fix: add a new tnm regex pattern by default.

Author: LucasDedieu

edsnlp/pipes/ner/tnm/patterns.py

@@ -1,4 +1,4 @@
-prefix_pattern = r"(?P<prefix>[cpPyraum]p?)"
+prefix_pattern = r"(?P<tumour_prefix>[cpPyraum]p?)"
 tumour_pattern = r"T\s?(?P<tumour>([0-4o]|is))?(?P<tumour_specification>[abcdx]|mi)?"
 tumour_pattern += r"(?:\((?P<tumour_suffix>[^()]{1,10})\))?"
 node_pattern = r"(\s{,2}\/?\s{,2}([cpPyraum]p?)?\s{,2}N\s?(?P<node>[0-3o]|x)"
@@ -9,7 +9,7 @@
 metastasis_pattern = (
     r"(\s{,2}\/?\s{,2}([cpPyraum]p?)?\s{,2}M\s?(?P<metastasis>([01o]|x))x?)"  # noqa: E501
 )
-resection_completeness = r"(\s{,2}\/?\s{,2}R\s?(?P<resection_completeness>[012]))"
+resection_pattern = r"(\s{,2}\/?\s{,2}R\s?(?P<resection>[012]))"
 
 version_pattern = (
     r"\(?(?P<version>uicc|accj|tnm|UICC|ACCJ|TNM)"
@@ -23,6 +23,6 @@
 tnm_pattern += prefix_pattern + r"\s{,2}?" + f"({tumour_pattern})"
 tnm_pattern += r"(\s{,2}" + f"{node_pattern})?"
 tnm_pattern += r"(\s{,2}" + f"{metastasis_pattern})?"
-tnm_pattern += r"(\s{,2}" + f"{resection_completeness})?"
+tnm_pattern += r"(\s{,2}" + f"{resection_pattern})?"
 tnm_pattern += f"({spacer}{version_pattern})?"
 tnm_pattern = r"(?:\b|^)" + tnm_pattern + r"(?:\b|$)"

edsnlp/pipes/ner/tnm/patterns_new.py

@@ -0,0 +1,92 @@
+tumour_pattern = (
+    r"(?P<tumour_prefix>[cpyramP]{1,2}\s?)?"  # Optional tumour prefix
+    r"T\s?"  # 'T' followed by optional space
+    r"(?P<tumour>([0-4]|is|[Xx]))"  # Tumour size (required if 'T' is present)
+    r"(?P<tumour_specification>[abcdx]|mi)?"  # Optional tumour specification
+    r"(?:\s?\((?P<tumour_suffix>[^()]{1,10})\))?"  # Optional tumour suffix
+)
+
+node_pattern = (
+    r"(?P<node_prefix>[cpyraP]{1,2}\s?)?"  # Optional node prefix
+    r"N\s?"  # 'N' followed by optional space
+    r"(?P<node>[Xx01234\+])"  # Node size/status (required if 'N' is present)
+    r"(?P<node_specification>[abcdx]|mi|sn|i[-,+]|mol[-,+]|\(mi\)|\(sn\)|"
+    r"\(i[-,+]\)|\(mol[-,+]\)|\(\d+\s*/\s*\d+\))?"  # Optional node specification
+    r"(?:\s?\((?P<node_suffix>[^()]{1,10})\))?"  # Optional node suffix
+)
+
+metastasis_pattern = (
+    r"(?P<metastasis_prefix>[cpyraP]{1,2}\s?)?"  # Optional metastasis prefix
+    r"M\s?"  # 'M' followed by optional space
+    r"(?P<metastasis>[Xx0123\+])"  # Metastasis status (required if 'M' is present)
+    r"(?P<metastasis_specification>[abcd]|i\+|mol\+|cy\+|\(i\+\)|\(mol\+\)|"
+    r"\(cy\+\)|PUL|OSS|HEP|BRA|LYM|OTH|MAR|PLE|PER|ADR|SKI)?"  # Optional specification
+)
+
+pleura_pattern = (
+    r"PL\s?(?P<pleura>([0123]|x))?"  # Optional pleura status (for lung cancer)
+)
+
+resection_pattern = (
+    r"R\s?"
+    r"(?P<resection>[Xx012])?"  # Optional resection completeness
+    r"(?P<resection_specification>(is|cy\+|\(is\)|\(cy\+\))?)?"  # Optional spec
+    r"(?P<resection_loc>(\((?P<r_loc>[a-z]+)\)[,;\s]*)*)?"  # Optional localization
+)
+
+version_pattern = (
+    r"\(?(?P<version>uicc|accj|tnm|UICC|ACCJ|TNM)"  # TNM version
+    r"\s+([éeE]ditions|[éeE]d\.?)?\s*"
+    r"(?P<version_year>\d{4}|\d{2})\)?"  # Year of the version
+)
+
+TNM_space = r"(\s*[,\/]?\s*|\n)"  # Allow space, comma, or slash as delimiters
+
+# We need te exclude pattern like 'T1', 'T2' if they are not followed by node or
+# metastasis sections.
+exclude_pattern = (
+    r"(?!T\s*[0-4]\s*[.,\/](?!\s*"
+    + node_pattern
+    + "?"
+    + TNM_space
+    + "?"
+    + metastasis_pattern
+    + "?"
+    + "))"
+)
+
+tnm_pattern_new = (
+    r"(?:\b|^)"
+    + exclude_pattern
+    + r"(?:"
+    + r"(?P<T_component>"
+    + tumour_pattern
+    + ")"
+    + TNM_space
+    + "?"
+    + r"(?P<N_component>"
+    + node_pattern
+    + ")?"
+    + TNM_space
+    + "?"
+    + r"(?P<M_component>"
+    + metastasis_pattern
+    + ")?"
+    + TNM_space
+    + "?"
+    + r"(?P<PL_component>"
+    + pleura_pattern
+    + ")?"
+    + TNM_space
+    + "?"
+    + r"(?P<R_component>"
+    + resection_pattern
+    + ")?"
+    + TNM_space
+    + "?"
+    + r"(?P<V_component>"
+    + version_pattern
+    + ")?"
+    + r")"
+    + r"(?:\b|$|\n)"
+)

edsnlp/pipes/ner/tnm/tnm.py

@@ -12,7 +12,7 @@
 from edsnlp.utils.typing import cast
 
 from .model import TNM
-from .patterns import tnm_pattern
+from .patterns_new import tnm_pattern_new
 
 
 class TNMMatcher(BaseNERComponent):
@@ -75,7 +75,7 @@ def __init__(
         nlp: Optional[PipelineProtocol],
         name: str = "tnm",
         *,
-        pattern: Optional[Union[List[str], str]] = tnm_pattern,
+        pattern: Optional[Union[List[str], str]] = tnm_pattern_new,
         attr: str = "TEXT",
         label: str = "tnm",
         span_setter: SpanSetterArg = {"ents": True, "tnm": True},