VALIDITY ANALYSIS OF COMPUTER-AIDED DIAGNOSIS (CAD4TB) VS. GENEXPERT FOR TUBERCULOSIS DETECTION IN VERBAL SCREENED PRESUMPTIVE PULMONARY TB PATIENTS IN QUETTA, BALUCHISTAN

Zafar Iqbal Khosti; Dr Samia Ahmed Khan; Shagufta Naseer; Sher Afghan; Syeda Farhat Bukhari; Mohsin Jamil Siddiqui

Authors

Zafar Iqbal Khosti Author
Dr Samia Ahmed Khan Author
Shagufta Naseer Author
Sher Afghan Author
Syeda Farhat Bukhari Author
Mohsin Jamil Siddiqui Author

Keywords:

Tuberculosis, Computer Aided Diagnosis, GeneXpert

Abstract

Background: Choosing diagnostic tools for serious diseases like tuberculosis requires careful consideration of parameters like sensitivity, specificity, and predictive values. Currently, both Computer-Aided Diagnosis for TB (CAD4TB) and GeneXpert are used in tandem for tuberculosis detection, but their relative effectiveness has not been conclusively established in the literature.

Objective: This study aims to evaluate the validity of CAD4TB by comparing it against GeneXpert, considered the gold standard.

Methods: A cross-sectional validation analysis was conducted using secondary data obtained from the Provincial TB Control Program, sourced from public sector healthcare institutes in Quetta, where both CAD4TB and GeneXpert facilities were available. Validity was calculated, yielding sensitivity (89.4%), specificity (10.2%), positive predictive value (14.7%), and negative predictive value (84.9%). Significant associations were further evaluated for trend association with the Moses Extreme Reactions test. Sensitivity and specificity were established, and Receiver Operating Curve analysis confirmed validity. Statistical significance was set at p < 0.05, and SPSS v 25.0 was used for data processing and analysis.

Results: Out of 573 presumptive patients, 514 were included in the final analysis (response rate: 89.7%). The majority were over 54 years old (260, 50.5%), and the cohort was predominantly male (282, 54.8%). The mean CAD4TB score was 80.15 ± 12.93, while GeneXpert detected 84.8% of cases. Significant associations were found between CAD4TB and Xpert scores, with moderately strong relationship interpreted. Moses Extreme Reaction test confirmed a significant relationship favoring GeneXpert. CAD4TB reported an area under the curve value of 0.497, indicating poor discrimination. This reinforced the preference for GeneXpert over CAD4TB.

Conclusion: This study demonstrates that GeneXpert provides a more sensitive and specific categorization of patients compared to CAD4TB.