Skip to main content

Table 1 Summary of NLP studies focusing on actionable radiology reports (ML: Machine Learning, DL: Deep Learning, BERT: Bidirectional Encoding Representation of Transformer).

From: Automatic text classification of actionable radiology reports of tinnitus patients using bidirectional encoder representations from transformer (BERT) and in-domain pre-training (IDPT)

Author(s)

Language

Number of radiology reports

Algorithm

Section of report

Research objective

Carrodeguas et al. [33]

English

2306

ML/DL

Impression

Classifying recommendation

Helibrun et al. [34]

English

851

Rule-based

Impression

Detecting critical finding

Lou et al. [35]

English

6000

ML

Not mentioned

Classifying recommendation

Esteban et al. [36]

English

3401

Software

Findings, impression

Classifying recommendation

Morioka et al. [37]

English

1402

Rule-based

Not mentioned

Classifying disease condition

Fu et al. [38]

English

1000

Rule-based ML/DL

Not mentioned

Classifying disease condition

Nakamura et al. [39]

Japanese

63646

BERT

Order, findings, impression

Detecting critical finding

Jujjavarapu et al. [40]

English

871

ML

Not mentioned

Classifying disease condition

Liu et al.. [15]

Chinese

1089

BERT/ML

Findings

Classifying disease condition

Zhang et al. [41]

Chinese

359

BERT Pre-training

Findings

Classifying disease condition

Zaman et al. [42]

English

1503

BERT Pre-training

Findings

Classifying disease condition

Liu et al.. [43]

English

594

BERT

Not mentioned

Classifying certainty

Proposed study

Chinese

5864

BERT Pre-training DL

Findings

Classifying disease condition