Overview of the main ML applications in urinalysis included in this review.
| Reference . | Format . | Patient population . | Purpose . | Data set (train/test split) . | Used features . | Best model . | Result (cross)-validation set . | Result test set . |
|---|---|---|---|---|---|---|---|---|
| Jang et al., 2023 (5) | Multicenter, retrospective | Heterogenous inpatient population (university hospital and diabetes center) for development; outpatients for external validation | Predict impaired eGFR using ML models comprising urine test strip parameters, age, and gender | 357.434 patients (198.015/22.003/74.380/62.945; training/internal validation/external validation 1/external validation 2) | Age, gender, 5 urine test strip parameters (protein, blood, glucose, pH, specific gravity) | XGBoost | eGFR <60 mL/min/1.73 m2: AUC of 0.91 (95%CI: 0.91–0.92); eGFR <45 mL/min/1.73 m2: AUC of 0.94 (95%CI: 0.94–0.95) | Test set 1: eGFR <60 mL/min/ 1.73 m2: AUC of 0.91 (95%CI: 0.90–0.92); eGFR <45 mL/min/ 1.73 m2: AUC of 0.95 (95%CI: 0.93–0.96) Test set 2: eGFR <60 mL/min/ 1.73 m2: AUC of 0.92 (95%CI: 0.91–0.93); eGFR <45 mL/min/ 1.73 m2: AUC of 0.94 (95%CI: 0.93–0.96) |
| Taylor et al., 2018 (49) | Single center, retrospective | Emergency department | Identifying the AI algorithm that has the highest diagnostic performance for UTI diagnosis using clinical symptoms and urine particle analysis results | 80.387 patients (64.310/16.077) | Age, gender, WBC, nitrates, leukocytes, bacteria, blood, epithelial cells, history of previous UTI, dysuria | XGBoost | AUC: 0.904 (95%CI: 0.898–0.910) Sensitivity: 61.7% (95%CI: 60.0–63.3%) Specificity: 94.9% (95%CI: 94.5–95.3%) | AUC: 0.858 (95%CI: 0.853–0.863) Sensitivity: 73.8% (95%CI: 72.3–75.2%) Specificity: 89.2% (95%CI: 88.6–89.8%) |
| Burton RJ et al., 2019 (50) | Single center, Retrospective | Heterogenous hospital population | Using AI to reduce the number of urinary cultures without compromising the detection of UTI | 212.554 urine reports (157.645/67.562) | Demographics, historical urine culture results, urine sediment results, clinical information | XGBoost | AUC: 0.910 Sensitivity: 96.7% (95%CI: 96.52–96.86) Specificity: 54.1% (95%CI: 53.5–54.8%) | Sensitivity: 95.2% (95%CI: 95.0%–95.4%) Specificity: 60.9% (95%CI: 60.3–61.6%) |
| Advanced analytics group of pediatric urology, 2019 (51) | Observational cohort study | Children | Identifying children with an initial UTI who are at risk for rUTI and VUR | 500 children (440/79) | Age, gender, race, weight, systolic blood pressure percentile, dysuria, urine albumin/creatinine ratio, prior antibiotic exposure, medication | Optimal classification tree | AUC: 0.761 (95%CI: 0.714–0.808)a | None |
| Wilkes et al., 2018 (52) | Single center, retrospective | Routine clinical practice | Application of ML algorithms to the automated interpretation of urine steroid profiles | 4916 urine steroid profiles | Up to 45 different features including steroid metabolites quantified by GC–MS and demographic data | WSRF model for binary classification, RF for multiclass classification | WSRF (normal versus abnormal): AUC of 0.955 (95%CI, 0.949–0.961). RF (multiclass): mean balanced accuracy of 0.873 (0.865– 0.880) | None |
| Chortis et al., 2019 (53) | Multicenter, longitudinal | Patients with histologically confirmed ACC, who had undergone microscopically complete (R0) tumor resection | Evaluating the performance of urine steroid metabolomics as a tool for postoperative recurrence detection after microscopically complete (R0) resection of ACC | 135 patients | Steroid metabolites quantified by gas chromatography–mass spectrometry | RF | AUC: 0.89 (95%CI 0.86–0.91) Sensitivity = specificity = 81% | None |
| Ni et al., 2021 (54) | Single center, retrospective | Patients with ovarian carcinoma (73 malignant and 59 benign) | Develop a classifier incorporating a urinary protein panel to classify benign and malignant ovarian tumors | 132 patients Train/test/extra validation set: 70/20/42 | Five proteins: WFDC2, PTMA, PVRL4, FIBA, and PVRL2 | RF | AUC: 0.980, sensitivity 0.967, specificity 0.900 | Test: AUC: 0.970, sensitivity 0.900, specificity 0.900 Extra validation set: AUC 0.952, sensitivity 0.895, specificity 0.913 |
| Bifarin et al., 2021 (55) | Single center, prospective | 105 patients with RCC and 179 controls | RCC status prediction Using multiplatform metabolomics | 256 patients (62/194) | 7-metabolite panel for RCC that included 2-phenylacetamide, Lys-Ile (or Lys-leu), dibutylamine, hippuric acid, mannitol hippurate, 2-mercaptobenzothiazole, and N-acetyl-glucosaminic acid | Linear SVM | Not provided | 88% accuracy, 94% sensitivity, 85% specificity, and 0.98 AUC |
| Cani et al., 2022 (56) | Single center, retrospective | 109 patients representing the spectrum of disease (benign to GG 5 prostate cancer) | Development of a next-generation RNA-sequencing assay for early detection of aggressive prostate cancer | 109 patients Training/validation split: 73/36 | 15 targets including TMPRSS2-ERG splicing isoforms, additional mRNAs, lncRNAs, and other current clinical biomarkers | RF feature-reduction process followed by logistic regression | AUC: 0.82 (95%CI 0.65–0.98) | None |
| Wang et al., 2021 (57) | Multicenter, prospective | Patients with bladder cancer (n = 270) and controls (=261) | Development of a gene expression assay for noninvasive detection of bladder cancer | 531 patients (211/320) | 32-gene signature | SVM | Accuracy: 92.68% | Accuracy: 89.9% (95%CI, 86%–93%) Sensitivity: 82.6% (95%CI, 75%–88%) Specificity: 95.1% (95%CI, 91%–98%) AUC: 0.932 (95%CI: 90%–96%) |
| Reference . | Format . | Patient population . | Purpose . | Data set (train/test split) . | Used features . | Best model . | Result (cross)-validation set . | Result test set . |
|---|---|---|---|---|---|---|---|---|
| Jang et al., 2023 (5) | Multicenter, retrospective | Heterogenous inpatient population (university hospital and diabetes center) for development; outpatients for external validation | Predict impaired eGFR using ML models comprising urine test strip parameters, age, and gender | 357.434 patients (198.015/22.003/74.380/62.945; training/internal validation/external validation 1/external validation 2) | Age, gender, 5 urine test strip parameters (protein, blood, glucose, pH, specific gravity) | XGBoost | eGFR <60 mL/min/1.73 m2: AUC of 0.91 (95%CI: 0.91–0.92); eGFR <45 mL/min/1.73 m2: AUC of 0.94 (95%CI: 0.94–0.95) | Test set 1: eGFR <60 mL/min/ 1.73 m2: AUC of 0.91 (95%CI: 0.90–0.92); eGFR <45 mL/min/ 1.73 m2: AUC of 0.95 (95%CI: 0.93–0.96) Test set 2: eGFR <60 mL/min/ 1.73 m2: AUC of 0.92 (95%CI: 0.91–0.93); eGFR <45 mL/min/ 1.73 m2: AUC of 0.94 (95%CI: 0.93–0.96) |
| Taylor et al., 2018 (49) | Single center, retrospective | Emergency department | Identifying the AI algorithm that has the highest diagnostic performance for UTI diagnosis using clinical symptoms and urine particle analysis results | 80.387 patients (64.310/16.077) | Age, gender, WBC, nitrates, leukocytes, bacteria, blood, epithelial cells, history of previous UTI, dysuria | XGBoost | AUC: 0.904 (95%CI: 0.898–0.910) Sensitivity: 61.7% (95%CI: 60.0–63.3%) Specificity: 94.9% (95%CI: 94.5–95.3%) | AUC: 0.858 (95%CI: 0.853–0.863) Sensitivity: 73.8% (95%CI: 72.3–75.2%) Specificity: 89.2% (95%CI: 88.6–89.8%) |
| Burton RJ et al., 2019 (50) | Single center, Retrospective | Heterogenous hospital population | Using AI to reduce the number of urinary cultures without compromising the detection of UTI | 212.554 urine reports (157.645/67.562) | Demographics, historical urine culture results, urine sediment results, clinical information | XGBoost | AUC: 0.910 Sensitivity: 96.7% (95%CI: 96.52–96.86) Specificity: 54.1% (95%CI: 53.5–54.8%) | Sensitivity: 95.2% (95%CI: 95.0%–95.4%) Specificity: 60.9% (95%CI: 60.3–61.6%) |
| Advanced analytics group of pediatric urology, 2019 (51) | Observational cohort study | Children | Identifying children with an initial UTI who are at risk for rUTI and VUR | 500 children (440/79) | Age, gender, race, weight, systolic blood pressure percentile, dysuria, urine albumin/creatinine ratio, prior antibiotic exposure, medication | Optimal classification tree | AUC: 0.761 (95%CI: 0.714–0.808)a | None |
| Wilkes et al., 2018 (52) | Single center, retrospective | Routine clinical practice | Application of ML algorithms to the automated interpretation of urine steroid profiles | 4916 urine steroid profiles | Up to 45 different features including steroid metabolites quantified by GC–MS and demographic data | WSRF model for binary classification, RF for multiclass classification | WSRF (normal versus abnormal): AUC of 0.955 (95%CI, 0.949–0.961). RF (multiclass): mean balanced accuracy of 0.873 (0.865– 0.880) | None |
| Chortis et al., 2019 (53) | Multicenter, longitudinal | Patients with histologically confirmed ACC, who had undergone microscopically complete (R0) tumor resection | Evaluating the performance of urine steroid metabolomics as a tool for postoperative recurrence detection after microscopically complete (R0) resection of ACC | 135 patients | Steroid metabolites quantified by gas chromatography–mass spectrometry | RF | AUC: 0.89 (95%CI 0.86–0.91) Sensitivity = specificity = 81% | None |
| Ni et al., 2021 (54) | Single center, retrospective | Patients with ovarian carcinoma (73 malignant and 59 benign) | Develop a classifier incorporating a urinary protein panel to classify benign and malignant ovarian tumors | 132 patients Train/test/extra validation set: 70/20/42 | Five proteins: WFDC2, PTMA, PVRL4, FIBA, and PVRL2 | RF | AUC: 0.980, sensitivity 0.967, specificity 0.900 | Test: AUC: 0.970, sensitivity 0.900, specificity 0.900 Extra validation set: AUC 0.952, sensitivity 0.895, specificity 0.913 |
| Bifarin et al., 2021 (55) | Single center, prospective | 105 patients with RCC and 179 controls | RCC status prediction Using multiplatform metabolomics | 256 patients (62/194) | 7-metabolite panel for RCC that included 2-phenylacetamide, Lys-Ile (or Lys-leu), dibutylamine, hippuric acid, mannitol hippurate, 2-mercaptobenzothiazole, and N-acetyl-glucosaminic acid | Linear SVM | Not provided | 88% accuracy, 94% sensitivity, 85% specificity, and 0.98 AUC |
| Cani et al., 2022 (56) | Single center, retrospective | 109 patients representing the spectrum of disease (benign to GG 5 prostate cancer) | Development of a next-generation RNA-sequencing assay for early detection of aggressive prostate cancer | 109 patients Training/validation split: 73/36 | 15 targets including TMPRSS2-ERG splicing isoforms, additional mRNAs, lncRNAs, and other current clinical biomarkers | RF feature-reduction process followed by logistic regression | AUC: 0.82 (95%CI 0.65–0.98) | None |
| Wang et al., 2021 (57) | Multicenter, prospective | Patients with bladder cancer (n = 270) and controls (=261) | Development of a gene expression assay for noninvasive detection of bladder cancer | 531 patients (211/320) | 32-gene signature | SVM | Accuracy: 92.68% | Accuracy: 89.9% (95%CI, 86%–93%) Sensitivity: 82.6% (95%CI, 75%–88%) Specificity: 95.1% (95%CI, 91%–98%) AUC: 0.932 (95%CI: 90%–96%) |
Abbreviations: WSRF, weighted-subspace RF; FIBA, fibrinogen alpha chain; GG, grade group; PTMA, prothymosin alpha; PVRL2, poliovirus receptor-related 2; PVRL4, poliovirus receptor-related 4; WFDC2, WAP four-disulfide core domain protein 2.
aThe authors do not mention a sensitivity and specificity associated with the AUC.
Overview of the main ML applications in urinalysis included in this review.
| Reference . | Format . | Patient population . | Purpose . | Data set (train/test split) . | Used features . | Best model . | Result (cross)-validation set . | Result test set . |
|---|---|---|---|---|---|---|---|---|
| Jang et al., 2023 (5) | Multicenter, retrospective | Heterogenous inpatient population (university hospital and diabetes center) for development; outpatients for external validation | Predict impaired eGFR using ML models comprising urine test strip parameters, age, and gender | 357.434 patients (198.015/22.003/74.380/62.945; training/internal validation/external validation 1/external validation 2) | Age, gender, 5 urine test strip parameters (protein, blood, glucose, pH, specific gravity) | XGBoost | eGFR <60 mL/min/1.73 m2: AUC of 0.91 (95%CI: 0.91–0.92); eGFR <45 mL/min/1.73 m2: AUC of 0.94 (95%CI: 0.94–0.95) | Test set 1: eGFR <60 mL/min/ 1.73 m2: AUC of 0.91 (95%CI: 0.90–0.92); eGFR <45 mL/min/ 1.73 m2: AUC of 0.95 (95%CI: 0.93–0.96) Test set 2: eGFR <60 mL/min/ 1.73 m2: AUC of 0.92 (95%CI: 0.91–0.93); eGFR <45 mL/min/ 1.73 m2: AUC of 0.94 (95%CI: 0.93–0.96) |
| Taylor et al., 2018 (49) | Single center, retrospective | Emergency department | Identifying the AI algorithm that has the highest diagnostic performance for UTI diagnosis using clinical symptoms and urine particle analysis results | 80.387 patients (64.310/16.077) | Age, gender, WBC, nitrates, leukocytes, bacteria, blood, epithelial cells, history of previous UTI, dysuria | XGBoost | AUC: 0.904 (95%CI: 0.898–0.910) Sensitivity: 61.7% (95%CI: 60.0–63.3%) Specificity: 94.9% (95%CI: 94.5–95.3%) | AUC: 0.858 (95%CI: 0.853–0.863) Sensitivity: 73.8% (95%CI: 72.3–75.2%) Specificity: 89.2% (95%CI: 88.6–89.8%) |
| Burton RJ et al., 2019 (50) | Single center, Retrospective | Heterogenous hospital population | Using AI to reduce the number of urinary cultures without compromising the detection of UTI | 212.554 urine reports (157.645/67.562) | Demographics, historical urine culture results, urine sediment results, clinical information | XGBoost | AUC: 0.910 Sensitivity: 96.7% (95%CI: 96.52–96.86) Specificity: 54.1% (95%CI: 53.5–54.8%) | Sensitivity: 95.2% (95%CI: 95.0%–95.4%) Specificity: 60.9% (95%CI: 60.3–61.6%) |
| Advanced analytics group of pediatric urology, 2019 (51) | Observational cohort study | Children | Identifying children with an initial UTI who are at risk for rUTI and VUR | 500 children (440/79) | Age, gender, race, weight, systolic blood pressure percentile, dysuria, urine albumin/creatinine ratio, prior antibiotic exposure, medication | Optimal classification tree | AUC: 0.761 (95%CI: 0.714–0.808)a | None |
| Wilkes et al., 2018 (52) | Single center, retrospective | Routine clinical practice | Application of ML algorithms to the automated interpretation of urine steroid profiles | 4916 urine steroid profiles | Up to 45 different features including steroid metabolites quantified by GC–MS and demographic data | WSRF model for binary classification, RF for multiclass classification | WSRF (normal versus abnormal): AUC of 0.955 (95%CI, 0.949–0.961). RF (multiclass): mean balanced accuracy of 0.873 (0.865– 0.880) | None |
| Chortis et al., 2019 (53) | Multicenter, longitudinal | Patients with histologically confirmed ACC, who had undergone microscopically complete (R0) tumor resection | Evaluating the performance of urine steroid metabolomics as a tool for postoperative recurrence detection after microscopically complete (R0) resection of ACC | 135 patients | Steroid metabolites quantified by gas chromatography–mass spectrometry | RF | AUC: 0.89 (95%CI 0.86–0.91) Sensitivity = specificity = 81% | None |
| Ni et al., 2021 (54) | Single center, retrospective | Patients with ovarian carcinoma (73 malignant and 59 benign) | Develop a classifier incorporating a urinary protein panel to classify benign and malignant ovarian tumors | 132 patients Train/test/extra validation set: 70/20/42 | Five proteins: WFDC2, PTMA, PVRL4, FIBA, and PVRL2 | RF | AUC: 0.980, sensitivity 0.967, specificity 0.900 | Test: AUC: 0.970, sensitivity 0.900, specificity 0.900 Extra validation set: AUC 0.952, sensitivity 0.895, specificity 0.913 |
| Bifarin et al., 2021 (55) | Single center, prospective | 105 patients with RCC and 179 controls | RCC status prediction Using multiplatform metabolomics | 256 patients (62/194) | 7-metabolite panel for RCC that included 2-phenylacetamide, Lys-Ile (or Lys-leu), dibutylamine, hippuric acid, mannitol hippurate, 2-mercaptobenzothiazole, and N-acetyl-glucosaminic acid | Linear SVM | Not provided | 88% accuracy, 94% sensitivity, 85% specificity, and 0.98 AUC |
| Cani et al., 2022 (56) | Single center, retrospective | 109 patients representing the spectrum of disease (benign to GG 5 prostate cancer) | Development of a next-generation RNA-sequencing assay for early detection of aggressive prostate cancer | 109 patients Training/validation split: 73/36 | 15 targets including TMPRSS2-ERG splicing isoforms, additional mRNAs, lncRNAs, and other current clinical biomarkers | RF feature-reduction process followed by logistic regression | AUC: 0.82 (95%CI 0.65–0.98) | None |
| Wang et al., 2021 (57) | Multicenter, prospective | Patients with bladder cancer (n = 270) and controls (=261) | Development of a gene expression assay for noninvasive detection of bladder cancer | 531 patients (211/320) | 32-gene signature | SVM | Accuracy: 92.68% | Accuracy: 89.9% (95%CI, 86%–93%) Sensitivity: 82.6% (95%CI, 75%–88%) Specificity: 95.1% (95%CI, 91%–98%) AUC: 0.932 (95%CI: 90%–96%) |
| Reference . | Format . | Patient population . | Purpose . | Data set (train/test split) . | Used features . | Best model . | Result (cross)-validation set . | Result test set . |
|---|---|---|---|---|---|---|---|---|
| Jang et al., 2023 (5) | Multicenter, retrospective | Heterogenous inpatient population (university hospital and diabetes center) for development; outpatients for external validation | Predict impaired eGFR using ML models comprising urine test strip parameters, age, and gender | 357.434 patients (198.015/22.003/74.380/62.945; training/internal validation/external validation 1/external validation 2) | Age, gender, 5 urine test strip parameters (protein, blood, glucose, pH, specific gravity) | XGBoost | eGFR <60 mL/min/1.73 m2: AUC of 0.91 (95%CI: 0.91–0.92); eGFR <45 mL/min/1.73 m2: AUC of 0.94 (95%CI: 0.94–0.95) | Test set 1: eGFR <60 mL/min/ 1.73 m2: AUC of 0.91 (95%CI: 0.90–0.92); eGFR <45 mL/min/ 1.73 m2: AUC of 0.95 (95%CI: 0.93–0.96) Test set 2: eGFR <60 mL/min/ 1.73 m2: AUC of 0.92 (95%CI: 0.91–0.93); eGFR <45 mL/min/ 1.73 m2: AUC of 0.94 (95%CI: 0.93–0.96) |
| Taylor et al., 2018 (49) | Single center, retrospective | Emergency department | Identifying the AI algorithm that has the highest diagnostic performance for UTI diagnosis using clinical symptoms and urine particle analysis results | 80.387 patients (64.310/16.077) | Age, gender, WBC, nitrates, leukocytes, bacteria, blood, epithelial cells, history of previous UTI, dysuria | XGBoost | AUC: 0.904 (95%CI: 0.898–0.910) Sensitivity: 61.7% (95%CI: 60.0–63.3%) Specificity: 94.9% (95%CI: 94.5–95.3%) | AUC: 0.858 (95%CI: 0.853–0.863) Sensitivity: 73.8% (95%CI: 72.3–75.2%) Specificity: 89.2% (95%CI: 88.6–89.8%) |
| Burton RJ et al., 2019 (50) | Single center, Retrospective | Heterogenous hospital population | Using AI to reduce the number of urinary cultures without compromising the detection of UTI | 212.554 urine reports (157.645/67.562) | Demographics, historical urine culture results, urine sediment results, clinical information | XGBoost | AUC: 0.910 Sensitivity: 96.7% (95%CI: 96.52–96.86) Specificity: 54.1% (95%CI: 53.5–54.8%) | Sensitivity: 95.2% (95%CI: 95.0%–95.4%) Specificity: 60.9% (95%CI: 60.3–61.6%) |
| Advanced analytics group of pediatric urology, 2019 (51) | Observational cohort study | Children | Identifying children with an initial UTI who are at risk for rUTI and VUR | 500 children (440/79) | Age, gender, race, weight, systolic blood pressure percentile, dysuria, urine albumin/creatinine ratio, prior antibiotic exposure, medication | Optimal classification tree | AUC: 0.761 (95%CI: 0.714–0.808)a | None |
| Wilkes et al., 2018 (52) | Single center, retrospective | Routine clinical practice | Application of ML algorithms to the automated interpretation of urine steroid profiles | 4916 urine steroid profiles | Up to 45 different features including steroid metabolites quantified by GC–MS and demographic data | WSRF model for binary classification, RF for multiclass classification | WSRF (normal versus abnormal): AUC of 0.955 (95%CI, 0.949–0.961). RF (multiclass): mean balanced accuracy of 0.873 (0.865– 0.880) | None |
| Chortis et al., 2019 (53) | Multicenter, longitudinal | Patients with histologically confirmed ACC, who had undergone microscopically complete (R0) tumor resection | Evaluating the performance of urine steroid metabolomics as a tool for postoperative recurrence detection after microscopically complete (R0) resection of ACC | 135 patients | Steroid metabolites quantified by gas chromatography–mass spectrometry | RF | AUC: 0.89 (95%CI 0.86–0.91) Sensitivity = specificity = 81% | None |
| Ni et al., 2021 (54) | Single center, retrospective | Patients with ovarian carcinoma (73 malignant and 59 benign) | Develop a classifier incorporating a urinary protein panel to classify benign and malignant ovarian tumors | 132 patients Train/test/extra validation set: 70/20/42 | Five proteins: WFDC2, PTMA, PVRL4, FIBA, and PVRL2 | RF | AUC: 0.980, sensitivity 0.967, specificity 0.900 | Test: AUC: 0.970, sensitivity 0.900, specificity 0.900 Extra validation set: AUC 0.952, sensitivity 0.895, specificity 0.913 |
| Bifarin et al., 2021 (55) | Single center, prospective | 105 patients with RCC and 179 controls | RCC status prediction Using multiplatform metabolomics | 256 patients (62/194) | 7-metabolite panel for RCC that included 2-phenylacetamide, Lys-Ile (or Lys-leu), dibutylamine, hippuric acid, mannitol hippurate, 2-mercaptobenzothiazole, and N-acetyl-glucosaminic acid | Linear SVM | Not provided | 88% accuracy, 94% sensitivity, 85% specificity, and 0.98 AUC |
| Cani et al., 2022 (56) | Single center, retrospective | 109 patients representing the spectrum of disease (benign to GG 5 prostate cancer) | Development of a next-generation RNA-sequencing assay for early detection of aggressive prostate cancer | 109 patients Training/validation split: 73/36 | 15 targets including TMPRSS2-ERG splicing isoforms, additional mRNAs, lncRNAs, and other current clinical biomarkers | RF feature-reduction process followed by logistic regression | AUC: 0.82 (95%CI 0.65–0.98) | None |
| Wang et al., 2021 (57) | Multicenter, prospective | Patients with bladder cancer (n = 270) and controls (=261) | Development of a gene expression assay for noninvasive detection of bladder cancer | 531 patients (211/320) | 32-gene signature | SVM | Accuracy: 92.68% | Accuracy: 89.9% (95%CI, 86%–93%) Sensitivity: 82.6% (95%CI, 75%–88%) Specificity: 95.1% (95%CI, 91%–98%) AUC: 0.932 (95%CI: 90%–96%) |
Abbreviations: WSRF, weighted-subspace RF; FIBA, fibrinogen alpha chain; GG, grade group; PTMA, prothymosin alpha; PVRL2, poliovirus receptor-related 2; PVRL4, poliovirus receptor-related 4; WFDC2, WAP four-disulfide core domain protein 2.
aThe authors do not mention a sensitivity and specificity associated with the AUC.
This PDF is available to Subscribers Only
View Article Abstract & Purchase OptionsFor full access to this pdf, sign in to an existing account, or purchase an annual subscription.
