Table 1.
Characteristics and diagnostic performance of included studies on AI-assisted tools for Schistosoma haematobium detection.
Fig 1.
PRISMA flow diagram of the study selection process.
The flowchart outlines the systematic identification, screening, and inclusion of studies investigating the accuracy of AI-assisted diagnostic tools for Schistosoma haematobium.
Fig 2.
Summary of risk of bias and applicability concerns of included studies.
Methodological quality was assessed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool. (A) Risk of bias and applicability concerns individual studies across the four QUADAS-2 domains: patient selection, index test, reference standard, and flow and timing. (B) Summary plot showing the proportion of studies judged as low, high, or unclear risk of bias (left panel) and applicability concerns (right panel) for each domain.
Fig 3.
Forest plots of pooled sensitivity and specificity for AI-assisted diagnostic tools in detecting Schistosoma haematobium.
The forest plots illustrate the diagnostic performance of AI-assisted tools across 15 included datasets. The left panel represents the Sensitivity 0.88 (95% CI: 0.83-0.91), and the right panel represents the Specificity 0.89 (95% CI: 0.83-0.93).
Fig 4.
Forest plots for pooled positive and negative Diagnostic Likelihood Ratios (DLRs).
The forest plots display the diagnostic performance of AI-assisted tools for Schistosoma haematobium across 15 datasets. The left panel shows the pooled Positive Diagnostic Likelihood Ratio (DLR+) of 7.61 (95% CI: 4.98-11.64), indicating a moderate-to-strong increase in the likelihood of infection given a positive AI result. The right panel shows the pooled Negative Diagnostic Likelihood Ratio (DLR-) of 0.14 (95%CI: 0.10-0.20), suggesting that a negative AI result significantly reduces the probability of infection.
Fig 5.
Forest plots for pooled diagnostic score and diagnostic odds ratio (DOR).
The forest plots illustrate the overall diagnostic efficacy of AI-assisted tools for Schistosoma haematobium detection across 15 datasets. The left panel shows a pooled Diagnostic Score of 3.99 (95%CI: 3.41-4.56), indicating a high level of discrimination between infected and non-infected individuals. The right panel displays the pooled Diagnostic Odds Ratio (DOR) of 54.00 (95%CI: 30.41-95.88), which represents the ratio of the odds of a positive AI result in infected individuals compared to non-infected individuals.
Fig 6.
Summary receiver operating characteristic (SROC) curve for AI-assisted diagnostic tools.
The SROC curve illustrates the overall diagnostic performance of AI-assisted platforms in detecting Schistosoma haematobium across 15 datasets. The Summary Operating Point (represented by the grey diamond) indicates a pooled sensitivity of 0.88 (95% CI: 0.83-0.91) and a pooled specificity of 0.89 (95% CI: 0.83-0.93). The Area Under the Curve (AUC) is 0.94 (95%CI: 0.92-0.96).
Fig 7.
Fagan nomogram for the clinical utility of AI-assisted diagnostic tools.
The Fagan nomogram displays the relationship between the pre-test probability, likelihood ratios, and post-test probability for Schistosoma haematobium detection. Assuming a baseline pre-test probability (prevalence) of 20%, a positive AI result (LR-Positive = 8) increases the post-test probability of infection to 66%. Conversely, a negative AI result (LR-Negative = 0.14) reduces the post-test probability of infection to 3%.
Table 2.
Subgroup analysis of pooled sensitivity and specificity of AI-assisted diagnostic tools for schistosomiasis, subgroup by publication year, endemicity, AI model type, and country setting.
Fig 8.
Univariable meta-regression and subgroup analyses evaluating the effect of study year, country setting, and AI model type on pooled sensitivity and specificity of AI-assisted diagnosis of schistosomiasis.
Each point represents the pooled diagnostic performance estimate within the corresponding subgroup, with horizontal lines indicating the 95% confidence intervals (CI). The left panel presents subgroup effects on pooled sensitivity, while the right panel shows subgroup effects on pooled specificity. Subgroups include publication year category, country setting, and type of AI model.
Fig 9.
Deeks’ funnel plot asymmetrical test assessing small-study effects in the meta-analysis of AI-assisted schistosomiasis diagnostics.
The figure presents Deeks’ funnel plot of diagnostic log odds ratios against the inverse root of the effective sample size. The regression line and its 95% confidence limits are shown. Asymmetry in the funnel plot indicates potential small-study effects or publication bias. The statistical significance of asymmetry was evaluated using Deeks’ regression test, with a P-value <0.10 indicating evidence of small-study effects.