Eike Elfert, Wolfgang E Kaminski, Christian Matek, Gregor Hoermann, Eyvind W Axelsen, Carsten Marr, Armin P Piehler
{"title":"Expert-level detection of M-proteins in serum protein electrophoresis using machine learning.","authors":"Eike Elfert, Wolfgang E Kaminski, Christian Matek, Gregor Hoermann, Eyvind W Axelsen, Carsten Marr, Armin P Piehler","doi":"10.1515/cclm-2024-0222","DOIUrl":null,"url":null,"abstract":"<p><strong>Objectives: </strong>Serum protein electrophoresis (SPE) in combination with immunotyping (IMT) is the diagnostic standard for detecting monoclonal proteins (M-proteins). However, interpretation of SPE and IMT is weakly standardized, time consuming and investigator dependent. Here, we present five machine learning (ML) approaches for automated detection of M-proteins on SPE on an unprecedented large and well-curated data set and compare the performance with that of laboratory experts.</p><p><strong>Methods: </strong>SPE and IMT were performed in serum samples from 69,722 individuals from Norway. IMT results were used to label the samples as M-protein present (positive, n=4,273) or absent (negative n=65,449). Four feature-based ML algorithms and one convolutional neural network (CNN) were trained on 68,722 randomly selected SPE patterns to detect M-proteins. Algorithm performance was compared to that of an expert group of clinical pathologists and laboratory technicians (n=10) on a test set of 1,000 samples.</p><p><strong>Results: </strong>The random forest classifier showed the best performance (F1-Score 93.2 %, accuracy 99.1 %, sensitivity 89.9 %, specificity 99.8 %, positive predictive value 96.9 %, negative predictive value 99.3 %) and outperformed the experts (F1-Score 61.2 ± 16.0 %, accuracy 89.2 ± 10.2 %, sensitivity 94.3 ± 2.8 %, specificity 88.9 ± 10.9 %, positive predictive value 47.3 ± 16.2 %, negative predictive value 99.5 ± 0.2 %) on the test set. Interestingly the performance of the RFC saturated, the CNN performance increased steadily within our training set (n=68,722).</p><p><strong>Conclusions: </strong>Feature-based ML systems are capable of automated detection of M-proteins on SPE beyond expert-level and show potential for use in the clinical laboratory.</p>","PeriodicalId":10390,"journal":{"name":"Clinical chemistry and laboratory medicine","volume":null,"pages":null},"PeriodicalIF":3.8000,"publicationDate":"2024-06-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Clinical chemistry and laboratory medicine","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1515/cclm-2024-0222","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"MEDICAL LABORATORY TECHNOLOGY","Score":null,"Total":0}
引用次数: 0
Abstract
Objectives: Serum protein electrophoresis (SPE) in combination with immunotyping (IMT) is the diagnostic standard for detecting monoclonal proteins (M-proteins). However, interpretation of SPE and IMT is weakly standardized, time consuming and investigator dependent. Here, we present five machine learning (ML) approaches for automated detection of M-proteins on SPE on an unprecedented large and well-curated data set and compare the performance with that of laboratory experts.
Methods: SPE and IMT were performed in serum samples from 69,722 individuals from Norway. IMT results were used to label the samples as M-protein present (positive, n=4,273) or absent (negative n=65,449). Four feature-based ML algorithms and one convolutional neural network (CNN) were trained on 68,722 randomly selected SPE patterns to detect M-proteins. Algorithm performance was compared to that of an expert group of clinical pathologists and laboratory technicians (n=10) on a test set of 1,000 samples.
Results: The random forest classifier showed the best performance (F1-Score 93.2 %, accuracy 99.1 %, sensitivity 89.9 %, specificity 99.8 %, positive predictive value 96.9 %, negative predictive value 99.3 %) and outperformed the experts (F1-Score 61.2 ± 16.0 %, accuracy 89.2 ± 10.2 %, sensitivity 94.3 ± 2.8 %, specificity 88.9 ± 10.9 %, positive predictive value 47.3 ± 16.2 %, negative predictive value 99.5 ± 0.2 %) on the test set. Interestingly the performance of the RFC saturated, the CNN performance increased steadily within our training set (n=68,722).
Conclusions: Feature-based ML systems are capable of automated detection of M-proteins on SPE beyond expert-level and show potential for use in the clinical laboratory.
期刊介绍:
Clinical Chemistry and Laboratory Medicine (CCLM) publishes articles on novel teaching and training methods applicable to laboratory medicine. CCLM welcomes contributions on the progress in fundamental and applied research and cutting-edge clinical laboratory medicine. It is one of the leading journals in the field, with an impact factor over 3. CCLM is issued monthly, and it is published in print and electronically.
CCLM is the official journal of the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) and publishes regularly EFLM recommendations and news. CCLM is the official journal of the National Societies from Austria (ÖGLMKC); Belgium (RBSLM); Germany (DGKL); Hungary (MLDT); Ireland (ACBI); Italy (SIBioC); Portugal (SPML); and Slovenia (SZKK); and it is affiliated to AACB (Australia) and SFBC (France).
Topics:
- clinical biochemistry
- clinical genomics and molecular biology
- clinical haematology and coagulation
- clinical immunology and autoimmunity
- clinical microbiology
- drug monitoring and analysis
- evaluation of diagnostic biomarkers
- disease-oriented topics (cardiovascular disease, cancer diagnostics, diabetes)
- new reagents, instrumentation and technologies
- new methodologies
- reference materials and methods
- reference values and decision limits
- quality and safety in laboratory medicine
- translational laboratory medicine
- clinical metrology
Follow @cclm_degruyter on Twitter!