natural language processing × radiology reports

/BUVSBMMBOHVBHFQSPDFTTJOH º 3BEJPMPHZ3FQPSUT

4VNNBSZ • ์ࣹઢϨϙʔτΛ༻͍ͨࣗવݴޠॲཧͷݚڀʹؔ͢ΔαʔϕΠ • ͜ͷ෼໺͸ੲʢ೥୅ʙʣ͔Βݚڀ͕͞Ε͍ͯΔ͕ɺۙ೥ͷػցֶशٕज़Λ༻͍ͨݚڀΛ ೺Ѳ͢ΔͨΊɺࠓճ͸ओʹ೥Ҏ߱ͷ࿦จΛத৺ʹݕࡧͨ͠ • ۙ೥ͷϨϏϡʔʢ$BTFZ FUBMʣΛࢀߟͱ͠ɺࣗ਎Ͱ೿ੜͤͯ͞࿦จΛݕࡧͨ͠ Casey
A, Davidson E, Poon M, Dong H, Duma D, Grivas A, et al. A systematic review of natural language processing applied to radiology reports. BMC Med Inf Decis Mak. 2020;21:179.

"QQMJDBUJPODBUFHPSZʢ$BTFZ FUBMʣ %JTFBTFJOGPSNBUJPO$MBTTJGJDBUJPO ü ࣬ױͷ༗ແͳͲΛ෼ྨ͢ΔͨΊͷ৘ใΛϨϙʔτ͔Βநग़͢Δ͜ͱΛ໨ࢦ͢ %JBHOPTUJDTVSWFJMMBODF ü ױऀͷ࣬ױঢ়ଶΛϑΥϩʔ͢ΔͨΊɺ࣬ױʹؔ͢Δ৘ใΛαʔϕΠϥϯε͢Δ͜ͱΛ໨ࢦ͢
2VBMJUZBOEDPNQMJBODF ü Ϩϙʔτͷ಺༰͔Β਍ྍͷ࣭΍҆શੑΛධՁ͢Δ͜ͱΛ໨ࢦ͢ $PIPSUBOEFQJEFNJPMPHZ ü Ϩϙʔτ͔ΒྟচݚڀͷͨΊͷίϗʔτΛ࡞੒͢Δ͜ͱΛ໨ࢦ͢ -BOHVBHFEJTDPWFSZLOPXMFEHF ü ϨϙʔτΛղੳ͠ɺ਍அࢧԉ΍ԁ׈ͳίϛϡχέʔγϣϯͷͨΊʹͲ͏࠷దԽ͢Δ͔ʹ͍ͭͯௐ΂Δ 5FDIOJDBM/-1 ü Ϩϙʔτʹ͓͚Δ൱ఆදݱͷݕग़΍εϖϧޡΓमਖ਼ͷΑ͏ͳࣗવݴޠॲཧͷࠜװతͳٕज़ʹؔ͢Δ΋ͷ Casey A, Davidson E, Poon M, Dong H, Duma D, Grivas A, et al. A systematic review of natural language processing applied to radiology reports. BMC Med Inf Decis Mak. 2020;21:179.

"QQMJDBUJPODBUFHPSZʢ$BTFZ FUBMʣ • ϨϏϡʔʢ$BTFZ FUBMʣͷΧςΰϦʔͰ͸ɺ ʮ%JTFBTFJOGPSNBUJPO$MBTTJGJDBUJPOʯͱ ʮ%JBHOPTUJDTVSWFJMMBODFʯ͕൒਺Ҏ্Λ઎Ί͍ͯͨ • ϨϏϡʔʢ$BTFZ FUBMʣ
ͰҾ༻͞Ε͍ͯΔ࿦ จ͸ͦͷ··ͷΧςΰϦʔΛ࢖༻͠ɺͦΕҎ֎ͷ࿦จ ͸ಠஅͰ෼ྨͨ͠ • ͨͩ͠ɺҙຯ߹͍͕ඃ͍ͬͯͨΓɺᐆດͳͱ͜Ζ΋͋ ΔͷͰɺ͋͘·Ͱࢀߟఔ౓ͷ෼ྨ • 5FDIOJDBM/-1͸ϨϏϡʔͰ͸ৄ͘͠ड़΂ΒΕ͍ͯͳ ͔ͬͨͷͰɺࠓճ͸ׂѪ

%FFQ-FTJPO 8BOHFUBM • $5ɾ.3ͳͲ༷ʑͳϞμϦςΟ͔Β੒ΔΞϊςʔγϣϯ෇͖ͷҩྍ ը૾ͷσʔληοτʢ/*)ͷ1"$4ʹ஝ੵ͞Ε͍ͯΔ΋ͷΛར༻ʣ • ը૾ʹ͸Ξϊςʔγϣϯπʔϧʢ3&$*45 ͳͲʣʹΑͬͯɺը૾಺ ͷපมʹΞϊςʔγϣϯ৘ใʢ#PPLNBSLʣ͕෇༩͞Ε͍ͯΔ͜ͱ ͕ଟ͍ʢӈਤྫʣ
• Ξϊςʔγϣϯ͕෇༩͞Ε͍ͯΔը૾ΛूΊͯɺ εϥΠε ʢ CPPLNBSLTʣͷσʔληοτΛߏங • #PPLNBSL৘ใΛCPVOEJOHCPYʹม׵ͯ͠෺ମݕग़ͷΞϊςʔ γϣϯ෇͖σʔληοτͱͯ͠ެ։ Yan K, Wang X, Lu L, Summers RM. DeepLesion: automated mining of large-scale lesion annotations and universal lesion detection with deep learning. J Med Imag. 2018;5(3):36501. RECIST .. response evaluation criteria in solid tumors

1FOHFUBM • Ϩϙʔτ͔Βපมͷಛ௃ʢҐஔɾαΠζɾܗ࣭ͳͲʣʹ ؔ͢Δ৘ใΛநग़ͨ͠ݚڀ • %FFQ-FTJPOͷCPPLNBSLͷ͋Δը૾ʹඥͮ͘Ϩϙʔτ Λར༻ʢӈਤͷΑ͏ʹ͕ϨϙʔτʹCPPLNBSL͕ຒΊࠐ ·Ε͍ͯΔʣ • /&3πʔϧʢ(&/*"5BHHFSʣΛ࢖ͬͯɺΤϯςΟςΟ
நग़Λߦ͍ɺ֤ΤϯςΟςΟ͕໨తͷCPPLNBSLʹඥͮ ͔͘ʹ͍ͭͯ$//ʴ4FMGBUUFOUJPOͰ෼ྨͨ͠ • ࠷ߴਫ਼౓ͱͯ͠'TDPSFΛه࿥ͨ͠ ը૾தͷ྘ͷۣܗ͕Ϩϙʔτ಺ͷ#00,."3,ʹඥͮ͘පม Peng Y, Yan K, Sandfort V, Summers RM, Lu Z. A self-attention based deep learning method for lesion attribute detection from CT reports. arXiv. 2019;1904.13018.

%BWJEFUBM • ಄෦.3*ϨϙʔτΛࣗಈ෼ྨ͢ΔϞσϧΛߏங • ෼ྨλεΫͱͯ͠ҎԼͷ̎ͭͷίʔύεΛ४උ • $PBSTFOPSNBMWTBCOPSNBM • (SBOVMBSEBNBHF WBTDVMBS
NBTT BDVUFTUSPLF 'B[FLBT • #JP#&35ʢ-FFFUBMʣ "UUFOUJPOͷ෼ྨثΛ࡞ͬͯɺ̎ͭͷ෼ྨ໰୊Λղ͍ͨ • $PBSTF (SBOVMBS$MBTTJGJDBUJPOͷͦΕͧΕͰBDDVSBDZɾΛه࿥ • (SBOVMBS$MBTTJGJDBUJPOͰ͸FYQFSJFODFEOFVSPSBEJPMPHJTUʹۇ͔ʹྼΔ΋ͷͷɺFYQFSJFODFE OFVSPMPHJTU΍ TUSPLFQIZTJDJBOΛ্ճΔ෼ྨੑೳΛه࿥ͨ͠ͱ͍ͯ͠Δ David A. Wood, Jeremy Lynch, Sina Kafiabadi, et al. 2020. Automated labelling using an attention model for radiology reports of MRI scans (ALARM). volume 121 of Proceedings of Machine Learning Research, pages 811– 826, Montreal, QC, Canada. PMLR. Lee J, Yoon W, Kim S, Kim D, Kim S, So CH, et al. BioBERT: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics. 2020;36(4):1234–40.

#SFTTFN FUBM • ڳ෦୯७9ઢϨϙʔτΛϚϧνϥϕϧ෼ྨͷλεΫΛߦ͍ɺϨϙʔτͷ ࣗಈ෼ྨΛࢼΈͨݚڀ • ࣗࢪઃͷϨϙʔτΛࣄલֶशʹ࢖༻ͨ͠Ϟσϧʢ3"%#&35ʣΛߏங ͠ɺ (FOFSBM#&35ͱͷਫ਼౓ൺֱΛߦͬͨʢ3"%#&35Ͱ͸ɺࣄલֶ श࣌ʹಠࣗͷࣙॻ΋༻ҙͯ͠ɺҰൠυϝΠϯͰ͸ΧόʔͰ͖ͳ͍ઐ໳
༻ޠͷαϒϫʔυ෼ׂʹ΋஫ҙΛ෷͍ͬͯΔʣ • ڳ෦୯७9ઢϨϙʔτͰධՁͨ݁͠Ռɺ 3"%#&35͸"6$ͰΛୡ ੒ʢͨͩ͠ɺ(FOFSBM#&35Ͱ΋ಉ༷ͷείΞʣ • ଞϞμϦςΟʢ$5ʣͷϨϙʔτͰධՁͨ͠ͱ͜Ζɺ 3"%#&35͸ "6$ͰΛه࿥͠ɺ༏ҐੑΛࣔͨ͠ʢ(FOFSBM#&35͸ʣ Bressem KK, Adams LC, Gaudin RA, Tröltzsch D, Hamm B, Makowski MR, et al. Highly accurate classification of chest radiographic reports using a deep learning natural language model pretrained on 3.8 million text reports. Bioinformatics. 2021;36(21):5255–61. • 'JOEJOH • $POHFTUJPO • 0QBDJUZ • &GGVTJPO • 1OFVNPUIPSBY • 5IPSBDJDESBJO • 7FOPVTDBUIFUFS • (BTUSJDUVCF • 5SBDIFBMUVCFDBOVMB • .JTQMBDFENFEJDBMEFWJDF

*SWJOFUBM$IF9QFSU Irvin J, Rajpurkar P, Ko M, Yu Y,
Ciurea-Ilcus S, Chute C, et al. CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison. arXiv. 2019;1901.0703. • ڳ෦୯७9ઢϨϙʔτʢ ຕʣʹ̍̐ͷॴݟϥϕ ϧΛ෇༩ͨ͠େن໛σʔληοτΛߏங • ֤ॴݟʹϥϕϧʢ1PTJUJWF6ODFSUBJO/FHBUJWFʣ ͕෇༩ʢʮॴݟͳ͠ʯ͸ɺ1PTJUJWFPS/FHBUJWFʣ͞Ε ͍ͯΔ • ϧʔϧϕʔεͰϑϨʔζநग़ɾ൱ఆ൑ఆͳͲΛߦ͍ɺ֤ ॴݟʹϥϕϧΛ෇༩ͨ͠΋ͷΛ܇࿅ηοτͱͯ͠༻ҙ • ධՁηοτͱͯ͠ɺෳ਺ͷ์ࣹઢՊҩ͕࡞੒ͨ͠σʔλ ηοτΛ४උ͠ɺσʔληοτͷ඼࣭ΛධՁͨ͠

4NJUFUBM$IF9CFSU • ڳ෦୯७9ઢϨϙʔτ͔Βॴݟͷ༗ແΛ෼ྨ͢ΔϞσϧͷఏҊ • .*.*$$93ͱ$IF9QFSU *SWJOFUBM ͷσʔληοτ ʢ֤Ϩϙʔτʹʮॴݟͳ͠ʯΛؚΉछྨͷॴݟϥϕϧ͕Ϛϧ νϥϕϧͱͯ͠෇༩͞Ε͍ͯΔʣΛར༻ɻ •
Ϟσϧ͸WBOJMMB#&35ʢϚϧνϥϕϧͷ෼ྨثΛग़ྗ૚ͱ͠ɺ ֤ϥϕϧΛʮ1PTJUJWF /FHBUJWF 6ODFSUBJO #MBOLʯͰ෼ྨʣ Λ༻ҙ Smit A, Jain S, Rajpurkar P, Pareek A, Ng AY, Lungren MP. CheXbert: Combining Automatic Labelers and Expert Annotations for Accurate Radiology Report Labeling Using BERT. arXiv. 2020;2004.09167. Irvin J, Rajpurkar P, Ko M, Yu Y, Ciurea-Ilcus S, Chute C, et al. CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison. arXiv. 2019;1901.0703.

4NJUFUBM$IF9CFSU • ࣄલֶशϞσϧͱͯ͠͸ɺʮ#JP#&35 -FFFUBM ɾ$MJOJDBM#&35 "MUPTBBS FUBM ɾ#MVF#&35 1FOHFUBM
ʯͳͲΛൺֱ • ·ͨɺ຋༁λεΫͰ༻͍ΒΕΔ#BDLUSBOTMBUJPOΛ༻͍ͯ%BUBBVHNFOUBUJPOͨ݁͠Ռͱൺֱ • ݁Ռ͸ʮ#MVF#&35 #BDLUSBOTMBUJPO"VHNFOUBUJPOʯ͕࠷ߴਫ਼౓ʢ'TDPSFʣ • ॴݟͷ༗ແΛͷॴݟϥϕϧʹݶఆ͓ͯ͠ΓɺͦͷଞͷॴݟʹదԠ͕ग़དྷͳ͍఺͸՝୊ͱ͍ͯ͠Δ Lee J, Yoon W, Kim S, Kim D, Kim S, So CH, et al. BioBERT: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics. 2020;36(4):1234–40. Huang K, Altosaar J. ClinicalBert: Modeling Clinical Notes and Predicting Hospital Readmission. arXiv. 2019;1904.05342. Peng Y, Yan S, Lu Z. Transfer Learning in Biomedical Natural Language Processing: An Evaluation of BERT and ELMo on Ten Benchmarking Datasets. In: Proceedings ofthe BioNLP 2019 workshop. Association for Computational Linguistics; 2019. p. 58–65. Smit A, Jain S, Rajpurkar P, Pareek A, Ng AY, Lungren MP. CheXbert: Combining Automatic Labelers and Expert Annotations for Accurate Radiology Report Labeling Using BERT. arXiv. 2020;2004.09167.

+BJOFUBM3BE(SBQI • ڳ෦୯७9ઢϨϙʔτʢ.*.*$$93 $IF9QFSUʣ ʹΤϯςΟςΟͱϦϨʔγϣϯͷϥϕϧΛ෇༩ͨ͠ σʔληοτΛެ։ • ʮ"OBUPNZɾ0CTFSWBUJPOʯΤϯςΟςΟΛ༻ҙ • 0CTFSWBUJPO͸ʮ%FGJOJUFMZ1SFTFOUɾ6ODFSUBJOɾ
%FGJOJUFMZ"CTFOUʯͷ"TTFSUJPO৘ใΛ෇༩ • ϦϨʔγϣϯ͸ʮ4VHHFTUJWF0Gɾ-PDBUFE"Uɾ .PEJGZʯͱ͠ɺ݁ՌΛ༗޲άϥϑͰදݱʢӈදʣ Jain S, Agrawal A, Saporta A, Truong SQ, Duong DN, Bui T, et al. RadGraph: Extracting Clinical Entities and Relations from Radiology Reports. arXiv. 2021;2016.14463.

+BJOFUBM3BE(SBQI • ΤϯςΟςΟநग़͸WBOJMMB#&35ʢUPLFODMBTTJGJDBUJPOʣ • ϦϨʔγϣϯநग़͸3#&35 8VFUBM • ֤λεΫͷ࠷ߴείΞʢ'TDPSFʣ͸ҎԼͷ௨Γ • /&3ʢ.*.*$$93$IF9QFSUʣ
• 3&ʢ.*.*$$93$IF9QFSUʣ • σʔληοτʹ͓͚ΔΞϊςʔλʔؒͷෆҰகʢΤϯςΟ ςΟͷཻ౓ͳͲʣͷ՝୊ʹ͍ͭͯߟ࡯͍ͯ͠Δ Jain S, Agrawal A, Saporta A, Truong SQ, Duong DN, Bui T, et al. RadGraph: Extracting Clinical Entities and Relations from Radiology Reports. arXiv. 2021;2016.14463. Wu S, He Y. Enriching Pre-trained Language Model with Entity Information for Relation Classification. arXiv. 2019;1905.08284. R-BERT

+BJOFUBM7JTVBM$IF9CFSU • ҩ༻ը૾ೝࣝͷϥϕϧσʔλΛਓखͰ४උ͢Δͷ͸ίετͱ͕͔͔࣌ؒΔͨΊɺ୅ΘΓʹը૾ʹඥͮ͘ Ϩϙʔτʹهड़͞Εͨॴݟ৘ใΛநग़ͯ͠୅༻͢Δͱ͍͏ํ๏͕༗༻ͱ͞Ε͍ͯΔ • ҰํͰͦͷΑ͏ʹͯ͠࡞੒͞Εͨσʔληοτ͸ը૾৘ใΛਖ਼͘͠൓ө͓ͯ͠Βͣɺͦͷ඼࣭ͷ௿͕͞ ࢦఠ͞Ε͍ͯΔʢ-VLFʣ • ຊݚڀͰ͸ɺϨϙʔτ͔Βநग़ͨ͠ॴݟ৘ใͷ໰୊఺ͳͲΛ໌Β͔ʹ͠ɺ௚઀ϨϙʔτશମΛೖྗͱ͢ ΔϞσϧʢ7JTVBM$IF9CFSUʣΛ࡞Γɺͦͷ༗༻ੑΛࣔͨ͠
Jain S, Smit A, Qh S, Vinbrain T, Chanh V, Nguyen DT, et al. VisualCheXbert: Addressing the Discrepancy Between Radiology Report Labels and Image Labels; VisualCheXbert: Addressing the Discrepancy Between Radiology Report Labels and Image Labels. In: Proceedings of the Conference on Health, Inference, and Learning. 2021. p. 105–115. Luke Oakden-Rayner. 2019. Exploring large scale public medical image datasets. arXiv:1907.12720

+BJOFUBM7JTVBM$IF9CFSU • .*.*$$93ͱ$IF9QFSU *SWJOFUBM ͷσʔληοτΛར༻ • ֤Ϩϙʔτʹ͸छྨͷॴݟϥϕϧ͕Ϛϧνϥϕϧͱͯ͠෇༩͞Ε͍ͯΔ • ਓͷ์ࣹઢՊҩ͕ը૾ΛಡӨ͠ɺछྨͷॴݟϥϕϧͷ༗ແΛ෇༩͠ɺಉ࣌ʹରԠ͢ΔϨϙʔτʹ΋ ಉ͡ॴݟϥϕϧʢQPTJUJWFOFHBUJWFVODFSUBJOCMBOLʣΛ෇༩ͨ͠
• ը૾͔Β෇༩ͨ͠ϥϕϧΛ(SPVOE5SVUIͱͯ͠ɺϨϙʔτ͔Β෇༩ͨ͠ॴݟϥϕϧͱͷҰக཰Λܭࢉ ͨ͠ͱ͜ΖɺฏۉͰ'είΞ͕ͱ௿͍݁ՌͰ͋ͬͨ Jain S, Smit A, Qh S, Vinbrain T, Chanh V, Nguyen DT, et al. VisualCheXbert: Addressing the Discrepancy Between Radiology Report Labels and Image Labels; VisualCheXbert: Addressing the Discrepancy Between Radiology Report Labels and Image Labels. In: Proceedings of the Conference on Health, Inference, and Learning. 2021. p. 105–115. Irvin J, Rajpurkar P, Ko M, Yu Y, Ciurea-Ilcus S, Chute C, et al. CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison. arXiv. 2019;1901.0703.

+BJOFUBM7JTVBM$IF9CFSU • Τϥʔ෼ੳͷ݁Ռɺॴݟͷʮཻ౓ʯͷҧ͍͕ݪҼͷ̍ͭͰ͋ͬ ͨɻྫ͑͹ʮ-VOH0QBDJUZʯ͸ʮ&EFNBʯͷ্Ґ֓೦ͱఆٛ ͞ΕΔ͕ɺ ϨϙʔτͰʮ&EFNBʯͷॴݟ͕͋ͬͨ৔߹Ͱ΋ ʮ-VOH0QBDJUZʯ͸໌Β͔Ͱ͸ͳ͍ͷͰʮ6ODFSUBJOʯͱϥϕ Ϧϯά͍ͯͨ͠ • ·ͨɺϨϙʔτ͸աڈͷ৘ใΛࢀরͯ͠࡞੒͞ΕΔ͕ɺը૾͸
ͦͷ࣌఺ͷ৘ใ͔͠ࢀরͰ͖ͳ͍఺ɺϨϙʔτ͕ʮ*NQSFTTJPO ηΫγϣϯʯ͔ΒͷΈϥϕϦϯά͢ΔͷͰ৘ใ͕མ͍ͪͯΔՄ ೳੑ͕͋Δ఺ͳͲ͕ݪҼͱͳ͍ͬͯΔՄೳੑ͕ࢦఠ͞Εͨ Jain S, Smit A, Qh S, Vinbrain T, Chanh V, Nguyen DT, et al. VisualCheXbert: Addressing the Discrepancy Between Radiology Report Labels and Image Labels; VisualCheXbert: Addressing the Discrepancy Between Radiology Report Labels and Image Labels. In: Proceedings of the Conference on Health, Inference, and Learning. 2021. p. 105–115. Irvin J, Rajpurkar P, Ko M, Yu Y, Ciurea-Ilcus S, Chute C, et al. CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison. arXiv. 2019;1901.0703.

+BJOFUBM7JTVBM$IF9CFSU • ʮϨϙʔτ͔ΒಘͨϥϕϧʯΛʮը૾͔Βಘͨϥϕ ϧʯʹม׵͢ΔͨΊɺ#&35ͷग़ྗΛಛ௃ྔͱͯ͠ɺ ը૾ͷ֤ϥϕϧͷ༗ແΛ൑ఆ͢ΔϩδεςΟοΫճ ؼϞσϧΛॴݟຖʹֶशͨ͠ • ͜ͷ݁Ռɺ'είΞ͕Λୡ੒͠ɺςΩετͷ ॴݟϥϕϧͷʮVODFSUBJOCMBOLʯͳͲΛϧʔϧ Ͱʮ1PTJUJWF/FHBUJWFʯʹஔ׵͢Δख๏ͱൺֱ
ͯ͠༗ҙͳվળ͕֬ೝ͞Εͨ Jain S, Smit A, Qh S, Vinbrain T, Chanh V, Nguyen DT, et al. VisualCheXbert: Addressing the Discrepancy Between Radiology Report Labels and Image Labels; VisualCheXbert: Addressing the Discrepancy Between Radiology Report Labels and Image Labels. In: Proceedings of the Conference on Health, Inference, and Learning. 2021. p. 105–115.

,FIM FUBM • ഏ؞ͷױऀΛूΊͯɺͦͷϨϙʔτ͔Βഏ؞ͷ༗ແɺܦ࣌తมԽ ʢѱԽʗվળʣɺಛఆͷ৔ॴ΁ͷసҠͷ༗ແͳͲΛ෼ྨ͢ΔϞσ ϧΛ։ൃ • Ϟσϧ͸ϨϙʔτΛܥྻσʔλͱͯ͠ѻ͍ɺ$//ʹೖྗͯ͠ɺഏ ؞ͷ༗ແΛ෼ྨ͢Δग़ྗ૚Λ༻ҙ͠ɺ߹Θͤͯܦ࣌తมԽɺಛఆ ͷ৔ॴ΁ͷసҠͷ༗ແͳͲͷ֤Ϋϥεͷ༧ଌ݁ՌΛ෼ྨ͢Δग़ྗ
૚Λ༻ҙͨ͠ • ഏ؞ͷ༗ແʹ͍ͭͯ͸"6$͕ɺѱԽͷ෼ྨ͸ɺվળͷ෼ ྨ͸ͷਫ਼౓Λୡ੒ͨ͠ Kehl KL, Elmarakeby H, Nishino M, Van Allen EM, Lepisto EM, Hassett MJ, et al. Assessment of Deep Natural Language Processing in Ascertaining Oncologic Outcomes From Radiology Reports. JAMA Oncol. 2019;5(10):1421–9.

,FIM FUBM • ݁Ռͷଥ౰ੑʢ৴པੑʣͷධՁͱͯ͠ɺҎԼΛ४උ • %'4 %JTFBTFGSFFTVSWJWBM • 1'4 1SPHSFTTJPOGSFFTVSWJWBM
• ܦ࣌తมԽʢѱԽʣΛΠϕϯτͱͯ͠ɺͦΕ͕ظؒ಺ʹൃ ੜ͢Δ͔ɺଧͪ੾ΓʹͳΔ͔Λ෼ੳͨ͠ • ਓखͰऩूͯ͠࡞੒ͨ͠ੜଘۂઢͱϞσϧ͔Βಘͨ݁Ռ͕ ͍ۙ݁ՌʹͳΔ͜ͱΛ֬ೝͨ͠ Kehl KL, Elmarakeby H, Nishino M, Van Allen EM, Lepisto EM, Hassett MJ, et al. Assessment of Deep Natural Language Processing in Ascertaining Oncologic Outcomes From Radiology Reports. JAMA Oncol. 2019;5(10):1421–9.

#P[LVSUFUBM • ͕Μͷਐߦঢ়گ΍࣏ྍޮՌͳͲΛධՁ͢ΔͨΊʹ͸ɺ පมͷେ͖͞͸ॏཁͳࢦඪͷ̍ͭ • NFBTVSFNFOUʢFHYDNʣͱͦΕʹؔ࿈͢ΔΤϯςΟςΟ৘ใΛநग़͢Δݚڀ • ࣗࢪઃͷ໿$5ɾ.3ϨϙʔτʢϞμϦςΟ͸ଟ਺ʣΛ࢖༻࣮ͯ͠ݧΛߦ͍ͬͯΔ • NFBTVSFNFOU͸ਖ਼نදݱͰநग़ɺͦͷଞͷΤϯςΟςΟ৘ใ͸$3'Ͱநग़
Bozkurt S, Alkim E, Banerjee I, Rubin DL. Automated Detection of Measurements and Their Descriptors in Radiology Reports Using a Hybrid Natural Language Processing Algorithm. J Digit Imaging. 2019;32:544–53.

#P[LVSUFUBM • நग़ͨ͠NFBTVSFNFOUͱͦͷಛ௃৘ใ͸ϧʔϧ ϕʔεͰඥ෇͚͍ͯΔ • ඥ෇͚ޙɺӈਤͷΑ͏ͳNFBTVSFNFOUΛओͱͨ͠ ϑϨʔϜΛߏங͢Δ • NFBTVSFNFOUͷநग़ਫ਼౓͸͕ͩɺߏஙͨ͠ ϑϨʔϜࣗମ͕ਖ਼ղσʔλͱ׬શҰகͨ͠ͷ͸શ
ମͷͩͬͨ Bozkurt S, Alkim E, Banerjee I, Rubin DL. Automated Detection of Measurements and Their Descriptors in Radiology Reports Using a Hybrid Natural Language Processing Algorithm. J Digit Imaging. 2019;32:544–53.

:FUJTHFO FUBM • ը૾਍அͰ͸ɺױऀʹॏେͳ݈߁ϦεΫΛٴ΅͢༧ظͤ͵ۮൃతॴݟΛൃݟ͞ΕΔ͜ͱ͕͋Δ • Ϩϙʔτʹۮൃతॴݟ΍ͦΕʹର͢ΔϑΥϩʔΞοϓͷਪ঑ͷهड़͕͋ͬͨͱͯ͠΋ɺͦΕΒ͕ݟಀ͞ ΕΔ͜ͱ͕͋Δɻ·ͨɺͦΕΒ͸ґཔҩͷؔ৺֎ͷॴݟͷͨΊɺͦͷޙͷέΞ͕ߦΘΕͳ͍͜ͱ͕͋Δ • ݚڀͰ͸ɺϨϙʔτ͔ΒྟচతʹॏཁͳϑΥϩʔΞοϓΛಛఆ͢Δ͜ͱΛ໨ࢦ͢ •
۩ମతʹ͸ϨϙʔτΛจ୯Ґʹ෼ׂ͠ɺ͋Δจ͕ʮϑΥϩʔΞοϓΛਪ঑͢Δจʯ͔Ͳ͏͔Λ൑ఆ͢Δ ೋ஋෼ྨثΛߏங͠ɺ໨తͷϨϙʔτΛݕग़͢Δ͜ͱΛ໨ࢦͨ͠ • ݚڀʹ͸ࣗࢪઃͷ୯७9ઢɾ.3*ɾ$5ɾ௒Ի೾ͷϨϙʔτʢຕʣΛར༻͠ɺ ໊̎ͷҩࢣ͕ʮϑΥ ϩʔΞοϓΛਪ঑͢ΔจʯʹΞϊςʔγϣϯΛ࣮ࢪͨ͠ʢ,BQQB4DPSFʣ Yetisgen-Yildiz M, Gunn ML, Xia F, Payne TH. A text processing pipeline to extract recommendations from radiology reports. J Biomed Inform. 2013;46(2):354–62.

:FUJTHFO FUBM • ෼ྨϞσϧͱͯ͠͸ʮର਺ઢܗճؼʢ࠷େΤϯτϩϐʔ๏ʣʯΛ࢖༻ • 104 ʢ1BSUPG4QFFDIʣɾOHSBNTʢ ʣʹՃ͑ͯɺ.FUB.BQΛࢀরͯ͠ૉੑΛ࡞੒ •
࠷ߴείΞͰ1SFDJTJPOɾ3FDBMMʢ'4DPSFʣ • ِཅੑͷසग़ύλʔϯͱͯ͠͸ʮྟচతͳݒ೦͕͋Δ৔߹͸.3*Λਪ঑͢ΔʯͷΑ͏ͳϔοδจΛϙδ ςΟϒྫͱͯ͠෼ྨ͍ͯͨ͠ʢΞϊςʔγϣϯͰ͸͜ͷΑ͏ͳϔοδจ͸ෛྫͱ͍ͯͨ͠ʣ • Ϩϙʔτͷશମͷ͏ͪʮϑΥϩʔΞοϓΛਪ঑͢Δจʯ͸͘͝Ұ෦ͳͷͰɺσʔλͷ෼෍͕ඇৗʹෆۉ ߧʢਖ਼ྫ͕ྫʀʣͩͬͨ͜ͱ΋͋Γɺଟ͘ͷِӄੑ͕ൃੜ͍ͯͨ͠ Yetisgen-Yildiz M, Gunn ML, Xia F, Payne TH. A text processing pipeline to extract recommendations from radiology reports. J Biomed Inform. 2013;46(2):354–62.

Lou et al. 2020) • Ϩϙʔτ͔ΒϑΥϩʔΞοϓͷཁ൱Λ෼ྨ͢ΔݚڀʢݚڀͷϞνϕʔγϣϯ͸:FUJTHFOΒͱಉ༷ʣ • ࢖༻σʔληοτʹ͸ɺઌߦݚڀ ;BGBS FUBM
ʹͯଁث಺ͷGPDBMNBTTʹ͍ͭͯɺѱੑ౓Λࣔ ͢ϥϕϧʢ࣍ϖʔδʣ͕طʹ෇༩͞Ε͓ͯΓɺຊݚڀͰ͸ͦͷ৘ใΛΰʔϧυελϯμʔυͷϥϕϧͱ ͯ͠ར༻ͨ͠ • :FUJTHFOΒͷख๏͸ʮ͋Δจ͕ʹϑΥϩʔΞοϓΛࢦࣔ͢͠Ωʔϫʔυʯ͕ग़ݱ͢Δ͔Λݟ͍ͯΔ͚ͩ ͳͷͰɺ௚઀తͳΩʔϫʔυ͕ଘࡏ͠ͳ͍৔߹ʹͦͷϨϙʔτΛݟམͱ͢ϦεΫ͕͋Δͱ͍ͯ͠Δ • ຊख๏Ͱ͸ɺʮϑΥϩʔΞοϓʯͷΑ͏ͳΩʔϫʔυΛर্͍͛ΔͷͰ͸ͳ͘ɺର৅Λѱੑजᙾʹ ϑΥʔΧε͠ɺͦΕΒΛػցֶशͰݕग़͢Δ͜ͱΛ໨ࢦͨ͠ Lou R, Lalevic D, Chambers C, Zafar HM, Cook TS. Automated Detection of Radiology Reports that Require Follow-up Imaging Using Natural Language Processing Feature Engineering and Machine Learning Classification. J Digit Imaging. 2020;33(1):131–6. Zafar HM, Chadalavada SC, Kahn CE, Cook TS, Sloan CE, Lalevic D et al.: Code abdomen: an assessment coding scheme for abdominal imaging findings possibly representing cancer. J Am Coll Radiol JACR. 12(9):947–950, 2015 Yetisgen-Yildiz M, Gunn ML, Xia F, Payne TH. A text processing pipeline to extract recommendations from radiology reports. J Biomed Inform. 2013;46(2):354–62.

Zafar et al. 2015) Zafar HM, Chadalavada SC, Kahn CE,
Cook TS, Sloan CE, Lalevic D et al.: Code abdomen: an assessment coding scheme for abdominal imaging findings possibly representing cancer. J Am Coll Radiol JACR. 12(9):947–950, 2015

Lou et al. 2020) • ϥϕϧͷ͏ͪʮJOEFUFSNJOBUFɾTVTQJDJPVTʯΛʮཁϑΥϩʔΞοϓʯͱͯ͠ݕग़͢Δର৅ͱͨ͠ • ʮཁϑΥϩʔΞοϓʯͷϨϙʔτΛ෼ྨ͢Δೋ஋෼ྨ໰୊ͱͯ͠ѻ͍ɺOHSBNTʢ ʣΛಛ௃ྔͱ
ͯ͠ʮφΠʔϒϕΠζɾܾఆ໦ɾର਺ઢܗճؼʯͳͲͷΫϥγΧϧͳػցֶशϞσϧͰධՁ • '4DPSFͩͱɺܾఆ໦͕࠷΋ྑ͔ͬͨʢʣɻ ཁϑΥϩʔΞοϓͷ-3ʢMJLFMJIPPESBUJPʣ͕ߴ ͔ͬͨOHSBNT͸ʮSFOBMOFPQMBTNɾ FWBMV XJUIFOIBODʯͳͲ • ݚڀͷσβΠϯ্ɺط஌ͷѱੑजᙾ͸ʮཁϑΥϩʔΞοϓʯͱ෼ྨ͠ͳ͍Α͏ʹ͓ͯ͠Γɺѱੑ౓͕ ʮதؒతʯͳϨϙʔτͷΈΛݕग़͢Δͱ͍͏৚͕݅λεΫΛ೉͍ͯͨ͘͠͠ͱ͍ͯ͠Δ Lou R, Lalevic D, Chambers C, Zafar HM, Cook TS. Automated Detection of Radiology Reports that Require Follow-up Imaging Using Natural Language Processing Feature Engineering and Machine Learning Classification. J Digit Imaging. 2020;33(1):131–6.

,BOHFUBM • ഏ͕ΜͷϦεΫίϯτϩʔϧΛߦ͏ͨΊʮJODJEFOUBMMZMVOHOPEVMFʢ*-/ʣʯͷϚωδϝϯτ͕ॏཁ • ͨͩ͠ɺίετͷ؍఺͔ΒશͯͷױऀΛ௥ՃͰݕࠪ͢ΔͷͰ͸ͳ͘ɺϦεΫʹԠͨ͡ద੾ͳϑΥϩʔ ΞοϓΛߦ͏͜ͱ͕๬·ΕΔ • 'MFJTDIOFS 4PDJFUZ͔Β*-/ͷϚωδϝϯτʹؔ͢ΔΨΠυϥΠϯ͕ग़͞Ε͓ͯΓɺ͜ͷϧʔϧʹԊͬ ͨӡ༻͕ظ଴͞ΕΔ
• ͔͠͠ɺϨϙʔτ͸ϑϦʔςΩετͰߏ଄Խ͞Ε͍ͯͳ͍ͷͰɺ͔ͦ͜Β໨ࢹͰ಺༰Λ֬ೝͯ͠ΨΠυ ϥΠϯʹద༻ͤ͞Δͷ͸ඇৗʹ͕͔͔࣌ؒΔ • ຊݚڀͰ͸ɺϑΥϩʔΞοϓ͕ඞཁͳ*-/Λಛఆ͢ΔΞϧΰϦζϜΛ։ൃ͠ɺͦͷޙɺͦͷϨϙʔτ͕ 'MFJTDIOFS 4PDJFUZHVJEFMJOFʹԊͬͨϑΥϩʔΞοϓ͕ߦΘΕ͔ͨΛධՁͨ͠ S. K. Kang, K. Garry, R. Chung, W. H. Moore, E. Iturrate, J. L. Swartz, D. C. Kim, L. I. Horwitz, and S. Blecker, ‘‘Natural language processing for identification of incidental pulmonary nodules in radiology reports,’’ J. Amer. College Radiol., vol. 16, no. 11, pp. 1587–1594, Nov. 2019 MacMahon H, Naidich DP, Goo JM, Lee KS, Leung ANC, Mayo JR, et al. Guidelines for Management of Incidental Pulmonary Nodules Detected on CT Images: From the Fleischner Society 2017. Radiology. 2017;284(1):228–43.

'MFJTDIOFS 4PDJFUZ(VJEFMJOFT Ұ෦ൈਮ

,BOHFUBM • ݚڀͰ͸ڳ෦$5ϨϙʔτΛ࢖༻ ˎط஌ͷݪൃ૥ɾʢ"$3DSJUFSJBʹԊͬͨʣഏ͕Μͷಛ௃Λ༗͢ΔױऀͳͲ͸আ֎ • ظؒ಺ʹෳ਺ճͷ$5͕ߦΘΕ͍ͯΔױऀ͸ॳճͷϨϙʔτͷΈΛղੳ • *-/ͷఆٛͱͯ͠ɺ͜Ε·ͰͷϨϙʔτͰࢦఠ͞Ε͓ͯΒͣɺط஌ͷѱੑजᙾɺഏԌͳͲͷྟচతഎܠ Λ࣋ͨͳ͍݁અΛର৅ͱͨ͠ɻ •
݁અ͕ʮมԽͳ͠ɺ҆ఆ͍ͯ͠ΔʯͳͲͱॻ͔Ε͍ͯΔ৔߹΋JODJEFOUBMͷج४͔Βআ֎ɻಉ༷ʹྑੑ Ͱ͋Δ͜ͱ͕֬ఆతͳ݁અʢੴփԽͷྑੑύλʔϯʣͳͲ΋JODJEFOUBMͷج४͔Βআ֎ɻ • ͜ΕΒͷϧʔϧʹԊͬͯ์ࣹઢՊҩ͕Ϩϙʔτ಺ͷ*-/ʹΞϊςʔγϣϯΛߦͬͨ ʢຕͷϨϙʔτͷ͏ͪɺຕͷϨϙʔτʹ*-/ͷΞϊςʔγϣϯ͕෇༩͞Εͨʣ S. K. Kang, K. Garry, R. Chung, W. H. Moore, E. Iturrate, J. L. Swartz, D. C. Kim, L. I. Horwitz, and S. Blecker, ‘‘Natural language processing for identification of incidental pulmonary nodules in radiology reports,’’ J. Amer. College Radiol., vol. 16, no. 11, pp. 1587–1594, Nov. 2019 Swartz J, Koziatek C, Theobald J, Smith S, Iturrate E. Creation of a simple natural language processing tool to support an imaging utilization quality dashboard. Int J Med Inform. 2017;101:93–9.

,BOHFUBM • ΤϯςΟςΟநग़ثͱͯ͠ɺ4JNQMF/-1ͱݺ͹ΕΔϧʔϧϕʔεͷϞσϧΛར༻ͯ͠ɺ໨తͷϑϨʔ ζʢʴ൱ఆදݱͷݕग़ʣΛߦͬͨ • 4FOTJUJWJUZͷείΞͰ*-/Λݕग़ʢ1PTJUJWF1SFEJDUJWF7BMVF͸ʣ • Ϩϙʔτ಺༰ͷධՁͱͯ͠ɺ *-/͕ݕग़͞ΕͨϨϙʔτΛϨϏϡʔͨ͠ͱ͜Ζɺ໿ͷ*-/ʹؔ࿈͢ ΔϨίϝϯσʔγϣϯͷهࡌؚ͕·Ε͍ͯͨ
• ͦͷ͏ͪɺͷϨίϝϯσʔγϣϯͷ಺༰͕ΨΠυϥΠϯͱҰக͍ͯͨ͠ S. K. Kang, K. Garry, R. Chung, W. H. Moore, E. Iturrate, J. L. Swartz, D. C. Kim, L. I. Horwitz, and S. Blecker, ‘‘Natural language processing for identification of incidental pulmonary nodules in radiology reports,’’ J. Amer. College Radiol., vol. 16, no. 11, pp. 1587–1594, Nov. 2019 Swartz J, Koziatek C, Theobald J, Smith S, Iturrate E. Creation of a simple natural language processing tool to support an imaging utilization quality dashboard. Int J Med Inform. 2017;101:93–9.

9JF FUBM • Ϩϙʔτ͔ΒΤϯςΟςΟͱͦͷؔ܎ͷநग़Λߦͬͨݚڀ • 5VNPSʹϑΥʔΧεͨ͠ΤϯςΟςΟͷεΩʔϚΛ༻ҙʢӈදʣ • ̎ࢪઃ͔ΒूΊͨڳ෦$5Ϩϙʔτʢ໿ ݅ʣΛ࢖༻ •
Ϟσϧ͸(36Λ࢖༻ɻೖྗͷຒΊࠐΈදݱͱͯ͠ɺ୯ޠʹՃ͑ͯ ʮจࣈɾ෦टʯͳͲͷ৘ใΛར༻ • நग़ͨ͠ΤϯςΟςΟͷඥ෇͚ʢάϧʔϓԽʣ͸ΤϯςΟςΟͷ Ґஔ৘ใΛ༻͍ͯϧʔϧϕʔεͰॲཧ • ΤϯςΟςΟநग़ͷੑೳ͸ʢ'4DPSFʣͱߴ͔͕ͬͨɺ άϧʔϓԽͷੑೳ͸ʢ"DDVSBDZʣͱϧʔϧϕʔεͷख๏ʹ վળͷ༨஍͕͋ͬͨͱ͍ͯ͠Δ Xie Z, Yang Y, Wang M, Li M, Huang H, Zheng D, et al. Introducing Information Extraction to Radiology Information Systems to Improve the Efficiency on Reading Reports. Methods Inf Med. 2019;58:94–106.

Liu et al. 2020) • ΤϯςΟςΟநग़ɾਖ਼نԽɾؔ܎நग़ͷύΠϓϥΠ ϯΛߏங͠ɺෲ෦$5Ϩϙʔτͷߏ଄ԽΛࢼΈͨ • ྟচԠ༻ͱͯ͠ɺૣظൃݟ͕๬·ΕΔ؊͕Μͷױऀ ʹয఺Λ౰ͯͯɺ༧ଌϞσϧͷ݁ՌΛධՁ͍ͯ͠Δ
• σʔληοτͱͯ͠ɺ$5ϨϙʔτʹΞϊςʔγϣ ϯΛߦ͍ɺຕͷ؊͕Μʹؔ͢ΔϨϙʔτΛूΊ ͨɻͦͷଞɺ؊ߗมɾ؊೯๔ɾ݂؅जͳͲͷ঱ྫʹ ؔ͢ΔϨϙʔτΛຕूΊͨɻ Liu H, Xu Y, Zhang Z, Wang N, Huang Y, Hu Y, et al. A Natural Language Processing Pipeline of Chinese Free-text Radiology Reports for Liver Cancer Diagnosis. IEEE 2020 Aug 28;8:159110-159119.

Liu et al. 2020) • ΤϯςΟςΟநग़Ϟσϧͱͯ͠-45. $3'Λ࢖༻͠ɺจࣈྻͷ෼ࢄදݱΛಛ௃ྔͱͯ͠ೖྗ • จࣈྻ৘ใʹՃ͑ͯɺࣄલʹ༻ҙͨ͠ޠኮϦετͱͷϚονϯά݁ՌΛಛ௃ྔʹؚΊͨ •
ΤϯςΟςΟͷछྨ͸ʮ-PDBUJPOɾ.PSQIPMPHZɾ%FOTJUZɾ&OIBODFNFOUɾ.PEJGJFSʯͷ̑ͭ • நग़ͨ͠ΤϯςΟςΟͷ༻ޠ͸ɺࣄલʹ༻ҙͨ͠ಉٛޠϦετʹج͖ͮɺਖ਼نԽ͞Εͨ • ΤϯςΟςΟநग़ͷੑೳ͸ɺશମͷ'4DPSFͰΛୡ੒ Liu H, Xu Y, Zhang Z, Wang N, Huang Y, Hu Y, et al. A Natural Language Processing Pipeline of Chinese Free-text Radiology Reports for Liver Cancer Diagnosis. IEEE 2020 Aug 28;8:159110-159119.

Liu et al. 2020) • நग़ͨ͠ΤϯςΟςΟ͔Βɺ-PDBUJPOΛΩʔͱͯ͠ɺϧʔ ϧϕʔεͰΤϯςΟςΟͷ૊Έ߹ΘͤΛ࡞Δʢӈྫʣ • Ϩϙʔτ͔Βநग़͞Εͨ૊Έ߹Θͤͷ༗ແͷόΠφϦ஋Λ ಛ௃ྔͱͯ͠ɺ؊͕Μͷ༗ແΛ༧ଌ͢Δ෼ྨثΛߏங
• ෳ਺ͷ෼ྨثͰ࣮ݧΛߦ͍ɺ3BOEPN'PSFTUͰ࠷ߴੑೳ Λه࿥ʢ'4DPSFʀʣ • Τϥʔ෼ੳͷ݁Ռɺ؊͕Μͱࣅͨಛ௃Λ࣋ͭ؊ߗมʹؔ͢ ΔϨϙʔτِ͕ཅੑͱͯ͠ଟ͘ݕग़͞Εͨͱ͍ͯ͠Δ Liu H, Xu Y, Zhang Z, Wang N, Huang Y, Hu Y, et al. A Natural Language Processing Pipeline of Chinese Free-text Radiology Reports for Liver Cancer Diagnosis. IEEE 2020 Aug 28;8:159110-159119.

8BEJBFUBM • $BODFS$BSF5SBDLJOH4ZTUFN $$54 Ͱ͸ɺಡӨ࣌ʹ์ࣹઢՊҩͱ/-1ΞϧΰϦζϜ͕ɺഏ͕Μͷ ͍͕ٙ͋ΔϨϙʔτʹίʔυΛ෇༩͍ͯ͠Δ • $$54ʹ͓͚ΔਓखͷίʔσΟϯάʢ.BOVBMʣͱ/-1ΞϧΰϦζϜͷ݁ՌΛൺֱ͠ɺੑೳධՁΛߦ͏ • /-1ΞϧΰϦζϜͱͯ͠͸ʮD5",&4ʢ4BWPWB
FUBMʣʯΛ༻͍ͨ෼ྨϞσϧΛ༻ҙ • ίʔσΟϯάࡁΈͷϨϙʔτ͔ΒຕΛαϯϓϦϯά͠ɺੑೳධՁΛߦͬͨ • ධՁηοτͱͯ͠ɺఆٛͨ͠ϧʔϧʹج͖ͮɺෳ਺ͷ์ࣹઢՊҩͰ֤ϨϙʔτʹόΠφϦ஋Λ෇༩ Wadia R, Akgun K, Brandt C, Fenton BT, Levin W, Marple AH, Garla V, Rose MG, Taddei T TC. Comparison of Natural Language Processing and Manual Coding for the Identification of Cross-Sectional Imaging Reports Suspicious for Lung Cancer. JCO Clin Cancer Inf. 2018;2:1–7. Savova GK, Masanz JJ, Ogren PV, et al: Mayo clinical Text Analysis and Knowledge Extraction System (cTAKES): Architecture, component evaluation and applications. J Am Med Inform Assoc 17:507-513, 2010

8BEJBFUBM • .BOVBMͱൺֱͯ͠ɺ/-1͸ײ౓͸ߴ͔͕ͬͨɺ ಛҟ౓͸௿͔ͬͨʢཅੑ༧ଌ஋͸ಉఔ౓ʣ • $PNCJOFEʢͲͪΒ͔͕1PTJUJWFͰ͋Ε͹ 1PTJUJWFʣͷ৔߹ɺײ౓͸Λه࿥ • ґཔප໊ผͰݟΔͱɺײ౓ͷࠩ͸ʮ͕ΜҎ֎ͷ ґཔʯͷϨϙʔτͰΑΓݦஶͩͬͨʢ.BOVBM
ͷ৔߹ɺ ʮ͕ΜҎ֎ͷಡӨґཔʯͩͱײ౓͕ େ͖͘௿Լ͍ͯͨ͠ʣ Wadia R, Akgun K, Brandt C, Fenton BT, Levin W, Marple AH, Garla V, Rose MG, Taddei T TC. Comparison of Natural Language Processing and Manual Coding for the Identification of Cross-Sectional Imaging Reports Suspicious for Lung Cancer. JCO Clin Cancer Inf. 2018;2:1–7.

1BOJDFL FUBM • ์ࣹઢϨϙʔτʹ༻͍ΒΕΔ਍அͷʮෆ࣮֬ੑʢ6ODFSUBJOUZʣ ʯ Λࣔ͢දݱ͸์ࣹઢՊҩʹΑ༷ͬͯʑͰ͋Γɺ·ͨ֬৴౓ͷ౓߹͍ ʹ͍ͭͯ͸ղऍ͕Ұக͠ͳ͍͜ͱ͕ଟ͍ʢ,IPSBTBOJ FUBMʣ • ༻ޠͷղऍͷζϨ΍ίϛϡχέʔγϣϯϛεΛݮΒͨ͢Ίɺӈදͷ
Α͏ͳʮ$FSUBJOUZUFSNTʯͱ਍அͷ໬΋Β͠͞Λࣔ͢ج४Λ࡞ͬͨ • ͜ΕΒͷϧʔϧΛࣗӃͰద༻͠ɺ์ࣹઢՊҩʹ͸ಡӨ࣌ʹ͜ͷج४ Λࢀߟͱ͢ΔΑ͏ʹଅͨ͠ • ͜ͷج४͕࠷దͱ͸ݶΒͳ͍͕ɺґཔҩͱಡӨҩͷؒͰڞ௨ͷϧʔ ϧΛڞ༗͢Δ͜ͱͰίϛϡχέʔγϣϯ͕վળͰ͖Δͱ͍ͯ͠Δ Panicek DM, Hricak H. How sure are you, doctor? A standardized lexicon to describe the radiologist’s level of certainty. AJR Am J Roentgenol 2016;207:2–3 Khorasani R, Bates DW, Teeger S, Rothschild JM, Adams DF, Seltzer SE. Is Terminology Used Effectively to Convey Diagnostic Certainty in Radiology Reports? Acad Radiol. 2003;10:685–8.

$BMMFOFUBM • ࣗࢪઃͷϨϙʔτΛ༻͍ͯʮෆ࣮֬ੑʯΛࣔ͢දݱͷಛ௃ʹ͍ͭͯ ෼ੳͨ͠࿦จ • ͋Δදݱͷʮෆ࣮֬ੑʯͷ౓߹͍͸ґཔҩͱಡӨҩͰͷζϨ΋େ͖ ͘ίϛϡχέʔγϣϯͷᴥᴪʹͭͳ͕Δ͜ͱ΋͋Δʢ3PTFOLSBOU[ FUBMʣ • ෼ੳͨ͠Ϩϙʔτͷ͏ͪɺʹʮෆ࣮֬ੑʯΛࣔ͢༻ޠ͕࢖༻
͞Ε͍ͯͨ • ʮෆ࣮֬ੑʯΛࣔ͢දݱͷස౓ͱܦྺͱͷ૬ؔ͸ݟΒΕͳ͔ͬͨ Callen AL, Dupont SM, Price A, Laguna B, Mccoy D, Do B, et al. Between Always and Never: Evaluating Uncertainty in Radiology Reports Using Natural Language Processing. J Digit Imaging. 2020;33(5):1194–201. Rosenkrantz AB, Kiritsy M, Kim S. How “consistent” is “consistent”? A clinician-based assessment ofthe reliability ofexpressions used by radiologists to communicate diagnostic confidence. Clin Radiol. 2014;69(7):745–9

-JVFUBM • ʮෆ࣮֬ੑʯΛධՁͨ͠ઌߦݚڀ͸͋Δ͕ɺͦͷଟ͘͸ʮΩʔϫʔυϕʔεʯͰ͋ΓɺίϯςΩετͷ ৘ใΛߟྀͨ͠ʮ਍அͷ໬΋Β͠͞ʯΛධՁͰ͖͍ͯͳ͍ • ຊݚڀ͸ϨϙʔτͷॴݟηΫγϣϯͷʮจ୯Ґͷ໬΋Β͠͞ʯΛػցֶशͰ෼ྨ͢Δख๏ΛఏҊ • Ϋϥεͱͯ͠͸ʮ/PO%FGJOJUJWF%FGJOJUJWF4USPOH%FGJOJUJWF.JME0UIFSʯΛ༻ҙ • ಄෦.3*Ϩϙʔτʢຕʀจʣʹ্هϥϕϧΛ෇༩ͯ͠ɺΞϊςʔγϣϯσʔλΛ࡞੒ͨ͠
• #JP#&35 -FFFUBM ΛGJOFUVOJOHͨ݁͠Ռ͕࠷ߴείΞʢ"6$ʣΛه࿥ͨ͠ Liu F, Zhou P, Baccei SJ, Masciocchi MJ, Amornsiripanitch N, Kiefe CI, et al. Qualifying Certainty in Radiology Reports through Deep Learning-Based Natural Language Processing. Am J Neuroradiol. 2021;42(10):1755–61. Panicek DM, Hricak H. How sure are you, doctor? A standardized lexicon to describe the radiologist’s level of certainty. AJR Am J Roentgenol 2016;207:2–3 Lee J, Yoon W, Kim S, Kim D, Kim S, So CH, et al. BioBERT: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics. 2020;36(4):1234–40.

%BUUBFUBM • ςΩετதͷʮۭؒ৘ใʯΛೝࣝ͢Δʮ4QBUJBMSPMFMBCFMJOH ,PSEKBNTIJEJ FUBM ʯ ʢ4Q3-ʣΛԠ༻ͯ͠ɺڳ෦9ઢϨϙʔτͷʮۭؒ৘ใʯΛߏ଄Խͨ͠ݚڀ • 4Q3-͸ɺओʹҎԼͷΑ͏ͳεΩʔϚ·Ͱߏ੒͞ΕΔ
ؔ৺ͷ͋Δର৅෺Λࣔ͢ʮ53"+&$503ʯ ͦͷ৔ॴΛࣔ͢ʮ-"/%."3,ʯͱ ͱͷؔ܎Λࣔ͢લஔࢺͷʮ41"5*"-*/%*$"503ʯ • Ϩϙʔτ಺ͷʮ41"5*"-*/%*$"503ʯΛೝࣝ͠ɺͦΕʹඥͮ͘4QBUJBMSPMFʢ53"+&$503 -"/%."3,)&%(&%*"(/04*4ʣΛநग़͠ɺߏ଄Խ͢Δ͜ͱΛࢼΈͨ P. Kordjamshidi, M.V. Otterlo, M.-F. Moens, Spatial Role Labeling: Task Definition and Annotation Scheme, in: Proceedings of the Language Resources & Evaluation Conference, 2010, pp. 413–420. Datta S, Si Y, Rodriguez L, Shooshan SE, Demner-Fushman D, Roberts K. Understanding spatial language in radiology: Representation framework, annotation, and spatial relation extraction from chest X-ray reports using deep learning. J Biomed Inform. 2020;32.

%BUUBFUBM • ֶश͸41"5*"-*/%*$"503ͱ4QBUJBMSPMFΛ෼͚ͯೋஈ֊ͷ/&3λεΫͱͯ͠ߦͬͨ • ·ͣɺ4QBUJBMSPMFؒͷڮ౉͠Λߦ͏41"5*"-*/%*$"503Λೝࣝ͢Δ • ࣍ʹɺೝࣝͨ͠41"5*"-*/%*$"503ʹඥͮ͘4QBUJBMSPMFΛநग़͢ΔλεΫΛֶश • 9-/FUΛ࢖ͬͨϞσϧ͕࠷ߴੑೳΛه࿥͠ɺ'είΞͰͦΕͧΕɺΛୡ੒ͨ͠ •
ݸผͷϥϕϧͩͱʮ)&%(&%*"(/04*4ʯ͸ֶशαϯϓϧ਺͕গͳ͔ͬͨ͜ͱɺ 41"5*"- */%*$"503ͱͷڑ཭͕େ͖͔ͬͨ͜ͱ͕ݪҼͰείΞ͕௿͔ͬͨͱߟ࡯͍ͯ͠Δ Datta S, Si Y, Rodriguez L, Shooshan SE, Demner-Fushman D, Roberts K. Understanding spatial language in radiology: Representation framework, annotation, and spatial relation extraction from chest X-ray reports using deep learning. J Biomed Inform. 2020;32.

ͦͷଞ • $5Ϩϙʔτ͔Βͷଁثʹؔ͢ΔʮసҠʯͷ༗ແΛػցֶशͰ෼ྨͨ͠ݚڀ 1BUUFSOTPG.FUBTUBUJD%JTFBTFJO1BUJFOUTXJUI$BODFS%FSJWFEGSPN/BUVSBM-BOHVBHF1SPDFTTJOHPG4USVDUVSFE$5 3BEJPMPHZ3FQPSUTPWFSBZFBS1FSJPE3BEJPMPHZ • $5Ϩϙʔτ͔Βʮഏ݁અʯͱͦͷ෦Ґɾಛ௃ɾαΠζͳͲΛϧʔϧϕʔεͰநग़ͨ͠ݚڀ /BUVSBM-BOHVBHF1SPDFTTJOHUP*EFOUJGZ1VMNPOBSZ/PEVMFTBOE&YUSBDU/PEVMF$IBSBDUFSJTUJDT'SPN3BEJPMPHZ 3FQPSUT$IFTU/PW

natural language processing × radiology reports

natural language processing × radiology reports

kento sugimoto

More Decks by kento sugimoto

Other Decks in Science

Featured

Transcript

/BUVSBMMBOHVBHFQSPDFTTJOH º 3BEJPMPHZ3FQPSUT

"QQMJDBUJPODBUFHPSZʢ$BTFZ FUBMʣ %JTFBTFJOGPSNBUJPO$MBTTJGJDBUJPO ü ࣬ױͷ༗ແͳͲΛ෼ྨ͢ΔͨΊͷ৘ใΛϨϙʔτ͔Βநग़͢Δ͜ͱΛ໨ࢦ͢ %JBHOPTUJDTVSWFJMMBODF ü ױऀͷ࣬ױঢ়ଶΛϑΥϩʔ͢ΔͨΊɺ࣬ױʹؔ͢Δ৘ใΛαʔϕΠϥϯε͢Δ͜ͱΛ໨ࢦ͢

"QQMJDBUJPODBUFHPSZʢ$BTFZ FUBMʣ • ϨϏϡʔʢ$BTFZ FUBMʣͷΧςΰϦʔͰ͸ɺ ʮ%JTFBTFJOGPSNBUJPO$MBTTJGJDBUJPOʯͱ ʮ%JBHOPTUJDTVSWFJMMBODFʯ͕൒਺Ҏ্Λ઎Ί͍ͯͨ • ϨϏϡʔʢ$BTFZ FUBMʣ

"QQMJDBUJPODBUFHPSZʢ$BTFZ FUBMʣ %JTFBTFJOGPSNBUJPO$MBTTJGJDBUJPO ü ࣬ױͷ༗ແͳͲΛ෼ྨ͢ΔͨΊͷ৘ใΛϨϙʔτ͔Βநग़͢Δ͜ͱΛ໨ࢦ͢ %JBHOPTUJDTVSWFJMMBODF ü ױऀͷ࣬ױঢ়ଶΛϑΥϩʔ͢ΔͨΊɺ࣬ױʹؔ͢Δ৘ใΛαʔϕΠϥϯε͢Δ͜ͱΛ໨ࢦ͢

%FFQ-FTJPO 8BOHFUBM • $5ɾ.3ͳͲ༷ʑͳϞμϦςΟ͔Β੒ΔΞϊςʔγϣϯ෇͖ͷҩྍ ը૾ͷσʔληοτʢ/)ͷ1"$4ʹ஝ੵ͞Ε͍ͯΔ΋ͷΛར༻ʣ • ը૾ʹ͸Ξϊςʔγϣϯπʔϧʢ3&$45 ͳͲʣʹΑͬͯɺը૾಺ ͷපมʹΞϊςʔγϣϯ৘ใʢ#PPLNBSLʣ͕෇༩͞Ε͍ͯΔ͜ͱ ͕ଟ͍ʢӈਤྫʣ

1FOHFUBM • Ϩϙʔτ͔Βපมͷಛ௃ʢҐஔɾαΠζɾܗ࣭ͳͲʣʹ ؔ͢Δ৘ใΛநग़ͨ͠ݚڀ • %FFQ-FTJPOͷCPPLNBSLͷ͋Δը૾ʹඥͮ͘Ϩϙʔτ Λར༻ʢӈਤͷΑ͏ʹ͕ϨϙʔτʹCPPLNBSL͕ຒΊࠐ ·Ε͍ͯΔʣ • /&3πʔϧʢ(&/*"5BHHFSʣΛ࢖ͬͯɺΤϯςΟςΟ

%BWJEFUBM • ಄෦.3*ϨϙʔτΛࣗಈ෼ྨ͢ΔϞσϧΛߏங • ෼ྨλεΫͱͯ͠ҎԼͷ̎ͭͷίʔύεΛ४උ • $PBSTFOPSNBMWTBCOPSNBM • (SBOVMBSEBNBHF WBTDVMBS

*SWJOFUBM$IF9QFSU Irvin J, Rajpurkar P, Ko M, Yu Y,

4NJUFUBM$IF9CFSU • ڳ෦୯७9ઢϨϙʔτ͔Βॴݟͷ༗ແΛ෼ྨ͢ΔϞσϧͷఏҊ • ..$$93ͱ$IF9QFSU *SWJOFUBM ͷσʔληοτ ʢ֤Ϩϙʔτʹʮॴݟͳ͠ʯΛؚΉछྨͷॴݟϥϕϧ͕Ϛϧ νϥϕϧͱͯ͠෇༩͞Ε͍ͯΔʣΛར༻ɻ •

4NJUFUBM$IF9CFSU • ࣄલֶशϞσϧͱͯ͠͸ɺʮ#JP#&35 -FFFUBM ɾ$MJOJDBM#&35 "MUPTBBS FUBM ɾ#MVF#&35 1FOHFUBM

+BJOFUBM3BE(SBQI • ڳ෦୯७9ઢϨϙʔτʢ..$$93 $IF9QFSUʣ ʹΤϯςΟςΟͱϦϨʔγϣϯͷϥϕϧΛ෇༩ͨ͠ σʔληοτΛެ։ • ʮ"OBUPNZɾ0CTFSWBUJPOʯΤϯςΟςΟΛ༻ҙ • 0CTFSWBUJPO͸ʮ%FGJOJUFMZ1SFTFOUɾ6ODFSUBJOɾ

+BJOFUBM3BE(SBQI • ΤϯςΟςΟநग़͸WBOJMMB#&35ʢUPLFODMBTTJGJDBUJPOʣ • ϦϨʔγϣϯநग़͸3#&35 8VFUBM • ֤λεΫͷ࠷ߴείΞʢ'TDPSFʣ͸ҎԼͷ௨Γ • /&3ʢ..$$93$IF9QFSUʣ

"QQMJDBUJPODBUFHPSZʢ$BTFZ FUBMʣ %JTFBTFJOGPSNBUJPO$MBTTJGJDBUJPO ü ࣬ױͷ༗ແͳͲΛ෼ྨ͢ΔͨΊͷ৘ใΛϨϙʔτ͔Βநग़͢Δ͜ͱΛ໨ࢦ͢ %JBHOPTUJDTVSWFJMMBODF ü ױऀͷ࣬ױঢ়ଶΛϑΥϩʔ͢ΔͨΊɺ࣬ױʹؔ͢Δ৘ใΛαʔϕΠϥϯε͢Δ͜ͱΛ໨ࢦ͢

,FIM FUBM • ݁Ռͷଥ౰ੑʢ৴པੑʣͷධՁͱͯ͠ɺҎԼΛ४උ • %'4 %JTFBTFGSFFTVSWJWBM • 1'4 1SPHSFTTJPOGSFFTVSWJWBM

#P[LVSUFUBM • நग़ͨ͠NFBTVSFNFOUͱͦͷಛ௃৘ใ͸ϧʔϧ ϕʔεͰඥ෇͚͍ͯΔ • ඥ෇͚ޙɺӈਤͷΑ͏ͳNFBTVSFNFOUΛओͱͨ͠ ϑϨʔϜΛߏங͢Δ • NFBTVSFNFOUͷநग़ਫ਼౓͸͕ͩɺߏஙͨ͠ ϑϨʔϜࣗମ͕ਖ਼ղσʔλͱ׬શҰகͨ͠ͷ͸શ

:FUJTHFO FUBM • ෼ྨϞσϧͱͯ͠͸ʮର਺ઢܗճؼʢ࠷େΤϯτϩϐʔ๏ʣʯΛ࢖༻ • 104 ʢ1BSUPG4QFFDIʣɾOHSBNTʢ ʣʹՃ͑ͯɺ.FUB.BQΛࢀরͯ͠ૉੑΛ࡞੒ •

Lou et al. 2020) • Ϩϙʔτ͔ΒϑΥϩʔΞοϓͷཁ൱Λ෼ྨ͢ΔݚڀʢݚڀͷϞνϕʔγϣϯ͸:FUJTHFOΒͱಉ༷ʣ • ࢖༻σʔληοτʹ͸ɺઌߦݚڀ ;BGBS FUBM

Zafar et al. 2015) Zafar HM, Chadalavada SC, Kahn CE,

Lou et al. 2020) • ϥϕϧͷ͏ͪʮJOEFUFSNJOBUFɾTVTQJDJPVTʯΛʮཁϑΥϩʔΞοϓʯͱͯ͠ݕग़͢Δର৅ͱͨ͠ • ʮཁϑΥϩʔΞοϓʯͷϨϙʔτΛ෼ྨ͢Δೋ஋෼ྨ໰୊ͱͯ͠ѻ͍ɺOHSBNTʢ ʣΛಛ௃ྔͱ

"QQMJDBUJPODBUFHPSZʢ$BTFZ FUBMʣ %JTFBTFJOGPSNBUJPO$MBTTJGJDBUJPO ü ࣬ױͷ༗ແͳͲΛ෼ྨ͢ΔͨΊͷ৘ใΛϨϙʔτ͔Βநग़͢Δ͜ͱΛ໨ࢦ͢ %JBHOPTUJDTVSWFJMMBODF ü ױऀͷ࣬ױঢ়ଶΛϑΥϩʔ͢ΔͨΊɺ࣬ױʹؔ͢Δ৘ใΛαʔϕΠϥϯε͢Δ͜ͱΛ໨ࢦ͢

'MFJTDIOFS 4PDJFUZ(VJEFMJOFT Ұ෦ൈਮ

"QQMJDBUJPODBUFHPSZʢ$BTFZ FUBMʣ %JTFBTFJOGPSNBUJPO$MBTTJGJDBUJPO ü ࣬ױͷ༗ແͳͲΛ෼ྨ͢ΔͨΊͷ৘ใΛϨϙʔτ͔Βநग़͢Δ͜ͱΛ໨ࢦ͢ %JBHOPTUJDTVSWFJMMBODF ü ױऀͷ࣬ױঢ়ଶΛϑΥϩʔ͢ΔͨΊɺ࣬ױʹؔ͢Δ৘ใΛαʔϕΠϥϯε͢Δ͜ͱΛ໨ࢦ͢

9JF FUBM • Ϩϙʔτ͔ΒΤϯςΟςΟͱͦͷؔ܎ͷநग़Λߦͬͨݚڀ • 5VNPSʹϑΥʔΧεͨ͠ΤϯςΟςΟͷεΩʔϚΛ༻ҙʢӈදʣ • ̎ࢪઃ͔ΒूΊͨڳ෦$5Ϩϙʔτʢ໿ ݅ʣΛ࢖༻ •

Liu et al. 2020) • ΤϯςΟςΟநग़ɾਖ਼نԽɾؔ܎நग़ͷύΠϓϥΠ ϯΛߏங͠ɺෲ෦$5Ϩϙʔτͷߏ଄ԽΛࢼΈͨ • ྟচԠ༻ͱͯ͠ɺૣظൃݟ͕๬·ΕΔ؊͕Μͷױऀ ʹয఺Λ౰ͯͯɺ༧ଌϞσϧͷ݁ՌΛධՁ͍ͯ͠Δ

Liu et al. 2020) • ΤϯςΟςΟநग़Ϟσϧͱͯ͠-45. $3'Λ࢖༻͠ɺจࣈྻͷ෼ࢄදݱΛಛ௃ྔͱͯ͠ೖྗ • จࣈྻ৘ใʹՃ͑ͯɺࣄલʹ༻ҙͨ͠ޠኮϦετͱͷϚονϯά݁ՌΛಛ௃ྔʹؚΊͨ •

Liu et al. 2020) • நग़ͨ͠ΤϯςΟςΟ͔Βɺ-PDBUJPOΛΩʔͱͯ͠ɺϧʔ ϧϕʔεͰΤϯςΟςΟͷ૊Έ߹ΘͤΛ࡞Δʢӈྫʣ • Ϩϙʔτ͔Βநग़͞Εͨ૊Έ߹Θͤͷ༗ແͷόΠφϦ஋Λ ಛ௃ྔͱͯ͠ɺ؊͕Μͷ༗ແΛ༧ଌ͢Δ෼ྨثΛߏங

8BEJBFUBM • $BODFS$BSF5SBDLJOH4ZTUFN $$54 Ͱ͸ɺಡӨ࣌ʹ์ࣹઢՊҩͱ/-1ΞϧΰϦζϜ͕ɺഏ͕Μͷ ͍͕ٙ͋ΔϨϙʔτʹίʔυΛ෇༩͍ͯ͠Δ • $$54ʹ͓͚ΔਓखͷίʔσΟϯάʢ.BOVBMʣͱ/-1ΞϧΰϦζϜͷ݁ՌΛൺֱ͠ɺੑೳධՁΛߦ͏ • /-1ΞϧΰϦζϜͱͯ͠͸ʮD5",&4ʢ4BWPWB

8BEJBFUBM • .BOVBMͱൺֱͯ͠ɺ/-1͸ײ౓͸ߴ͔͕ͬͨɺ ಛҟ౓͸௿͔ͬͨʢཅੑ༧ଌ஋͸ಉఔ౓ʣ • $PNCJOFEʢͲͪΒ͔͕1PTJUJWFͰ͋Ε͹ 1PTJUJWFʣͷ৔߹ɺײ౓͸Λه࿥ • ґཔප໊ผͰݟΔͱɺײ౓ͷࠩ͸ʮ͕ΜҎ֎ͷ ґཔʯͷϨϙʔτͰΑΓݦஶͩͬͨʢ.BOVBM

"QQMJDBUJPODBUFHPSZʢ$BTFZ FUBMʣ %JTFBTFJOGPSNBUJPO$MBTTJGJDBUJPO ü ࣬ױͷ༗ແͳͲΛ෼ྨ͢ΔͨΊͷ৘ใΛϨϙʔτ͔Βநग़͢Δ͜ͱΛ໨ࢦ͢ %JBHOPTUJDTVSWFJMMBODF ü ױऀͷ࣬ױঢ়ଶΛϑΥϩʔ͢ΔͨΊɺ࣬ױʹؔ͢Δ৘ใΛαʔϕΠϥϯε͢Δ͜ͱΛ໨ࢦ͢

1BOJDFL FUBM • ์ࣹઢϨϙʔτʹ༻͍ΒΕΔ਍அͷʮෆ࣮֬ੑʢ6ODFSUBJOUZʣ ʯ Λࣔ͢දݱ͸์ࣹઢՊҩʹΑ༷ͬͯʑͰ͋Γɺ·ͨ֬৴౓ͷ౓߹͍ ʹ͍ͭͯ͸ղऍ͕Ұக͠ͳ͍͜ͱ͕ଟ͍ʢ,IPSBTBOJ FUBMʣ • ༻ޠͷղऍͷζϨ΍ίϛϡχέʔγϣϯϛεΛݮΒͨ͢Ίɺӈදͷ

%BUUBFUBM • ςΩετதͷʮۭؒ৘ใʯΛೝࣝ͢Δʮ4QBUJBMSPMFMBCFMJOH ,PSEKBNTIJEJ FUBM ʯ ʢ4Q3-ʣΛԠ༻ͯ͠ɺڳ෦9ઢϨϙʔτͷʮۭؒ৘ใʯΛߏ଄Խͨ͠ݚڀ • 4Q3-͸ɺओʹҎԼͷΑ͏ͳεΩʔϚ·Ͱߏ੒͞ΕΔ