View in EDS

Natural language processing for kidney ultrasound analysis: correlating imaging reports with chronic kidney disease diagnosis.

Saved in:
Bibliographic Details
Title: Natural language processing for kidney ultrasound analysis: correlating imaging reports with chronic kidney disease diagnosis.
Authors: Wang C; Department of Computer Science, Stony Brook University, NY, USA., Banerjee R; Department of Computer Science, Stony Brook University, NY, USA., Kuperstein H; Renaissance School of Medicine, Stony Brook University, NY, USA., Malick H; Renaissance School of Medicine, Stony Brook University, NY, USA., Bano R; Department of Medicine, Stony Brook University, NY, USA., Cunningham RL; Department of Radiology, Stony Brook University, NY, USA., Tahir H; Department of Medicine, Stony Brook University, NY, USA., Sakhuja P; Department of Medicine, Stony Brook University, NY, USA., Hajagos J; Department of Biomedical Informatics, Stony Brook University, NY, USA., Koraishy FM; Department of Medicine, Stony Brook University, NY, USA.
Source: Renal failure [Ren Fail] 2025 Dec; Vol. 47 (1), pp. 2539938. Date of Electronic Publication: 2025 Aug 04.
Publication Type: Journal Article
Language: English
Journal Info: Publisher: Informa Healthcare Country of Publication: England NLM ID: 8701128 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1525-6049 (Electronic) Linking ISSN: 0886022X NLM ISO Abbreviation: Ren Fail Subsets: MEDLINE
Imprint Name(s): Publication: London : Informa Healthcare
Original Publication: New York, N.Y. : M. Dekker, c1987-
MeSH Terms: Natural Language Processing* , Renal Insufficiency, Chronic*/diagnostic imaging , Renal Insufficiency, Chronic*/diagnosis , Kidney*/diagnostic imaging , Kidney*/pathology, Humans ; Ultrasonography/methods ; Female ; Male ; Middle Aged ; Pilot Projects ; Aged ; Adult ; Logistic Models
Abstract: Introduction: Natural language processing (NLP) has been used to analyze unstructured imaging report data, yet its application in identifying chronic kidney disease (CKD) features from kidney ultrasound reports remains unexplored.
Methods: In a single-center pilot study, we analyzed 1,068 kidney ultrasound reports using NLP techniques. To identify kidney echogenicity as either "normal" or "increased," we used two methods: one that looks at individual words and another that analyzes full sentences. Kidney length was identified as "small" if its length was below the 10th percentile. Nephrologists reviewed 100 randomly selected reports to create the reference standard (ground truth) for initial model training followed by model validation on an independent set of 100 reports.
Results: The word-level NLP model outperformed the sentence-level approach in classifying increased echogenicity (accuracy: 0.96 vs. 0.89 for the left kidney; 0.97 vs. 0.92 for the right kidney). This model was then applied to the full dataset to assess associations with CKD. Multivariable logistic regression identified bilaterally increased echogenicity as the strongest predictor of CKD (odds ratio [OR] = 7.642, 95% confidence interval [CI]: 4.887-11.949; p  < 0.0001), followed by bilaterally small kidneys (OR = 4.981 [1.522, 16.300]; p  = 0.008). Among individuals without CKD, those with bilaterally increased echogenicity had significantly lower kidney function than those with normal echogenicity.
Conclusions: State-of-the-art NLP models can accurately extract CKD-related features from ultrasound reports, with the potential of providing a scalable tool for early detection and risk stratification. Future research should focus on validating these models across different healthcare systems.
References: Lancet. 2021 Aug 28;398(10302):786-802. (PMID: 34175022)
Clin J Am Soc Nephrol. 2023 Jun 1;18(6):806-808. (PMID: 36758147)
NPJ Digit Med. 2019 Apr 26;2:29. (PMID: 31304376)
Nat Methods. 2020 Mar;17(3):261-272. (PMID: 32015543)
Kidney Int Rep. 2017 Mar;2(2):271-281. (PMID: 28439569)
Exp Ther Med. 2014 May;7(5):1259-1264. (PMID: 24940422)
Cureus. 2019 Mar 12;11(3):e4241. (PMID: 31131164)
Comput Methods Programs Biomed. 2021 Sep;209:106329. (PMID: 34418814)
J Nephrol. 2018 Dec;31(6):863-879. (PMID: 30191413)
Kidney Dis (Basel). 2023 Jan 23;9(3):187-196. (PMID: 37497203)
Clin J Am Soc Nephrol. 2016 Dec 7;11(12):2150-2158. (PMID: 27927892)
Kidney Int. 2005 Apr;67(4):1515-20. (PMID: 15780105)
J Am Med Inform Assoc. 2016 Sep;23(5):1007-15. (PMID: 26911811)
Eur Arch Otorhinolaryngol. 2022 Nov;279(11):5363-5373. (PMID: 35767056)
N Engl J Med. 2021 Nov 4;385(19):1737-1749. (PMID: 34554658)
Adv Chronic Kidney Dis. 2022 Sep;29(5):465-471. (PMID: 36253030)
Ren Fail. 2013 Jul;35(6):896-900. (PMID: 23751145)
Radiol Bras. 2015 Jan-Feb;48(1):12-6. (PMID: 25798002)
J Biomed Inform. 2022 Jan;125:103971. (PMID: 34920127)
Sci Rep. 2019 Aug 14;9(1):11862. (PMID: 31413285)
J Am Med Inform Assoc. 2010 Sep-Oct;17(5):588-94. (PMID: 20819869)
Int J Mol Sci. 2021 Jan 02;22(1):. (PMID: 33401711)
BMC Nephrol. 2020 Jul 20;21(1):289. (PMID: 32689967)
Bioinformatics. 2020 Feb 15;36(4):1234-1240. (PMID: 31501885)
Contributed Indexing: Keywords: Natural language processing; chronic kidney disease; deep syntactic structures; kidney ultrasound reports; language embeddings
Entry Date(s): Date Created: 20250805 Date Completed: 20250805 Latest Revision: 20250807
Update Code: 20250807
PubMed Central ID: PMC12322987
DOI: 10.1080/0886022X.2025.2539938
PMID: 40760725
Database: MEDLINE
Description
Abstract:Introduction: Natural language processing (NLP) has been used to analyze unstructured imaging report data, yet its application in identifying chronic kidney disease (CKD) features from kidney ultrasound reports remains unexplored.<br />Methods: In a single-center pilot study, we analyzed 1,068 kidney ultrasound reports using NLP techniques. To identify kidney echogenicity as either "normal" or "increased," we used two methods: one that looks at individual words and another that analyzes full sentences. Kidney length was identified as "small" if its length was below the 10th percentile. Nephrologists reviewed 100 randomly selected reports to create the reference standard (ground truth) for initial model training followed by model validation on an independent set of 100 reports.<br />Results: The word-level NLP model outperformed the sentence-level approach in classifying increased echogenicity (accuracy: 0.96 vs. 0.89 for the left kidney; 0.97 vs. 0.92 for the right kidney). This model was then applied to the full dataset to assess associations with CKD. Multivariable logistic regression identified bilaterally increased echogenicity as the strongest predictor of CKD (odds ratio [OR] = 7.642, 95% confidence interval [CI]: 4.887-11.949; p  &lt; 0.0001), followed by bilaterally small kidneys (OR = 4.981 [1.522, 16.300]; p  = 0.008). Among individuals without CKD, those with bilaterally increased echogenicity had significantly lower kidney function than those with normal echogenicity.<br />Conclusions: State-of-the-art NLP models can accurately extract CKD-related features from ultrasound reports, with the potential of providing a scalable tool for early detection and risk stratification. Future research should focus on validating these models across different healthcare systems.
ISSN:1525-6049
DOI:10.1080/0886022X.2025.2539938