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

Figures & data

Figure 1 The flow diagram of the study.

Figure 2 The radiomics flow chart of the study.

Table 1 The Baseline Characteristics of the Patients

Characteristic	Value
Age (years)	37.2±12.4
Sex (male/female)	32/117
Average pressure (mmHg)	102.5±14.2
Creatinine level at biopsy (μmol/L)	100.8±49.1
eGFR at biopsy (mL/min/1.73m^2)	86.1±30.7
Proteinuria at biopsy (g/24h)	1.49±1.52
Activity index
>7	n=66
≤7	n=83
Chronic index
>3	n=34
≤3	n=115

Figure 3 Ten features were used to establish the radiomic features.

Table 2 Performance of the Five Model in the Training and Testing Cohorts

	Training					Testing
	MLP	LR	SVM	RF	XGB	MLP	LR	SVM	RF	XGB
AUC	0.89	0.93	0.90	0.95	0.95	0.82	0.81	0.79	0.78	0.77
ACC	81.7	86.5	84.6	89.4	87.5	73.3	75.6	75.6	62.2	60
SEN	83	87.2	74.5	95.7	84.2	78.9	73.7	57.9	73.7	78.9
SPE	80.7	86	93	84.2	77.2	69.2	76.9	88.5	53.8	46.2
PPV	78	83.7	89.7	83.3	78.3	65.2	70	78.6	53.8	51.7
NPV	85.2	89.1	81.5	96	100	81.8	80	74.2	73.7	75

Abbreviations: AUC, area under the curve; ACC, accuracy; SEN, sensitivity; SPE, specificity; PPV, positive likelihood ratio; NPV, negative likelihood ratio.

Figure 4 The receiver operating characteristic (ROC) curves of the five models. (A) five model ROC curves in the training cohort. (B) five model ROC curves in the testing cohort.

Figure 5 The confusion matrix of MLP model in the training (A) and testing cohorts (B).