Table 1 Clinical application areas and their definitions
Application area (number of papers) | Description | Subcategory | Description of included papers | Technical task (number of papers) | Anatomical scan region [Paper number—a full list of the papers by number is included in Additional file 1] |
|---|---|---|---|---|---|
Surveillance (45) | Using imaging reports for surveillance of disease at a population health or individual level either longitudinally or generating alerts | Disease surveillance | Monitoring occurrence of infectious disease Monitoring non-communicable disease patterns, including alerts for conditions | Information Extraction (3) | Thorax [25] Mixed [26] Cerebrovascular [19] |
Classification (4) | Thorax [27, 28] Abdomen [29] Mixed [30] | ||||
Prioritising reports | Generating alerts for reports requiring more urgent action | Classification (4) | Other [31] Mixed [32] Unspecified [33, 34] | ||
Incidental findings | Generating alerts for incidental findings | Information Extraction (1) | Mixed [23] | ||
Classification (3) | Thorax [20, 21] Cerebrovascular [24] | ||||
Patient surveillance | Pairing measurements and linking reports to Investigate/monitor conditions over time e.g. worsening prognosis/ response to treatment | Information Extraction (21) | Thorax [35] Abdomen [6, 8, 16–18] Mixed [1, 2, 9, 36, 37, 131] Breast [3, 10, 11, 13–15] Unspecified [7, 38, 39] | ||
Classification (5) | Thorax [4, 40] Breast [12] Mixed [5, 41] | ||||
Follow-up | Detecting follow-up recommendations, creating alerts and linking to see if carried out | Information Extraction (3) | Abdomen [42] Unspecified [43] Mixed [44] | ||
Classification (1) | Unspecified [45] | ||||
Disease information and classification (46) | Using imaging reports to identify information that may also be aggregated according to classification systems (no specific clinical purpose specified) | N/A | Extracting information about a disease/condition/function (e.g. LVEF) (no additional processing required) Staging e.g. using BIRADS or Lung-RADS Identifying sub-types of disease Classification of fractures Predicting ICD codes ICD codes used for ground truth | Information Extraction (14) | Cerebrovascular [47,54,55] Breast [59,65] Abdomen [66] Thorax [63,67–70] Mixed [71,72] Unspecified [73] |
Classification (32) | Cerebrovascular [46, 48–53, 74–76] Abdomen [77] Breast [56–58] Extremities [78–84] Mixed [85–89] Spine [22, 90] Thorax [60–62, 64] | ||||
Language discovery and knowledge structure (27) | Investigating the structure of language in imaging reports and ways in which this may be optimised to facilitate knowledge and decision support, communication (both internally between clinicians and outward facing communication with patients/public), and assist in improving NLP applications | Knowledge support for patients/public | Improving readability of reports and communications for public/patients | Lexicon/ontology discovery (3) | Mixed [91, 92, 99] |
Knowledge and decision support for clinicians | Providing information for clinician use (including using ontologies and lexicons) Finding relevant reports Improving reading efficiency Supporting radiological and clinical decision making Supporting clinician education | Clustering (1) | Mixed [100] | ||
Information Extraction (8) | Breast [94] Mixed [101–104] Thorax [93] Cerebrovascular [105] Unspecified [106] | ||||
Lexicon/Ontology (1) | Unspecified [107] | ||||
Variability, complexity and structure of language for NLP purposes | Investigating variability and complexity of language including free-test and structured reports Improving structure of language for NLP e.g. normalising phrases to support classification Normalising and disease specific phrases | Information Extraction (6) | Thorax [96, 108] Mixed [97, 109, 110] Unspecified [98] | ||
Lexicon/ontology discovery (4) | Unspecified [111, 112] Spine [113] Breast [114] | ||||
Classification (4) | Unspecified [95] Thorax) [115–117] | ||||
Quality and compliance (20) | Using imaging reports to assess quality and safety of radiology practice, clinical practice, and efficiency of healthcare services | Assessing imaging practices | Do imaging practices adhere to guidance including indications and protocol selection Impact of guideline changes on imaging practice Assessing imaging utilisation and yield | Information Extraction (4) | Abdomen [121] Thorax [122] Mixed [119, 123] |
Classification (11) | Mixed [118, 120, 125, 135] Cerebrovascular [124, 126, 127] Thorax [128, 129] Abdomen [130] Extremities [136] | ||||
Audit | Classification used for quality improvement in radiology and clinical practice Identifying reports for auditing Identifying and fixing errors in reports (e.g. gender/laterality) | Information Extraction (2) | Mixed [134] Cerebrovascular [133] | ||
Classification (3) | Thorax [132] Breast [137] Extremities [138] | ||||
Research (16) | Using imaging reports to create patient cohorts for research purposes | Cohort | Identifying cohorts for research purposes with specific medical conditions (sometimes in specified anatomical regions), with particular radiological findings, who have had certain healthcare interactions | Classification (7) | Abdomen [140–142] Cerebrovascular [144, 152] Mixed [151] Spine [150] |
Information Extraction (3) | Unspecified [153] Cerebrovascular [145] Spine [143] | ||||
Epidemiology | Identifying research cohorts as above but papers in which they go on to do epidemiological analyses and present their results | Information extraction (4) | Unspecified [154] Cerebrovascular [148] Abdomen [139] Thorax [147] | ||
Classification (2) | Mixed [146] Thorax [149] | ||||
Technical NLP (10) | Papers which do not fit to a specific category, often with a primarily technical aim | N/A | Studies encompassed a variety of purposes, such as negation detection, spelling correction, fact checking, methods for sample selection, crowd source annotation | Information Extraction (6) | Mixed [155,157,159] Thorax [160] Unspecified [161, 162] |
Classification (4) | Cerebrovascular [158] Mixed [163] Thorax [156] Unspecified [164] |