by a number of healthcare or-

422 Maedica A Journal of Clinical Medicine, Volume 17, No. 2, 2022


Integration of Artificial Intelligence (AI)

ganisations to improve mental health services, based on DIHM do not need extensive blood
telehealth as well as patient engagement and sample processing (16). Infectious outbreaks
wellbeing. Moreover, numerous researchers such as Ebola have necessitated the use of tech-
have highlighted the influence and role of emer­ nologically-based approaches such as logistic re-
ging healthcare platforms and electronic medi- gression (LR), SVM classifiers (with good accuracy
ums such as mobile health, 5G, telemedicine, profiles), single layer artificial neural network
internet of things, AI and others in the fight (ANN) and decision tree (DT), which have pro­
against pandemics by acting as advanced wea­ ven as a successful set of indicators to be used for
pons to prevent further spread of infection (8-10). various Ebola-related data configurations (17). To
Clinical microbiology informatics is progres- allow a better response, technology techniques
sively using AI. Genomic information from iso- should also integrate socioeconomic aspects in
lated bacteria, metagenomic microbial results order to support a consistent approach to infec-
from original specimens, mass spectra recorded tion diagnosis and treatment (18).
from grown bacterial isolates, and huge digital Artificial intelligence, which employs convo-
photographs are all examples of enormous data lutional networks, is predicted to play a signifi-
sets in clinical microbiology that may be used to cant role in cancer outcome prediction. A study
build AI diagnoses (11). Microscopy for traditio­ conducted by Lee Su-In et al demonstrated a
nal Gram stains, ova and parasite production, promising approach to identify robust molecular
and histopathology slides might be revolutio­ markers for targeted treatment of acute myeloid
nised by machine-learning-based image analysis. leukemia (AML) by introducing data from
For instance, a neural network could classify 30 AML patients including genome-wide gene
Gram stains from positive blood cultures into expression profiles and in vitro sensitivity to
Gram positives/negatives and cocci/rods with 160 chemotherapy drugs, a computational me­
amazing accuracy (12). Mathison et al offer a thod to identify reliable gene expression markers
computer vision validation for a novel applica- for drug sensitivity by incorporating multi-omic
tion: protozoa identification in trichrome-stained prior information relevant to each gene’s poten-
faecal smears. It is a complete validation of the tial to drive cancer. They also showed that their
computer vision programme, including studies of method outperformed several state-of-the-art
accuracy, precision and detection limit (13). A approaches in identifying molecular markers
systematic review documented that machine replicated invalidation data and predicting drug
learning was used for species identification and sensitivity accurately (4). Hirasawa et al built a
antibiotic susceptibility testing. Support vector system that could process a large number of
machines, genetic algorithms, artificial neural stored endoscopic pictures in a short amount of
networks and fast classifiers were among the time with clinically meaningful diagnostic capa-
most extensively utilised machine learning ap- bilities, which may be used in everyday clinical
proaches (14). Deep learning is particularly sig- practise to relieve endoscopists' workload (19).
nificant for omic analysis because it allows the Gulshan et al designed an algorithm for identi­
combination and interpretation of image-based fying referable diabetic retinopathy and macular
data with – omic information, allowing this data edoema with excellent sensitivity and specificity
to be used to generate new and more trust­worthy (20). Lee et al developed a deep learning-based
knowledge (15). Recently, infectious disorders computer-aided diagnosis approach for detec­
such as malaria, which have time-consuming di- ting cervical lymph node metastases by CT scan
agnosis criteria and need many health services, in thyroid cancer patients (21). Early lung cancer
are diagnosed using machine learning tech- chest CT scans with AI-assisted automated lear­
niques. The use of digital in-line holographic mi- ning showed good specificity and sensitivity for
croscopy (DIHM) data processing to detect in- early lung cancer detection and could be benefi-
fected RBCs in the blood of malaria patients is a cial in the future in assisting doctors in the early
suitable and cost-effective technology. Further- diagnosis of microscopic lung cancer nodules
more, for training and testing groups, various (22). Mobedarsany et al showed that AI was
machine learning techniques are effective and proven to be more accurate than surgical patho­
accurate in distinguishing healthy cells from in- logists in predicting patient outcomes. This study
fected ones. Artificial intelligence approaches gives light on the application of deep learning in

Maedica A Journal of Clinical Medicine, Volume 17, No. 2, 2022 423


Integration of Artificial Intelligence (AI)

medicine as well as the integration of histology tures, allowing for automated study of histology
and genetic data, and ways for dealing with chal- slides. This scanning, along with deep learning
lenges such intratumoral heterogeneity (23). algorithms, enables for the automated detection
Muneer et al, on the other hand, employed AI of lesions based on previously confirmed regions
approaches to identify glioma grade and their of interest (33). In a simulated setting, machine
findings were good, with better than 90% accu- learning algorithms produced potentially quicker
racy (24). Yala et al created a big database by and more accurate diagnoses than 11 patholo-
using a machine learning algorithm to identify gists, according to a new research (15). The fu-
important tumour features from breast pathology ture laboratory will be more automated and
reports (25). The algorithm-based smartphone dominated by robotics, as well as more net-
software "Skinvision" may assist a user through worked to take use of benefits provided by AI
performing frequent self-checks for skin cancer and the internet of things (34).
using a phone and a snapshot of a skin spot. The Artificial intelligence has also several admini­
algorithm, like a doctor, can determine the te­x­ strative applications in healthcare. These are re-
ture, colour, and form of the lesions. Users get an quired in healthcare since a typical US nurse, for
instant risk assessment for skin lesions in 30 se­ instance, spends 25% of her time on administra-
conds, and the algorithm has been shown to de- tive and regulatory tasks. RPA is the technology
tect 95% of skin cancer at an early stage (26). that would be most likely to be applicable to this
IBM's Watson has recently received a lot of goal. It has a wide range of healthcare applica-
attention for its focus on precision medicine, tions, including billing process, claims processing
particularly cancer detection and treatment. and clinical documentation management Based
Watson employs a mix of machine learning and on the DL NHS 111 algorithm, the AI-based cli­
natural language processing techniques. Users nical assessment service "National health service"-
discovered how difficult it was to train Watson NHS 24 is in the clinical testing phase in Scotland
how to handle certain forms of cancer and inte- to assist people with minor health concerns at
grate it into care procedures and systems, thus home through telephonic contact (36). Likewise,
initial excitement for this implementation of the another virtual care company, "Babylon Health,"
technology diminished (27, 28). Recently, Google uses semantic web technology to deliver com-
is working with health-care delivery networks to plementary digital services to improve clinical
develop big-data prediction algorithms that will results. The semantic web aims to make internet
alert physicians to high-risk illnesses like sepsis data machine-readable. Develop a clinical LDG
and heart failure (29). Nvidia, a renowned inter- (Linked Data Graph) to integrate various bioin-
national technology firm located in the United formatics-based biomedical data banks in a way
States, has revealed its ambition to create an AI that is understandable to the average individual
supercomputer for medical research and medi- who uses AI-based medical services (37).
cine delivery in November 2020 (30, 31). Before AI may become widespread in medi-
In recent years, the demand for automated cine, the issue of legal accountability must be
laboratory recommendation systems has risen in settled, particularly in imaging disciplines such as
order to provide more accurate and faster dia­ radiology and pathology. The lack of clarity and
gnosis. By incorporating image analysis and ma- misunderstanding around important concerns
chine learning into ordinary surgical pathology, such as the processing of sensitive personal infor-
the digital revolution is revolutionising the prac- mation and data gathering, consent, transparen-
tise of diagnostic surgical pathology. An automat- cy, storage and other difficulties further compli-
ed recommendation might help save healthcare cates and obscures the question of legal
resources by improving the accuracy and effi- accountability for AI-based judgments in medi-
ciency of test requests. In a study, it was shown cine. According to a recent study, depending too
that with limited variables from EHR, the DL much on decision support systems in radiology
model had the greatest discriminating ability for led to a higher rate of false negative diagnoses
all diagnostic laboratory tests, with a mean than when the computer-aided diagnostic sys-
AUROC micro of 0.98 and AUROC macro of tem was unavailable to the same group of radi-
0.94, respectively (32). WSI scanners can now ologists (38). In a simulated setting, machine
capture and store slides in the form of digital pic- learning algorithms produced potentially quicker

424 Maedica A Journal of Clinical Medicine, Volume 17, No. 2, 2022


Integration of Artificial Intelligence (AI)

and more accurate diagnoses than 11 patholo- used to prioritize novel drug targets. Such me­
gists according to a new study (15). thods can identify combinations of genetic va­
Similarly, bringing AI to clinical decision-ma­ riants or abnormalities that cause disease, inclu­
king also comes with a slew of difficulties. One ding cases where causal genes are either known
such challenge is equitable decision-making that or unknown. The progress in data integration
does not discriminate based on biases contained combined with novel AI/ML algorithms and di­
in datasets and methodologies used to construct sease causality modeling will probably shift the
the system. Some AI models lack of transparen- paradigm and establish unbiased ways for target
cy, which can have no cognitive similarity to the selection and prioritization (41). q
problem they are tackling, might exacerbate
these problems. Understanding the reasons be- CONCLUSION
hind a model's choice or suggestion is frequently
a difficult task. For instance, if a decision-support
tool suggests that a patient's total knee arthro- T echnological innovation in healthcare is ac-
celerating and has become increasingly inter-
woven into both our daily lives and medical
plasty surgery be postponed, the patient and
his/her care team might appreciate knowing the practice, such as smart health trackers and dia­
variables that led to that conclusion (39, 40). gnostic algorithms. Deep learning is now being
The laboratory of the future is expected to be used to address an increasing number of specia­
increasingly automated and dominated by ro- lised jobs in medicine. Algorithm help has the
botics as well as more networked with the use of ability to increase not only the sensitivity and ac-
AI and the benefits of internet of things (34). Ar- curacy of diagnoses, but also the turnaround
tificial intelligence can be applied across any time, according to the above-presented stu­ -
stream of medical field, including and not limi­ dies. q
ted to clinical and laboratory diagnosis only. One Conflicts of interest: none declared.
such another example is MetaPath that can be Financial support: none declared.

References
1. Smith KP, Kirby JE. Image analysis Haleem A. Artificial Intelligence (AI) Convolutional Neural Network.
and artificial intelligence in infectious applications for COVID-19 pandemic. J Clin Microbiol 2018;56:e01521-17.
disease diagnostics. Diabetes Metab Syndr 2020;14:337-339. 13. Mathison BA, Kohan JL, Walker JF, et al.
Clin Microbiol Infect Off Publ Eur Soc Clin 8. 4 Robotic Process Automation (RPA) Detection of Intestinal Protozoa in
Microbiol Infect Dis 2020;26:1318-1323. Applications in the Healthcare Industry | by Trichrome-Stained Stool Specimens by
2. Jiang F, Jiang Y, Zhi H, et al. Artificial Karl Utermohlen | Medium [Internet]. Use of a Deep Convolutional Neural
intelligence in healthcare: past, [cited 2022 May 4]. Available from: Network.
present and future. https://medium.com/@karl.utermohlen/4- J Clin Microbiol 2020;58:e02053-19.
Stroke Vasc Neurol 2017;2:230-243. robotic-process-automation-rpa- 14. Weis CV, Jutzeler CR, Borgwardt K.
3. Davenport T, Kalakota R. The potential applications-in-the-healthcare-industry- Machine learning for microbial
for artificial intelligence in healthcare. 4d449b24b613 identification and antimicrobial
Future Healthc J 2019;6:94. 9. Davenport TH, Glaser J. Just-in-time susceptibility testing on MALDI-TOF
4. Kaur I, Behl T, Aleya L, et al. Artificial delivery comes to knowledge mass spectra: a systematic review.
intelligence as a fundamental tool in management. Clin Microbiol Infect Off Publ Eur Soc Clin
management of infectious diseases and Harv Bus Rev 2002;80:107-111, 126. Microbiol Infect Dis 2020;26:1310-1317.
its current implementation in 10. Ye J. The Role of Health Technology and 15. Ahmad Z, Rahim S, Zubair M,
COVID-19 pandemic. Informatics in a Global Public Health Abdul-Ghafar J. Artificial intelligence
Environ Sci Pollut Res Int Emergency: Practices and Implications (AI) in medicine, current applications
2021;28:40515-40532. From the COVID-19 Pandemic. and future role with special emphasis on
5. Li L, Qin L, Xu Z, et al. Using Artificial JMIR Med Inform 2020;8:e19866. its potential and promise in pathology:
Intelligence to Detect COVID-19 and 11. Peiffer-Smadja N, Dellière S, present and future impact, obstacles
Community-acquired Pneumonia Based Rodriguez C, et al. Machine learning in including costs and acceptance among
on Pulmonary CT: Evaluation of the the clinical microbiology laboratory: has pathologists, practical and philosophical
Diagnostic Accuracy. the time come for routine practice? considerations. A comprehensive review.
Radiology 2020;296:E65-E71. Clin Microbiol Infect Off Publ Eur Soc Clin Diagn Pathol [Internet]. 2021 [cited 2022
6. Agrebi S, Larbi A. Use of artificial Microbiol Infect Dis 2020;26:1300-1309. Mar 17];16. Available from:
intelligence in infectious diseases. 12. Smith KP, Kang AD, Kirby JE. https://www.ncbi.nlm.nih.gov/labs/pmc/
Artif Intell Precis Health 2020;415-438. Automated Interpretation of Blood articles/PMC7971952/
7. Vaishya R, Javaid M, Khan IH, Culture Gram Stains by Use of a Deep 16. Go T, Kim JH, Byeon H, Lee SJ.

Maedica A Journal of Clinical Medicine, Volume 17, No. 2, 2022 425


Integration of Artificial Intelligence (AI)

Machine learning-based in-line 2018;115:E2970-E2979. 33. Zemouri R, Devalland C,


holographic sensing of unstained 24. Ahammed Muneer K Vnull, Valmary-Degano S, Zerhouni N.
malaria-infected red blood cells. Rajendran VR, K PJ. Glioma Tumor [Neural network: A future in pathology?].
J Biophotonics 2018;11:e201800101. Grade Identification Using Artificial Ann Pathol 2019;39:119-129.
17. Colubri A, Silver T, Fradet T, et al. Intelligent Techniques. 34. Kricka LJ. History of disruptions in
Transforming Clinical Data into J Med Syst 2019;43:113. laboratory medicine: what have we
Actionable Prognosis Models: 25. Yala A, Barzilay R, Salama L, et al. learned from predictions?
Machine-Learning Framework and Using machine learning to parse breast Clin Chem Lab Med
Field-Deployable App to Predict pathology reports. 2019;57:308-311.
Outcome of Ebola Patients. Breast Cancer Res Treat 2017;161:203-211. 35. Nurses Say Distractions Cut
PLoS Negl Trop Dis 2016;10:e0004549. 26. Malhi IS, Yiu ZZN. Algorithm-based Bedside Time by 25% | HealthLeaders
18. Kaur I, Behl T, Aleya L, et al. Artificial smartphone apps to assess risk of skin Media [Internet]. [cited 2022 May 4].
intelligence as a fundamental tool in cancer in adults: critical appraisal of a Available from:
management of infectious diseases and systematic review. https://www.healthleadersmedia.com/
its current implementation in Br J Dermatol 2021;184:638-639. nursing/nurses-say-distractions-cut-
COVID-19 pandemic. 27. IBM pitched Watson as a revolution in bedside-time-25
Environ Sci Pollut Res cancer care. It’s nowhere close [Internet]. 36. McCartney M. Margaret McCartney: AI
2021;28:40515-40532. [cited 2022 May 4]. Available from: in medicine must be rigorously tested.
19. Hirasawa T, Aoyama K, Tanimoto T, https://www.statnews.com/2017/09/05/ BMJ 2018;361:k1752.
et al. Application of artificial intelligence watson-ibm-cancer/ 37. Supporting Digital Healthcare Services
using a convolutional neural network for 28. Rule-based expert systems: The mycin Using Semantic Web Technologies
detecting gastric cancer in endoscopic experiments of the stanford heuristic « ISWC 2018 [Internet].
images. programming project: B.G. Buchanan and [cited 2022 May 4]. Available from:
Gastric Cancer Off J Int Gastric Cancer E.H. Shortliffe, (Addison-Wesley, http://iswc2018.semanticweb.org/
Assoc Jpn Gastric Cancer Assoc Reading, MA, 1984); 702 pages, $40.50 - sessions/supporting-digital-healthcare-
2018;21:653-660. ScienceDirect [Internet]. services-using-semantic-web-
20. Gulshan V, Peng L, Coram M, et al. [cited 2022 May 4]. Available from: technologies/index.html
Development and Validation of a https://www.sciencedirect.com/science/ 38. Alberdi E, Povykalo A, Strigini L,
Deep Learning Algorithm for Detection article/abs/pii/0004370285900670 Ayton P. Effects of incorrect computer-
of Diabetic Retinopathy in Retinal 29. Rysavy M. Evidence-based medicine: a aided detection (CAD) output on
Fundus Photographs. science of uncertainty and an art of decision-making in mammography.
JAMA 2016;316:2402-2410. probability. Acad Radiol 2004;11:909-918.
21. Lee JH, Ha EJ, Kim JH. Application of Virtual Mentor VM 2013;15:4-8. 39. Gilpin LH, Bau D, Yuan BZ, et al.
deep learning to the diagnosis of cervical 30. Buitrago PA, Nystrom NA, Gupta R, Explaining Explanations: An Overview of
lymph node metastasis from thyroid Saltz J. Delivering Scalable Deep Interpretability of Machine Learning. In:
cancer with CT. Learning to Research with Bridges-AI. 2018 IEEE 5th International Conference on
Eur Radiol 2019;29:5452-5457. In: CARLA, 2019. Data Science and Advanced Analytics
22. [Clinical Application of Artificial 31. Kochanny SE, Pearson AT. Academics (DSAA), 2018. pp 80-89.
Intelligence Recognition Technology in as leaders in the cancer artificial 40. Surmacz K, Kamath AF, Andel DV.
the Diagnosis of Stage T1 Lung Cancer]. intelligence revolution. Fairness in AI: How Can We Avoid Bias
- Abstract - Europe PMC [Internet]. [cited Cancer 2021;127:664-671. and Disparities in Orthopedic
2022 Mar 17]. Available from: 32. The performance of DL model based on Applications of Artificial Intelligence?
https://europepmc.org/article/ varying cut-offs for clinical... [Internet]. J Orthop Exp Innov 2021;4:25901.
med/31109442 ResearchGate. [cited 2022 Mar 17]. 41. Wan G, Du B, Pan S, Haffari G.
23. Mobadersany P, Yousefi S, Amgad M, Available from: Reinforcement Learning Based Meta-Path
et al. Predicting cancer outcomes from https://www.researchgate.net/figure/ Discovery in Large-Scale Heterogeneous
histology and genomics using The-performance-of-DL-model-based-on- Information Networks.
convolutional networks. varying-cut-offs-for-clinical-laboratory- Proc AAAI Conf Artif Intell
Proc Natl Acad Sci USA test_tbl1_351969929 2020;34:6094-6101.

426 Maedica A Journal of Clinical Medicine, Volume 17, No. 2, 2022

You might also like