Adds functions to call Comprehend Medical's ontology linking APIs

DESCRIPTION

@@ -9,7 +9,9 @@ Authors@R: c(person("Thomas J.", "Leeper", role = c("aut"),
              person("Antoine", "Sachet", role = c("aut", "cre"),
                     email = "antoine.sac@gmail.com"),
              person("Dave", "Kincaid", role = c("ctb"),
-                    email = "dave@kincaidlabs.ai"))
+                    email = "dave@kincaidlabs.ai"),
+             person("Robert", "Norberg", role = c("ctb"),
+                    email = "Robert.Norberg@moffitt.org"))
 Description: Client for 'AWS Comprehend' <https://aws.amazon.com/comprehend>, a cloud natural language processing service that can perform a number of quantitative text analyses, including language detection, sentiment analysis, and feature extraction.
 License: GPL (>= 2)
 URL: https://github.com/cloudyr/aws.comprehend

NAMESPACE

@@ -8,6 +8,9 @@ export(detect_medical_phi)
 export(detect_phrases)
 export(detect_sentiment)
 export(detect_syntax)
+export(infer_icd10)
+export(infer_rxnorm)
+export(infer_snowmed_ct)
 import(httr)
 importFrom(aws.signature,locate_credentials)
 importFrom(aws.signature,signature_v4_auth)

NEWS.md

@@ -1,5 +1,7 @@
 # aws.comprehend (development version)
 
+* Added `infer_icd10`, `infer_rxnorm`, and `infer_snowmed_ct` functions to call Comprehend Medical's ontology linking APIs.
+
 # aws.comprehend 0.2.1
 
 * Released on CRAN 2020-03-18

R/infer_icd10.R

@@ -0,0 +1,22 @@
+#' @title Detect conditions and associated ICD10 codes in a source medical text
+#' @description Detect detect possible medical conditions as entities and link them to ICD10 codes in a source medical text
+#' @param text A character string containing a text to entities analyze, or a character vector to perform analysis separately for each element.
+#' @param language A character string containing a two-letter language code. Currently only \dQuote{en} is supported.
+#' @param \dots Additional arguments passed to \code{\link{comprehendHTTP}}.
+#' @return A data frame
+#' @references \href{https://docs.aws.amazon.com/comprehend-medical/latest/dev/ontology-icd10.html}{AWS Comprehend Medical Developer Guide}
+#' @examples
+#' \dontrun{
+#'   # simple example
+#'   infer_icd10("Mrs. Smith comes in today complaining of shortness of breath.")
+#'
+#'   txt <-c("Mrs. Smith comes in today.",
+#'           "She is complaining of shortnesss of breath.")
+#'   infer_icd10(txt)
+#' }
+#' @export
+infer_icd10 <- function(text, language = "en", ...) {
+    bod <- list(Text = text, LanguageCode = language)
+    out <- comprehendHTTP(action = "InferICD10CM", body = bod, service = "comprehendmedical", ...)
+    return(cbind(Index = 0, out$Entities))
+  }

R/infer_rxnorm.R

@@ -0,0 +1,22 @@
+#' @title Identify medications in a source medical text
+#' @description Identify medications in a source medical text and link them to RxCUI codes
+#' @param text A character string containing a text to entities analyze, or a character vector to perform analysis separately for each element.
+#' @param language A character string containing a two-letter language code. Currently only \dQuote{en} is supported.
+#' @param \dots Additional arguments passed to \code{\link{comprehendHTTP}}.
+#' @return A data frame
+#' @references \href{https://docs.aws.amazon.com/comprehend-medical/latest/dev/ontology-RxNorm.html}{AWS Comprehend Medical Developer Guide}
+#' @examples
+#' \dontrun{
+#'   # simple example
+#'   infer_rxnorm("fluoride topical ( fluoride 1.1 % topical gel ) 1 application Topically daily. Patient is not on warfarin.")
+#'
+#'   txt <-c("fluoride topical ( fluoride 1.1 % topical gel ) 1 application Topically daily.",
+#'           "Patient is not on warfarin.")
+#'   infer_rxnorm(txt)
+#' }
+#' @export
+infer_rxnorm <- function(text, language = "en", ...) {
+  bod <- list(Text = text, LanguageCode = language)
+  out <- comprehendHTTP(action = "InferRxNorm", body = bod, service = "comprehendmedical", ...)
+  return(cbind(Index = 0, out$Entities))
+}

R/infer_snowmed_ct.R

@@ -0,0 +1,22 @@
+#' @title Identify clinical terms in a source medical text
+#' @description Identify clinical terms in a source medical text and link them to Systematized Nomenclature of Medicine, Clinical Terms (SNOMED CT) codes
+#' @param text A character string containing a text to entities analyze, or a character vector to perform analysis separately for each element.
+#' @param language A character string containing a two-letter language code. Currently only \dQuote{en} is supported.
+#' @param \dots Additional arguments passed to \code{\link{comprehendHTTP}}.
+#' @return A data frame
+#' @references \href{https://docs.aws.amazon.com/comprehend-medical/latest/dev/ontology-linking-snomed.html}{AWS Comprehend Medical Developer Guide}
+#' @examples
+#' \dontrun{
+#'   # simple example
+#'   infer_snowmed_ct("BHEENT : Boggy inferior turbinates. No oropharyngeal lesion.")
+#'
+#'   txt <-c("BHEENT : Boggy inferior turbinates.",
+#'           "No oropharyngeal lesion.")
+#'   infer_snowmed_ct(txt)
+#' }
+#' @export
+infer_snowmed_ct <- function(text, language = "en", ...) {
+  bod <- list(Text = text, LanguageCode = language)
+  out <- comprehendHTTP(action = "InferSNOMEDCT", body = bod, service = "comprehendmedical", ...)
+  return(cbind(Index = 0, out$Entities))
+}

tests/testthat/helper-medical_mocks.R

@@ -20,7 +20,10 @@ generate_mocks_medical <- function() {
   actions <- list(
     "DetectPHI" = body$medical,
     "DetectEntities" = body$medical,
-    "DetectEntitiesV2" = body$medical
+    "DetectEntitiesV2" = body$medical,
+    "InferICD10CM" = body$medical,
+    "InferRxNorm" = body$medical,
+    "InferSNOMEDCT" = body$medical
   )
 
   mapply(generate_mock_medical, names(actions), actions)

tests/testthat/test-infer_icd10.R

@@ -0,0 +1,35 @@
+context("infer_icd10")
+
+body = get_request_body()
+
+test_that("infer_icd10 works on single string", {
+  output <- with_mock(
+    comprehendHTTP = mock_comprehendHTTP_medical,
+    infer_icd10(text = body$medical$Text,
+                       language = body$medical$LanguageCode)
+  )
+
+  icd10_codes <- Reduce(rbind, output$ICD10CMConcepts)
+  expected_icd10_codes <- read.table(sep = "\t", text = "
+  Code	Description	Score
+  R45.83	Excessive crying of child, adolescent or adult	0.737255275249481
+  G47.9	Sleep disorder, unspecified	0.592359900474548
+  Z72.821	Inadequate sleep hygiene	0.291808724403381
+  Y93.84	Activity, sleeping	0.153297811746597
+  F51.9	Sleep disorder not due to a substance or known physiological condition, unspecified	0.146037772297859",
+  header = TRUE, stringsAsFactors = FALSE, strip.white = TRUE)
+  expect_similar(icd10_codes, expected_icd10_codes)
+
+  # These columns are lists of data.frames - complicated to test
+  output$Attributes <- NULL
+  output$ICD10CMConcepts <- NULL
+  output$Traits <- NULL
+
+  expected <- read.table(sep = "\t", text = "
+  Index	BeginOffset	Category	EndOffset	Id	Score	Text	Type
+  0	45	MEDICAL_CONDITION	61	1	0.678973436355591	Sleeping trouble	DX_NAME",
+  header = TRUE, stringsAsFactors = FALSE)
+
+  expect_similar(output, expected)
+
+})

tests/testthat/test-infer_rxnorm.R

@@ -0,0 +1,36 @@
+context("infer_rxnorm")
+
+body = get_request_body()
+
+test_that("infer_rxnorm works on single string", {
+  output <- with_mock(
+    comprehendHTTP = mock_comprehendHTTP_medical,
+    infer_rxnorm(text = body$medical$Text,
+                language = body$medical$LanguageCode)
+  )
+
+  rxcui_codes <- Reduce(rbind, output$RxNormConcepts)
+  expected_rxcui_codes <- read.table(sep = "\t", text = "
+  Code	Description	Score
+  2599	clonidine	0.840667128562927
+  884185	clonidine hydrochloride 0.2 mg oral tablet	0.208982422947884
+  884173	clonidine hydrochloride 0.1 mg oral tablet	0.189939811825752
+  216094	clinidine	0.129013881087303
+  884225	10 ml clonidine hydrochloride 0.5 mg/ml injection	0.0852060243487358",
+  colClasses = c("Code" = "character", "Description" = "character", "Score" = "numeric"),
+  header = TRUE, stringsAsFactors = FALSE, strip.white = TRUE)
+  expect_similar(rxcui_codes, expected_rxcui_codes)
+
+  # These columns are lists of data.frames - complicated to test
+  output$Attributes <- NULL
+  output$RxNormConcepts <- NULL
+  output$Traits <- NULL
+
+  expected <- read.table(sep = "\t", text = "
+  Index	BeginOffset	Category	EndOffset	Id	Score	Text	Type
+  0	83	MEDICATION	92	1	0.997781932353973	Clonidine	GENERIC_NAME",
+  header = TRUE, stringsAsFactors = FALSE)
+
+  expect_similar(output, expected)
+
+})

tests/testthat/test-infer_snowmed_ct.R

@@ -0,0 +1,36 @@
+context("infer_snowmed_ct")
+
+body = get_request_body()
+
+test_that("infer_snowmed_ct works on single string", {
+  output <- with_mock(
+    comprehendHTTP = mock_comprehendHTTP_medical,
+    infer_snowmed_ct(text = body$medical$Text,
+                 language = body$medical$LanguageCode)
+  )
+
+  snowmed_codes <- Reduce(rbind, output$SNOMEDCTConcepts)
+  expected_snowmed_codes <- read.table(sep = "\t", text = "
+  Code	Description	Score
+  301345002	Difficulty sleeping (finding)	0.0105650639161468
+  39898005	Sleep disorder (disorder)	0.00971199851483107
+  26677001	Sleep pattern disturbance (finding)	0.00622155796736479
+  89675003	Sleep terror disorder (disorder)	0.00587156973779202
+  248259004	Symptoms interfere with sleep (disorder)	0.00511229783296585",
+  colClasses = c("Code" = "character", "Description" = "character", "Score" = "numeric"),
+  header = TRUE, stringsAsFactors = FALSE, strip.white = TRUE)
+  expect_similar(snowmed_codes, expected_snowmed_codes)
+
+  # These columns are lists of data.frames - complicated to test
+  output$Attributes <- NULL
+  output$SNOMEDCTConcepts <- NULL
+  output$Traits <- NULL
+
+  expected <- read.table(sep = "\t", text = "
+  Index	BeginOffset	Category	EndOffset	Id	Score	Text	Type
+  0	45	MEDICAL_CONDITION	61	1	0.678973436355591	Sleeping trouble	DX_NAME",
+  header = TRUE, stringsAsFactors = FALSE)
+
+  expect_similar(output, expected)
+
+})