From immunology to artificial intelligence: revolutionizing latent tuberculosis infection diagnosis with machine learning

Table 1 Current TST methods used for LTBI diagnosis

Characteristics	PPD	C-TB	Diaskintest	EC-test
Time	> 100 years	2009	2010	2020
Type of reaction	DTH	DTH	DTH	DTH
Number of visits	2	2	2	2
Type of antigen	PPD	ESAT-6 and CFP-10	ESAT-6 and CFP-10	ESAT-6 and CFP-10
Outcome measures	Millimeters of induration	Millimeters of induration	Millimeters of induration	Millimeters of induration
Sensitivity	77% [190] 84% [191]	73.9% [189] 74.52% [192]	86% [193] 68% [193] 91.18 [192]	90.85% [194] 86.06% [192]
Specificity	97% (without BCG vaccination) and 59% (with BCG vaccination) [190] 100% (without BCG vaccination) and 79% (with BCG vaccination) [191]	97.85% [192]	98% [193]	89.83% [194]
Interpretation	Subjective	Subjective	Subjective	Subjective
False positive rate in immunosuppressed or BCG vaccinated population	High	Low	Low	Low
Distinguish between LTBI and ATB	No	No	No	No

ATB active tuberculosis, BCG Bacillus Calmette–Guérin, CFP-10 culture filtrate protein-10, DTH delayed type hypersensitivity, ESAT-6 early secreted antigen target-6, LTBI latent tuberculosis infection, TST tuberculin skin test, PPD purified protein derivative, C-TB a novel skin test based on ESAT-6 and CFP-10 proteins, EC recombinant Mycobacterium tuberculosis fusion protein of ESAT-6 and CFP-10

ISSN: 2054-9369