Table 1 Comparative analysis of extracted information from the included studies

[18]

Prostate cancer

2010–2018

5461

patients

NLTK [33]

SVM [34],

Rule-based algorithms,

ConText [35],

NegEx [36]

Imputation, Vectorization

Early-stage cancer patients

Clinical & pathological TNM staging

[19]

Ophthalmology

2013–2016

286 visits

R [37], Mobile devices, Numbers [38]

Bespoke Algorithms

Data drop

Ophthalmology outpatient

Clinical workflow analysis

[20]

Ophthalmology

2015–2016

8,703 visits

R 3.4.3 [37]

Linear regression [39],

RF [40]

Rule And

Condition

Pediatric ophthalmology outpatient

Outpatient

visit length

[21]

Non-small cell lung cancer

2010–2018

794 patients

Scikit-Learn 0.24.1

[41], LightGBM 3.2.0 [42],

SciPy 1.6.2 [43],

BERT [44]

Logistic regression [45],

RF [40], SVM [34],

Deep neural network [46]

NER, Rule-Based, NLP Relation Classification, Postprocessing Modules

CT-scanned non-small cell lung cancer patients

Preoperative prediction of lymph node metastasis

[22]

Type II diabetes

-

997 patients

Python 3.6,

PyTorch 1.0 [47],

NVIDIA Titan X GPU,

CUDA 9.0 [48], PyPhewas [49]

ADAM [50], 3D UNet [51], Fuzzy C means [52], Convolutional neural network

Segmentation & Slicing, Feature Extraction & Normalization, Annotation

CT scanned patients

with and without diabetes

Early

Detection of type II

diabetes

mellitus

[23]

Acute ischemic stroke

1992–2019

6,136, 686 patients

OHDSI tool [53], R [37], OMOP CDM [54]

Lasso logistic regression [55]

Rule-based processing

Patients

aged 45 + with first ischemic stroke

Early

prediction of

symptomatic intracerebral hemorrhage

[24]

Nasopharyngeal cancer

2008–2018

54,703 patients

-

-

ETL, Data Structurization & Normalization

Nasopharyngeal carcinoma patient receiving treatment

Platform development

for retrospective clinical

studies

[25]

No specific disease

1980–2014

704,587 patients

NCBO BioPortal [56],

Open Biomedical Annotator [57]

RF [40], PCA [58], GMM [59],

K-Means, ICA [60],

Multi-Layer Neural Network [46], LDA [61], SDA [62],

NegEx [36]

Denoising, Topic Modelling,

Negation

Patients with one recorded ICD code

Onset of

disease based

on EHRs

[26]

Cancer

1996–2012

7000 reports

Weka Software

3.6.11 [63],

Perl Lingua Stem module [64],

SAS 9.4 [65], MetaMap [66]

Logistic regression [45], Naive Bayes [67], K–NN [68],

RF [40],

J48 decision

tree [69],

NegEx [36]

Kullback-Leibler [70], NER,

Dictionary and Non-dictionary approach,

Rule-based

classifier

Patients with a recorded clinical note

Detect cancer cases using plaintext medical data

[27]

Inpatient Accidental Falling

2010–2014

46,241

patients

Ubuntu 14.04 LTS [71], R 3.1.2 [37], lme4 package [72], Epi [73]

Multilevel Logistic Regression [74]

Transformation, Mapping Values

Hospitalized inpatients with recorded data

Predict fall

risk to prevent injury

[28]

Pediatric Care

2008–2013

149,604

visits

Excel 2010 [75], Access 2010 [76]

-

Statistical

Analysis, Correlation, Interpolation

Pediatric physician visits

Compute physician & departmental performance

[29]

No specific: Evaluated in Colorectal Cancer

-

*20346 visits

LinkEHR [77, 78], XML [79],

Semantic tool [80], Saxon [81],

OWL [82], NCBO BioPortal [56], Protégé [83],

Hermit Reasoner

[84], UMLS [85], OpenEHR [86], SNOMED CT [87], SPARQL [88]

Bespoke

phenotyping algorithm,

Ontology mapping, Semantic Reasoning

Semantic Representation, Standardization

Colorectal cancer patients

Identification

of patient cohorts

[30]

No specific: Evaluated in HIV,

hepatitis C,

lab measurements

-

**Multiple

CogStack [89], Bio-YODIE [90], Elasticsearch [91], UMLS [92, 93], SPARQL [88],

SNOMED CT [87]

Bidirectional

recurrent neural network [94]

NER,

Normalization, Semantic Indexing

& Computation

Negation,

Indexing

Pertinent

clinical notes

for target use cases

Customized care, trial recruitment,

and research