Preliminary study on the application of renal ultrasonography radiomics in the classification of glomerulopathy

Machine learning-based diagnosis and prognosis of IgAN: A systematic review and meta-analysis

Purpose

Plenty of studies have explored the diagnosis and prognosis of IgA nephropathy (IgAN) based on machine learning (ML), but the accuracy lacks the support of evidence-based medical evidence. We aim at this problem to guide the precision treatment of IgAN.

Methods

Embase, Pubmed, Cochrane Library, and Web of Science were searched systematically until February 24th, 2024, for publications on ML-based diagnosis and prognosis of IgAN. Subgroup analysis or meta-regression was conducted according to modeling method, follow-up time, endpoint definition, and variable type. Further, the rank sum test was applied to compare the discrimination ability of prognosis.

Results

A total of 47 studies involving 51,935 patients were eligible. Among the 38 diagnostic models, the pooled C-index was 0.902 (95 % CI: 0.878-0.926) in 27 diagnostic models. Of the 162 prognostic models, the C-index for model discrimination of 144 prognostic models was 0.838 (95 % CI: 0.827-0.850) in training. The overall discrimination ability of prognosis was as follows: COX regression > new ML models (e.g. ANN, DT, RF, SVM, XGBoost) > traditional ML models (logistic regression) > Naïve Bayesian network (P < 0.05). External validation of IIgAN-RPT in 19 models showed a pooled C-index of 0.801 (95 % CI: 0.784-0.817).

Conclusions

New ML models have shown application values that are as good as traditional ML models, both in diagnosis and prognosis. In addition, future models are desired to use a more sensitive prognostic endpoint (albuminuria), improve predictive ability in moderate progression risk, and ultimately translate into clinically applicable intelligent tools.

Non-invasive prediction of the chronic degree of lupus nephropathy based on ultrasound radiomics

OBJECTIVE

Through machine learning (ML) analysis of the radiomics features of ultrasound extracted from patients with lupus nephritis (LN), this attempt was made to non-invasively predict the chronicity index (CI)of LN.

METHODS

A retrospective collection of 136 patients with LN who had renal biopsy was retrospectively collected, and the patients were randomly divided into training set and validation set according to 7:3. Radiomics features are extracted from ultrasound images, independent factors are obtained by using LASSO dimensionality reduction, and then seven ML models were used to establish predictive models. At the same time, a clinical model and an US model were established. The diagnostic efficacy of the model is evaluated by analysis of the receiver operating characteristics (ROC) curve, accuracy, specificity, and sensitivity. The performance of the seven machine learning models was compared with each other and with clinical and US models.

RESULTS

A total of 1314 radiomics features are extracted from ultrasound images, and 5 features are finally screened out by LASSO for model construction, and the average ROC of the seven ML is 0.683, among which the Xgboost model performed the best, and the AUC in the test set is 0.826 (95% CI: 0.681-0.936). For the same test set, the AUC of clinical model constructed based on eGFR is 0.560 (95% CI: 0.357-0.761), and the AUC of US model constructed based on Ultrasound parameters is 0.679 (95% CI: 0.489-0.853). The Xgboost model is significantly more efficient than the clinical and US models.

CONCLUSION

ML model based on ultrasound radiomics features can accurately predict the chronic degree of LN, which can provide a valuable reference for clinicians in the treatment strategy of LN patients.

Development of a novel combined nomogram model integrating deep learning radiomics to diagnose IgA nephropathy clinically

This study aimed to develop and validate a combined nomogram model based on superb microvascular imaging (SMI)-based deep learning (DL), radiomics characteristics, and clinical factors for noninvasive differentiation between immunoglobulin A nephropathy (IgAN) and non-IgAN.We prospectively enrolled patients with chronic kidney disease who underwent renal biopsy from May 2022 to December 2022 and performed an ultrasound and SMI the day before renal biopsy. The selected patients were randomly divided into training and testing cohorts in a 7:3 ratio. We extracted DL and radiometric features from the two-dimensional ultrasound and SMI images. A combined nomograph model was developed by combining the predictive probability of DL with clinical factors using multivariate logistic regression analysis. The proposed model's utility was evaluated using receiver operating characteristics, calibration, and decision curve analysis. In this study, 120 patients with primary glomerular disease were included, including 84 in the training and 36 in the test cohorts. In the testing cohort, the ROC of the radiomics model was 0.816 (95% CI:0.663-0.968), and the ROC of the DL model was 0.844 (95% CI:0.717-0.971). The nomogram model combined with independent clinical risk factors (IgA and hematuria) showed strong discrimination, with an ROC of 0.884 (95% CI:0.773-0.996) in the testing cohort. Decision curve analysis verified the clinical practicability of the combined nomogram. The combined nomogram model based on SMI can accurately and noninvasively distinguish IgAN from non-IgAN and help physicians make clearer patient treatment plans.

Noninvasive Evaluation of Lupus Nephritis Activity Using a Radiomics Machine Learning Model Based on Ultrasound

Introduction

To explore whether ultrasonic radiomics extracted by machine learning method can noninvasively evaluate lupus nephritis (LN) activity.

Materials and Methods

This retrospective study included 149 patients with LN diagnosed by renal biopsy. They were divided into a training cohort (n=104) and a test cohort (n=45). Ultrasonic radiomics features were extracted from the ultrasound images, and the radiomics features were constructed. Furthermore, five machine learning algorithms were compared to evaluate LN activity. The performance of the binary classification model was evaluated by the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

Results

The average AUC of the five machine learning models was 79.4, of which the MLP model was the best. The AUC of the training group was 89.1, with an accuracy of 81.7%, a sensitivity of 83%, a specificity of 80.7%, a negative predictive value of 85.2%, and a positive predictive value of 78%. The AUC of the test group was 82.2, the accuracy was 73.3%, the sensitivity was 78.9%, the specificity was 69.2%, the negative predictive value was 81.8%, and the positive predictive value was 65.2%.

Conclusion

Machine learning classifier based on ultrasonic radiomics has high accuracy for LN activity. The model can be used to noninvasively detect the activity of LN and can be an effective tool to assist the clinical decision-making process.

A novel clinical-radiomic nomogram for the crescent status in IgA nephropathy

OBJECTIVE

We used machine-learning (ML) models based on ultrasound radiomics to construct a nomogram for noninvasive evaluation of the crescent status in immunoglobulin A (IgA) nephropathy.

METHODS

Patients with IgA nephropathy diagnosed by renal biopsy (n=567) were divided into training (n=398) and test cohorts (n=169). Ultrasound radiomic features were extracted from ultrasound images. After selecting the most significant features using univariate analysis and the least absolute shrinkage and selection operator algorithm, three ML algorithms were assessed for final radiomic model establishment. Next, clinical, ultrasound radiomic, and combined clinical-radiomic models were compared for their ability to detect IgA crescents. The diagnostic performance of the three models was evaluated using receiver operating characteristic curve analysis.

RESULTS

The average area under the curve (AUC) of the three ML radiomic models was 0.762. The logistic regression model performed best, with AUC values in the training and test cohorts of 0.838 and 0.81, respectively. Among the final models, the combined model based on clinical characteristics and the Rad score showed good discrimination, with AUC values in the training and test cohorts of 0.883 and 0.862, respectively. The decision curve analysis verified the clinical practicability of the combined nomogram.

CONCLUSION

ML classifier based on ultrasound radiomics has a potential value for noninvasive diagnosis of IgA nephropathy with or without crescents. The nomogram constructed by combining ultrasound radiomic and clinical features can provide clinicians with more comprehensive and personalized image information, which is of great significance for selecting treatment strategies.

Artificial intelligence in glomerular diseases

In this narrative review, we focus on the application of artificial intelligence in the clinical history of patients with glomerular disease, digital pathology in kidney biopsy, renal ultrasonography imaging, and prediction of chronic kidney disease (CKD). With the development of natural language processing, the clinical history of a patient can be used to identify a computable phenotype. In kidney pathology, digital imaging has adopted innovative deep learning algorithms (DLAs) that can improve the predictive capability of the examined lesions. However, at this time, these applications can only be used in research because there is no recognized validation to replace the conventional diagnostic applications. Kidney ultrasonography, used in the clinical examination of patients, provides information about the progression of kidney damage. Machine learning algorithms (MLAs) with promising results for the early detection of CKD have been proposed, but, still, they are not solid enough to be incorporated into the clinical practice. A few tools for glomerulonephritis, based on MLAs, are available in clinical practice. They can be downloaded on computers and cellular phones but can only be applied to uniracial cohorts of patients. To improve their performance, it is necessary to organize large consortia with multiracial cohorts. Finally, in many studies MLA development has been carried out using retrospective cohorts. The performance of the models might differ in retrospective cohorts compared to real-world data. Therefore, the models should be validated in prospective external large cohorts.

MRI-based kidney radiomic analysis during chronic lithium treatment

BACKGROUND

Lithium therapy during bipolar disorder is associated with an increased risk of chronic kidney disease (CKD) that is slowly progressive and undetectable at early stages. We aimed at identifying kidney image texture features as possible imaging biomarkers of decreased measured glomerular filtration rate (mGFR) using radiomics of T2-weighted magnetic resonance imaging (MRI).

METHODS

One hundred and eight patients treated with lithium were evaluated including mGFR and kidney MRI, with T2-weighted sequence single-shot fast spin-echo. Computed radiomic analysis was performed after kidney segmentation. Significant features were selected to build a radiomic signature using multivariable Cox analysis to detect an mGFR <60 ml/min/1.73 m². The texture index was validated using a training and a validation cohort.

RESULTS

Texture analysis index was able to detect an mGFR decrease, with an AUC of 0.85 in the training cohort and 0.71 in the validation cohort. Patients with a texture index below the median were older (59 [42-66] vs. 46 [34-54] years, p = .001), with longer treatment duration (10 [3-22] vs. 6 [2-10] years, p = .02) and a lower mGFR (66 [46-84] vs. 83 [71-94] ml/min/1.73m², p < .001). Texture analysis index was independently and negatively associated with age (β = -.004 ± 0.001, p < .001), serum vasopressin (-0.005 ± 0.002, p = .02) and lithium treatment duration (-0.01 ± 0.003, p = .001), with a significant interaction between lithium treatment duration and mGFR (p = .02).

CONCLUSIONS

A renal texture index was developed among patients treated with lithium associated with a decreased mGFR. This index might be relevant in the diagnosis of lithium-induced renal toxicity.