@precisa-saude/fhir-ocr-utils 0.37.2 → 0.38.1

package/README.md

@@ -83,6 +83,20 @@ Por isso a correspondência é conservadora:
   a linha carregar um valor: um número, uma unidade conhecida, ou um termo
   qualitativo esperado ("Negativo", "Ausente", "Amarelo Citrino"). "Specimen
   type: Blood" não ancora; "Sangue Oculto: Negativo" ancora.
+- **Seção de urinálise** — debaixo de um cabeçalho de urinálise
+  ("URINALYSIS", "Urina tipo I", "Rotina de urina", "EAS", "Urinálise"), os
+  nomes nus do exame de urina passam a ser o analito da urina: "GLUCOSE"
+  ancora `Glucose_Urine`, "WBC" ancora `Leukocytes_Urine`, "RBC" ancora
+  `RBC_Urine`, "PH" ancora `pH_Urine`, e "COLOR", "KETONES", "PROTEIN" e
+  companhia ancoram mesmo sem valor na linha (texto em colunas). O código do
+  sangue não ancora pela mesma linha. A seção acaba na primeira linha com
+  exame de outro painel, no cabeçalho de outro painel ou no fim do texto. Uma
+  linha sem valor e sem nome conhecido pode ser as duas coisas ("COMPREHENSIVE
+  METABOLIC PANEL" ou "MUCUS"), e quem decide é a próxima linha decisiva: se só
+  pode ser da urina (código da urina, a palavra urina, ou /HPF e /LPF), a seção
+  segue; se é exame de outro painel, ou se o texto acaba sem nada decisivo, a
+  seção acaba ali, e o nome volta ao sentido de fora da seção. Fora dela nada
+  muda.
 - **Contexto genético é descartado** — símbolos de gene colidem com nomes de
   biomarcador (o gene `APOB` vs. a lipoproteína `ApoB`). Linhas com acesso
   RefSeq (`NM_000384.2`), notação HGVS (`p.Trp448*`, `c.1234A>G`), `rs` do dbSNP

package/dist/cli.js

@@ -135,6 +135,34 @@ var QUALITATIVE_VALUE_TERMS = /* @__PURE__ */ new Set([
   "undetectable",
   "yellow"
 ]);
+var GENETIC_CONTEXT_PATTERNS = [
+  /\b[nx][mrpc]_\d{6,}/,
+  // RefSeq: NM_000384.2, NP_, NR_, XM_
+  /\bens[gtp]\d{6,}/,
+  // Ensembl: ENSG00000084674
+  /\bp\.[a-z]{3}\d/,
+  // HGVS proteína: p.Trp448*
+  /\bc\.\d+[acgt]?[>_+-]/,
+  // HGVS codificante: c.1234A>G, c.76_78del
+  /\brs\d{4,}\b/,
+  // dbSNP
+  /\bgenes?\b/,
+  /\bvariante?s?\b/,
+  /\bexons?\b/,
+  /\bzygosity\b/,
+  /\bzigosidade\b/,
+  /\balleles?\b/,
+  /\balelos?\b/,
+  /\bmutations?\b/,
+  /\bmutac(ao|oes)\b/,
+  /\bpathogenic/,
+  /\bpatogenic/,
+  /\bheterozyg/,
+  /\bhomozyg/,
+  /\bheterozigot/,
+  /\bhomozigot/,
+  /\bsequence change\b/
+];
 
 // src/body-region.ts
 var WHOLE_BODY_COMPOSITION_CODES = /* @__PURE__ */ new Set([
@@ -251,6 +279,144 @@ function attachMethodVariants(matches, normalizedText, anchoredLineStarts, match
   }
 }
 
+// src/urinalysis-section.ts
+var HEADER_NAMES = String.raw`(?:urinalysis|urinalise|urina tipo i|rotina de urina|eas)(?![\p{L}\p{N}])`;
+var URINALYSIS_HEADER = new RegExp(`^${HEADER_NAMES}`, "u");
+var ANY_HEADER = new RegExp(String.raw`(?:^|\n)[^\S\n]*` + HEADER_NAMES, "u");
+function mentionsUrinalysisHeader(normalizedText) {
+  return ANY_HEADER.test(normalizedText);
+}
+var URINALYSIS_SECTION_NAMES = /* @__PURE__ */ new Map([
+  ["bacteria", "Bacteria_Urine"],
+  ["bacterias", "Bacteria_Urine"],
+  ["bilirrubina", "Bilirubin_Urine"],
+  ["bilirubin", "Bilirubin_Urine"],
+  ["cetonas", "Ketones_Urine"],
+  ["color", "Color_Urine"],
+  ["cor", "Color_Urine"],
+  ["glicose", "Glucose_Urine"],
+  ["glucose", "Glucose_Urine"],
+  ["hemacias", "RBC_Urine"],
+  ["ketones", "Ketones_Urine"],
+  ["leucocitos", "Leukocytes_Urine"],
+  ["nitrite", "Nitrite_Urine"],
+  ["nitrito", "Nitrite_Urine"],
+  ["ph", "pH_Urine"],
+  ["protein", "Protein_Urine"],
+  ["proteina", "Protein_Urine"],
+  ["rbc", "RBC_Urine"],
+  ["wbc", "Leukocytes_Urine"]
+]);
+function isSectionName(matched) {
+  return URINALYSIS_SECTION_NAMES.has(matched) || URINALYSIS_SECTION_NAMES.has(matched.replace(/s$/, ""));
+}
+var MENTIONS_URINE = /(?<![\p{L}\p{N}])urin[ae](?![\p{L}\p{N}])/u;
+var SEDIMENT_FIELD_UNIT = /\/(?:hpf|lpf)(?![\p{L}\p{N}])/u;
+function hasUrineCue(line) {
+  return MENTIONS_URINE.test(line) || SEDIMENT_FIELD_UNIT.test(line);
+}
+function kindOf(line) {
+  const text = line.text.trim();
+  if (!text) return "blank";
+  if (URINALYSIS_HEADER.test(text) && !/\d/.test(text)) return "header";
+  if (line.foreign) return "foreign";
+  if (line.urine) return "urine";
+  return line.known || line.hasValue ? "neutral" : "unknown";
+}
+function urinalysisLineIndexes(lines) {
+  const kinds = lines.map(kindOf);
+  const continuesAsUrine = (from) => {
+    const next = kinds.findIndex(
+      (kind, index) => index > from && (kind === "urine" || kind === "foreign" || kind === "header")
+    );
+    return next !== -1 && kinds[next] === "urine";
+  };
+  const inside = /* @__PURE__ */ new Set();
+  let open = false;
+  kinds.forEach((kind, index) => {
+    if (kind === "header") {
+      open = true;
+      return;
+    }
+    if (!open || kind === "blank") {
+      return;
+    }
+    if (kind === "foreign" || kind === "unknown" && !continuesAsUrine(index)) {
+      open = false;
+      return;
+    }
+    inside.add(index);
+  });
+  return inside;
+}
+function sectionDepsFrom(catalog, buildPattern, hasValue, resolve) {
+  const loincByCode = new Map(catalog.map((p) => [p.code, p.loinc]));
+  const isUrine = (p) => (Array.isArray(p.category) ? p.category : [p.category]).includes("urina");
+  return {
+    hasValue,
+    patterns: [...URINALYSIS_SECTION_NAMES].map(([name, code]) => {
+      const loinc = loincByCode.get(code);
+      const entry = { ambiguous: false, code, ...loinc && { loinc }, original: name };
+      return { entry, name, regex: buildPattern(name) };
+    }),
+    resolve,
+    urineCodes: new Set(catalog.filter(isUrine).map((p) => p.code))
+  };
+}
+function* matchesIn(text, name, regex) {
+  if (!text.includes(name)) return;
+  regex.lastIndex = 0;
+  for (let match = regex.exec(text); match !== null; match = regex.exec(text)) {
+    yield match;
+  }
+}
+function applyUrinalysisSection(normalizedText, candidates, deps) {
+  const resolved = deps.resolve(candidates);
+  if (!mentionsUrinalysisHeader(normalizedText)) {
+    return resolved;
+  }
+  const lineStarts = [0];
+  for (let i = normalizedText.indexOf("\n"); i !== -1; i = normalizedText.indexOf("\n", i + 1)) {
+    lineStarts.push(i + 1);
+  }
+  const lines = lineStarts.map((start, index) => {
+    const end = index + 1 < lineStarts.length ? lineStarts[index + 1] - 1 : normalizedText.length;
+    const text = normalizedText.slice(start, end);
+    const known = deps.patterns.some(
+      ({ name, regex }) => !matchesIn(text, name, regex).next().done
+    );
+    return { foreign: false, hasValue: deps.hasValue(text), known, text, urine: hasUrineCue(text) };
+  });
+  for (const candidate of resolved) {
+    let index = lineStarts.length - 1;
+    while (lineStarts[index] > candidate.start) index -= 1;
+    const line = lines[index];
+    line.known = true;
+    const matched = normalizedText.slice(candidate.start, candidate.end);
+    if (candidate.entries.some((e) => deps.urineCodes.has(e.code))) {
+      line.urine = true;
+    } else if (!isSectionName(matched)) {
+      line.foreign = true;
+    }
+  }
+  const inside = urinalysisLineIndexes(lines);
+  if (!inside.size) {
+    return resolved;
+  }
+  const sectionCandidates = [];
+  for (const index of inside) {
+    const start = lineStarts[index];
+    for (const { entry, name, regex } of deps.patterns) {
+      for (const match of matchesIn(lines[index].text, name, regex)) {
+        const at = start + match.index;
+        sectionCandidates.push({ end: at + match[0].length, entries: [entry], start: at });
+      }
+    }
+  }
+  const replaced = (c) => sectionCandidates.some((s) => s.start === c.start && s.end === c.end);
+  return deps.resolve([...candidates.filter((c) => !replaced(c)), ...sectionCandidates]);
+}
+
 // src/anchor.ts
 var CONFIDENCE_VALUE_ADJACENT = 1;
 var CONFIDENCE_NAME_ONLY = 0.7;
@@ -302,40 +468,13 @@ function foldCommas(text) {
     (comma, space, offset) => startsWithCatalogName(text.slice(offset + 1 + space.length)) ? comma : " "
   );
 }
-var GENETIC_CONTEXT_PATTERNS = [
-  /\b[nx][mrpc]_\d{6,}/,
-  // RefSeq: NM_000384.2, NP_, NR_, XM_
-  /\bens[gtp]\d{6,}/,
-  // Ensembl: ENSG00000084674
-  /\bp\.[a-z]{3}\d/,
-  // HGVS proteína: p.Trp448*
-  /\bc\.\d+[acgt]?[>_+-]/,
-  // HGVS codificante: c.1234A>G, c.76_78del
-  /\brs\d{4,}\b/,
-  // dbSNP
-  /\bgenes?\b/,
-  /\bvariante?s?\b/,
-  /\bexons?\b/,
-  /\bzygosity\b/,
-  /\bzigosidade\b/,
-  /\balleles?\b/,
-  /\balelos?\b/,
-  /\bmutations?\b/,
-  /\bmutac(ao|oes)\b/,
-  /\bpathogenic/,
-  /\bpatogenic/,
-  /\bheterozyg/,
-  /\bhomozyg/,
-  /\bheterozigot/,
-  /\bhomozigot/,
-  /\bsequence change\b/
-];
 var DIGIT_PATTERN = /\d/;
 var cachedUnitTokens = null;
 function getUnitTokens() {
   if (!cachedUnitTokens) {
+    const isExamName = (t) => URINALYSIS_SECTION_NAMES.has(t) || getNamePatterns().has(t);
     cachedUnitTokens = new Set(
-      Object.keys(UNIT_TO_UCUM).map((unit) => normalize(unit).trim()).filter(Boolean)
+      Object.keys(UNIT_TO_UCUM).map((unit) => normalize(unit).trim()).filter((unit) => unit && !isExamName(unit))
     );
   }
   return cachedUnitTokens;
@@ -497,6 +636,13 @@ function resolveOverlaps(candidates) {
   }
   return accepted;
 }
+var cachedSectionDeps = null;
+var getSectionDeps = () => cachedSectionDeps ??= sectionDepsFrom(
+  getPatterns(),
+  buildNamePattern,
+  hasValueEvidence,
+  resolveOverlaps
+);
 function findBiomarkersInText(ocrText) {
   const startTime = Date.now();
   const normalizedText = normalize(ocrText);
@@ -507,7 +653,7 @@ function findBiomarkersInText(ocrText) {
     ...collectCandidates(normalizedText),
     ...collectWrappedCandidates(normalizedText)
   ];
-  for (const candidate of resolveOverlaps(candidates)) {
+  for (const candidate of applyUrinalysisSection(normalizedText, candidates, getSectionDeps())) {
     const lineStart = getLineBounds(normalizedText, candidate.start).start;
     const lineEnd = getLineBounds(normalizedText, candidate.end - 1).end;
     const key = `${lineStart}:${lineEnd}`;
@@ -1093,7 +1239,7 @@ async function main() {
     strict: false
   });
   if (values.version) {
-    process.stdout.write(`${"0.37.2"}
+    process.stdout.write(`${"0.38.1"}
 `);
     return;
   }

package/dist/index.cjs

@@ -119,6 +119,34 @@ var QUALITATIVE_VALUE_TERMS = /* @__PURE__ */ new Set([
   "undetectable",
   "yellow"
 ]);
+var GENETIC_CONTEXT_PATTERNS = [
+  /\b[nx][mrpc]_\d{6,}/,
+  // RefSeq: NM_000384.2, NP_, NR_, XM_
+  /\bens[gtp]\d{6,}/,
+  // Ensembl: ENSG00000084674
+  /\bp\.[a-z]{3}\d/,
+  // HGVS proteína: p.Trp448*
+  /\bc\.\d+[acgt]?[>_+-]/,
+  // HGVS codificante: c.1234A>G, c.76_78del
+  /\brs\d{4,}\b/,
+  // dbSNP
+  /\bgenes?\b/,
+  /\bvariante?s?\b/,
+  /\bexons?\b/,
+  /\bzygosity\b/,
+  /\bzigosidade\b/,
+  /\balleles?\b/,
+  /\balelos?\b/,
+  /\bmutations?\b/,
+  /\bmutac(ao|oes)\b/,
+  /\bpathogenic/,
+  /\bpatogenic/,
+  /\bheterozyg/,
+  /\bhomozyg/,
+  /\bheterozigot/,
+  /\bhomozigot/,
+  /\bsequence change\b/
+];
 
 // src/body-region.ts
 var WHOLE_BODY_COMPOSITION_CODES = /* @__PURE__ */ new Set([
@@ -235,6 +263,144 @@ function attachMethodVariants(matches, normalizedText, anchoredLineStarts, match
   }
 }
 
+// src/urinalysis-section.ts
+var HEADER_NAMES = String.raw`(?:urinalysis|urinalise|urina tipo i|rotina de urina|eas)(?![\p{L}\p{N}])`;
+var URINALYSIS_HEADER = new RegExp(`^${HEADER_NAMES}`, "u");
+var ANY_HEADER = new RegExp(String.raw`(?:^|\n)[^\S\n]*` + HEADER_NAMES, "u");
+function mentionsUrinalysisHeader(normalizedText) {
+  return ANY_HEADER.test(normalizedText);
+}
+var URINALYSIS_SECTION_NAMES = /* @__PURE__ */ new Map([
+  ["bacteria", "Bacteria_Urine"],
+  ["bacterias", "Bacteria_Urine"],
+  ["bilirrubina", "Bilirubin_Urine"],
+  ["bilirubin", "Bilirubin_Urine"],
+  ["cetonas", "Ketones_Urine"],
+  ["color", "Color_Urine"],
+  ["cor", "Color_Urine"],
+  ["glicose", "Glucose_Urine"],
+  ["glucose", "Glucose_Urine"],
+  ["hemacias", "RBC_Urine"],
+  ["ketones", "Ketones_Urine"],
+  ["leucocitos", "Leukocytes_Urine"],
+  ["nitrite", "Nitrite_Urine"],
+  ["nitrito", "Nitrite_Urine"],
+  ["ph", "pH_Urine"],
+  ["protein", "Protein_Urine"],
+  ["proteina", "Protein_Urine"],
+  ["rbc", "RBC_Urine"],
+  ["wbc", "Leukocytes_Urine"]
+]);
+function isSectionName(matched) {
+  return URINALYSIS_SECTION_NAMES.has(matched) || URINALYSIS_SECTION_NAMES.has(matched.replace(/s$/, ""));
+}
+var MENTIONS_URINE = /(?<![\p{L}\p{N}])urin[ae](?![\p{L}\p{N}])/u;
+var SEDIMENT_FIELD_UNIT = /\/(?:hpf|lpf)(?![\p{L}\p{N}])/u;
+function hasUrineCue(line) {
+  return MENTIONS_URINE.test(line) || SEDIMENT_FIELD_UNIT.test(line);
+}
+function kindOf(line) {
+  const text = line.text.trim();
+  if (!text) return "blank";
+  if (URINALYSIS_HEADER.test(text) && !/\d/.test(text)) return "header";
+  if (line.foreign) return "foreign";
+  if (line.urine) return "urine";
+  return line.known || line.hasValue ? "neutral" : "unknown";
+}
+function urinalysisLineIndexes(lines) {
+  const kinds = lines.map(kindOf);
+  const continuesAsUrine = (from) => {
+    const next = kinds.findIndex(
+      (kind, index) => index > from && (kind === "urine" || kind === "foreign" || kind === "header")
+    );
+    return next !== -1 && kinds[next] === "urine";
+  };
+  const inside = /* @__PURE__ */ new Set();
+  let open = false;
+  kinds.forEach((kind, index) => {
+    if (kind === "header") {
+      open = true;
+      return;
+    }
+    if (!open || kind === "blank") {
+      return;
+    }
+    if (kind === "foreign" || kind === "unknown" && !continuesAsUrine(index)) {
+      open = false;
+      return;
+    }
+    inside.add(index);
+  });
+  return inside;
+}
+function sectionDepsFrom(catalog, buildPattern, hasValue, resolve) {
+  const loincByCode = new Map(catalog.map((p) => [p.code, p.loinc]));
+  const isUrine = (p) => (Array.isArray(p.category) ? p.category : [p.category]).includes("urina");
+  return {
+    hasValue,
+    patterns: [...URINALYSIS_SECTION_NAMES].map(([name, code]) => {
+      const loinc = loincByCode.get(code);
+      const entry = { ambiguous: false, code, ...loinc && { loinc }, original: name };
+      return { entry, name, regex: buildPattern(name) };
+    }),
+    resolve,
+    urineCodes: new Set(catalog.filter(isUrine).map((p) => p.code))
+  };
+}
+function* matchesIn(text, name, regex) {
+  if (!text.includes(name)) return;
+  regex.lastIndex = 0;
+  for (let match = regex.exec(text); match !== null; match = regex.exec(text)) {
+    yield match;
+  }
+}
+function applyUrinalysisSection(normalizedText, candidates, deps) {
+  const resolved = deps.resolve(candidates);
+  if (!mentionsUrinalysisHeader(normalizedText)) {
+    return resolved;
+  }
+  const lineStarts = [0];
+  for (let i = normalizedText.indexOf("\n"); i !== -1; i = normalizedText.indexOf("\n", i + 1)) {
+    lineStarts.push(i + 1);
+  }
+  const lines = lineStarts.map((start, index) => {
+    const end = index + 1 < lineStarts.length ? lineStarts[index + 1] - 1 : normalizedText.length;
+    const text = normalizedText.slice(start, end);
+    const known = deps.patterns.some(
+      ({ name, regex }) => !matchesIn(text, name, regex).next().done
+    );
+    return { foreign: false, hasValue: deps.hasValue(text), known, text, urine: hasUrineCue(text) };
+  });
+  for (const candidate of resolved) {
+    let index = lineStarts.length - 1;
+    while (lineStarts[index] > candidate.start) index -= 1;
+    const line = lines[index];
+    line.known = true;
+    const matched = normalizedText.slice(candidate.start, candidate.end);
+    if (candidate.entries.some((e) => deps.urineCodes.has(e.code))) {
+      line.urine = true;
+    } else if (!isSectionName(matched)) {
+      line.foreign = true;
+    }
+  }
+  const inside = urinalysisLineIndexes(lines);
+  if (!inside.size) {
+    return resolved;
+  }
+  const sectionCandidates = [];
+  for (const index of inside) {
+    const start = lineStarts[index];
+    for (const { entry, name, regex } of deps.patterns) {
+      for (const match of matchesIn(lines[index].text, name, regex)) {
+        const at = start + match.index;
+        sectionCandidates.push({ end: at + match[0].length, entries: [entry], start: at });
+      }
+    }
+  }
+  const replaced = (c) => sectionCandidates.some((s) => s.start === c.start && s.end === c.end);
+  return deps.resolve([...candidates.filter((c) => !replaced(c)), ...sectionCandidates]);
+}
+
 // src/anchor.ts
 var CONFIDENCE_VALUE_ADJACENT = 1;
 var CONFIDENCE_NAME_ONLY = 0.7;
@@ -286,40 +452,13 @@ function foldCommas(text) {
     (comma, space, offset) => startsWithCatalogName(text.slice(offset + 1 + space.length)) ? comma : " "
   );
 }
-var GENETIC_CONTEXT_PATTERNS = [
-  /\b[nx][mrpc]_\d{6,}/,
-  // RefSeq: NM_000384.2, NP_, NR_, XM_
-  /\bens[gtp]\d{6,}/,
-  // Ensembl: ENSG00000084674
-  /\bp\.[a-z]{3}\d/,
-  // HGVS proteína: p.Trp448*
-  /\bc\.\d+[acgt]?[>_+-]/,
-  // HGVS codificante: c.1234A>G, c.76_78del
-  /\brs\d{4,}\b/,
-  // dbSNP
-  /\bgenes?\b/,
-  /\bvariante?s?\b/,
-  /\bexons?\b/,
-  /\bzygosity\b/,
-  /\bzigosidade\b/,
-  /\balleles?\b/,
-  /\balelos?\b/,
-  /\bmutations?\b/,
-  /\bmutac(ao|oes)\b/,
-  /\bpathogenic/,
-  /\bpatogenic/,
-  /\bheterozyg/,
-  /\bhomozyg/,
-  /\bheterozigot/,
-  /\bhomozigot/,
-  /\bsequence change\b/
-];
 var DIGIT_PATTERN = /\d/;
 var cachedUnitTokens = null;
 function getUnitTokens() {
   if (!cachedUnitTokens) {
+    const isExamName = (t) => URINALYSIS_SECTION_NAMES.has(t) || getNamePatterns().has(t);
     cachedUnitTokens = new Set(
-      Object.keys(_fhir.UNIT_TO_UCUM).map((unit) => normalize(unit).trim()).filter(Boolean)
+      Object.keys(_fhir.UNIT_TO_UCUM).map((unit) => normalize(unit).trim()).filter((unit) => unit && !isExamName(unit))
     );
   }
   return cachedUnitTokens;
@@ -481,6 +620,13 @@ function resolveOverlaps(candidates) {
   }
   return accepted;
 }
+var cachedSectionDeps = null;
+var getSectionDeps = () => cachedSectionDeps ??= sectionDepsFrom(
+  getPatterns(),
+  buildNamePattern,
+  hasValueEvidence,
+  resolveOverlaps
+);
 function findBiomarkersInText(ocrText) {
   const startTime = Date.now();
   const normalizedText = normalize(ocrText);
@@ -491,7 +637,7 @@ function findBiomarkersInText(ocrText) {
     ...collectCandidates(normalizedText),
     ...collectWrappedCandidates(normalizedText)
   ];
-  for (const candidate of resolveOverlaps(candidates)) {
+  for (const candidate of applyUrinalysisSection(normalizedText, candidates, getSectionDeps())) {
     const lineStart = getLineBounds(normalizedText, candidate.start).start;
     const lineEnd = getLineBounds(normalizedText, candidate.end - 1).end;
     const key = `${lineStart}:${lineEnd}`;
@@ -881,7 +1027,7 @@ async function extractWithModel(text, options) {
   const tookMs = Date.now() - startedAt;
   try {
     return { payload: JSON.parse(stripFence(raw)), raw, tookMs };
-  } catch (e) {
+  } catch (e2) {
     throw new Error(`O modelo n\xE3o devolveu JSON analis\xE1vel. Resposta crua:
 ${raw.slice(0, 500)}`);
   }