A quantitative characterization of tuberculin skin test indurations using hyperspectral imaging to enable automated latent tuberculosis screening
| dc.contributor.advisor | Malila, Bessie | |
| dc.contributor.advisor | Mutsvangwa, Tinashe | |
| dc.contributor.author | Oladokun, Ajibola Samson | |
| dc.date.accessioned | 2026-01-20T12:55:16Z | |
| dc.date.available | 2026-01-20T12:55:16Z | |
| dc.date.issued | 2025 | |
| dc.date.updated | 2026-01-20T12:52:45Z | |
| dc.description.abstract | An estimated two billion people globally are believed to harbour latent tuberculosis infection (LTBI), which is a precursor to active tuberculosis (TB). Immunocompromising conditions, such as human immunodeficiency virus (HIV) infection, diabetes, chronic renal failure, and chemotherapy, are associated with the risk of progression from LTBI to active TB. People with LTBI represent a major reservoir for new active TB cases. Thus, the detection and management of LTBI is key to the world health organisation (WHO) goal to reduce TB incidence globally by 90% by the year 2035. The tuberculin skin test (TST) is the most widely used test for LTBI in low- and middle-income countries (LMICs). It involves the injection of tuberculin into the forearm skin of a participant. An induration – a subdermal lump – forms at the injection site 48 – 72 hours after the test. Using the Mantoux reading method, diagnosis is obtained using a pen and a ruler to measure the induration diameter and comparing it with consensus thresholds. Mantoux readings are subject to inter- and intra-observer subjective variations in readings. The manual nature of this method precludes any repeat analysis or validation after the 72-hour expiry of an induration. Furthermore, the Mantoux method does not provide any significant insight on the pathophysiology of indurations which could improve current understanding of LTBI. Recent applications of hyperspectral imaging (HSI) in other dermatological applications provides evidence of its viability in addressing the shortcomings of the TST. Thus, the aim of this research was to develop and validate an HSI-based approach for quantitative characterisation of TST indurations that can aid the diagnosis of LTBI. This aim was achieved through three objectives. The first objective was to develop and validate an HSI protocol for reliable and reproducible acquisition of spectral signatures of TST injection sites. An enclosure was designed to implement and validate the developed protocol. The validation was done using a range of light source types and ten participants with a wide range of skin tones. The effect of light source type on skin spectral signature was investigated to reveal factors that impact spectral reproducibility. The validated imaging protocol was utilized in the second objective to reliably capture the hyperspectral images of 70 participants (38 from South Africa and 32 from Vietnam). The objective was to develop and validate an HSI-model that can generate precise diameter estimations of TST indurations that are comparable to traditional Mantoux readings. Principal component analysis was utilized to transform the hyperspectral images (or hypercubes) into principal components and generate induration segmentation masks. Precise induration diameter estimations, which were comparable to corresponding Mantoux readings, were generated from the masks. An intraclass correlation coefficient score of 0.80, a median difference in reading of 0.29 mm, and a Pearson's correlation of 0.664 (p<0.001), was achieved between the HSI-based diameter estimations and corresponding Mantoux readings. This metrics surpassed the performance of similar studies. This is evidence that HSI can mitigate the subjectivity of the Mantoux method by generating comparable but more precise diameter estimations of TST indurations. The induration hypercubes and segmentation masks were utilized in the third objective to develop and validate an HSI framework capable of identifying biomarkers that accurately and consistently characterise TST indurations. The framework enabled the generation of chromophore maps from induration hypercubes. Radiomics was applied on the chromophore maps to generate a set of features. The features were used, in conjunction with a support vector machine model, to rank the ability of the generated chromophore maps to predict LTBI. Ferritin, water, and oxyhaemoglobin maps were identified as the most predictive chromophore maps for LTBI. The predictive accuracies for TST diagnosis using these chromophores were 90.0% across all participants, and 96.0% and 96.7% for the South African and Vietnamese participants. This is evidence that HSI can noninvasively facilitate the generation of chromophore biomarkers to characters TST indurations – a first of its kind in literature. These findings present HSI as a potential modality for mitigating the subjectivity of the Mantoux method and identifying digital biomarkers that noninvasively offer insights into induration pathophysiology. The potential impact of this research is improved health outcome for at-risk TBgroups via enhanced LTBI screening accuracy and precision. | |
| dc.identifier.apacitation | Oladokun, A. S. (2025). <i>A quantitative characterization of tuberculin skin test indurations using hyperspectral imaging to enable automated latent tuberculosis screening</i>. (). University of Cape Town ,Faculty of Health Sciences ,Department of Human Biology. Retrieved from http://hdl.handle.net/11427/42630 | en_ZA |
| dc.identifier.chicagocitation | Oladokun, Ajibola Samson. <i>"A quantitative characterization of tuberculin skin test indurations using hyperspectral imaging to enable automated latent tuberculosis screening."</i> ., University of Cape Town ,Faculty of Health Sciences ,Department of Human Biology, 2025. http://hdl.handle.net/11427/42630 | en_ZA |
| dc.identifier.citation | Oladokun, A.S. 2025. A quantitative characterization of tuberculin skin test indurations using hyperspectral imaging to enable automated latent tuberculosis screening. . University of Cape Town ,Faculty of Health Sciences ,Department of Human Biology. http://hdl.handle.net/11427/42630 | en_ZA |
| dc.identifier.ris | TY - Thesis / Dissertation AU - Oladokun, Ajibola Samson AB - An estimated two billion people globally are believed to harbour latent tuberculosis infection (LTBI), which is a precursor to active tuberculosis (TB). Immunocompromising conditions, such as human immunodeficiency virus (HIV) infection, diabetes, chronic renal failure, and chemotherapy, are associated with the risk of progression from LTBI to active TB. People with LTBI represent a major reservoir for new active TB cases. Thus, the detection and management of LTBI is key to the world health organisation (WHO) goal to reduce TB incidence globally by 90% by the year 2035. The tuberculin skin test (TST) is the most widely used test for LTBI in low- and middle-income countries (LMICs). It involves the injection of tuberculin into the forearm skin of a participant. An induration – a subdermal lump – forms at the injection site 48 – 72 hours after the test. Using the Mantoux reading method, diagnosis is obtained using a pen and a ruler to measure the induration diameter and comparing it with consensus thresholds. Mantoux readings are subject to inter- and intra-observer subjective variations in readings. The manual nature of this method precludes any repeat analysis or validation after the 72-hour expiry of an induration. Furthermore, the Mantoux method does not provide any significant insight on the pathophysiology of indurations which could improve current understanding of LTBI. Recent applications of hyperspectral imaging (HSI) in other dermatological applications provides evidence of its viability in addressing the shortcomings of the TST. Thus, the aim of this research was to develop and validate an HSI-based approach for quantitative characterisation of TST indurations that can aid the diagnosis of LTBI. This aim was achieved through three objectives. The first objective was to develop and validate an HSI protocol for reliable and reproducible acquisition of spectral signatures of TST injection sites. An enclosure was designed to implement and validate the developed protocol. The validation was done using a range of light source types and ten participants with a wide range of skin tones. The effect of light source type on skin spectral signature was investigated to reveal factors that impact spectral reproducibility. The validated imaging protocol was utilized in the second objective to reliably capture the hyperspectral images of 70 participants (38 from South Africa and 32 from Vietnam). The objective was to develop and validate an HSI-model that can generate precise diameter estimations of TST indurations that are comparable to traditional Mantoux readings. Principal component analysis was utilized to transform the hyperspectral images (or hypercubes) into principal components and generate induration segmentation masks. Precise induration diameter estimations, which were comparable to corresponding Mantoux readings, were generated from the masks. An intraclass correlation coefficient score of 0.80, a median difference in reading of 0.29 mm, and a Pearson's correlation of 0.664 (p<0.001), was achieved between the HSI-based diameter estimations and corresponding Mantoux readings. This metrics surpassed the performance of similar studies. This is evidence that HSI can mitigate the subjectivity of the Mantoux method by generating comparable but more precise diameter estimations of TST indurations. The induration hypercubes and segmentation masks were utilized in the third objective to develop and validate an HSI framework capable of identifying biomarkers that accurately and consistently characterise TST indurations. The framework enabled the generation of chromophore maps from induration hypercubes. Radiomics was applied on the chromophore maps to generate a set of features. The features were used, in conjunction with a support vector machine model, to rank the ability of the generated chromophore maps to predict LTBI. Ferritin, water, and oxyhaemoglobin maps were identified as the most predictive chromophore maps for LTBI. The predictive accuracies for TST diagnosis using these chromophores were 90.0% across all participants, and 96.0% and 96.7% for the South African and Vietnamese participants. This is evidence that HSI can noninvasively facilitate the generation of chromophore biomarkers to characters TST indurations – a first of its kind in literature. These findings present HSI as a potential modality for mitigating the subjectivity of the Mantoux method and identifying digital biomarkers that noninvasively offer insights into induration pathophysiology. The potential impact of this research is improved health outcome for at-risk TBgroups via enhanced LTBI screening accuracy and precision. DA - 2025 DB - OpenUCT DP - University of Cape Town KW - Skin KW - Tuberculosis LK - https://open.uct.ac.za PB - University of Cape Town PY - 2025 T1 - A quantitative characterization of tuberculin skin test indurations using hyperspectral imaging to enable automated latent tuberculosis screening TI - A quantitative characterization of tuberculin skin test indurations using hyperspectral imaging to enable automated latent tuberculosis screening UR - http://hdl.handle.net/11427/42630 ER - | en_ZA |
| dc.identifier.uri | http://hdl.handle.net/11427/42630 | |
| dc.identifier.vancouvercitation | Oladokun AS. A quantitative characterization of tuberculin skin test indurations using hyperspectral imaging to enable automated latent tuberculosis screening. []. University of Cape Town ,Faculty of Health Sciences ,Department of Human Biology, 2025 [cited yyyy month dd]. Available from: http://hdl.handle.net/11427/42630 | en_ZA |
| dc.language.iso | en | |
| dc.language.rfc3066 | eng | |
| dc.publisher.department | Department of Human Biology | |
| dc.publisher.faculty | Faculty of Health Sciences | |
| dc.publisher.institution | University of Cape Town | |
| dc.subject | Skin | |
| dc.subject | Tuberculosis | |
| dc.title | A quantitative characterization of tuberculin skin test indurations using hyperspectral imaging to enable automated latent tuberculosis screening | |
| dc.type | Thesis / Dissertation | |
| dc.type.qualificationlevel | Doctoral | |
| dc.type.qualificationlevel | PhD |