The initial evaluation of TBDx TM, “Proof-of-Concept” Evaluation of an Automated Sputum Smear Microscopy System for Tuberculosis Diagnosis, undertaken by The Aurum Institute, was published in the Public Library of Science (PLoS) One on November 29, 2012. PLoS One is a highly regarded and open source of scientific research in the TB community.
This peer-reviewed evaluation represents the state of the TBDxTM technology development as it existed in May, 2011. Several important observations are discussed in the publication:
- This study is the first large-scale evaluation of an automated microscopy system for tuberculosis in the published literature.
- Unlike several previously published studies of automated smear microscopy, the two strengths of this study can be found in:
- its use of culture as the ‘gold standard’, rather than microscopist-generated results,
- its classification of the patient smear as positive or negative rather than simply classifying individual objects as ”true Acid-Fast Bacillus (AFB) or not AFB,” which can lead to bias and patient case misclassifications.
- Automated sputum microscopy has the potential to substantially reduce the number of slides read by a microscopist.
- Although reaching a sensitivity of 75% (conventional microscopy ranges from 20%-60%), which represents an important advancement in TB disease detection, using it as a diagnostic substitute for a microscopist will require improvements in reducing false positives.
- Additional studies of TBDx TM deployed in field settings are warranted.
Improvements in the detection algorithms and in our operational software have been ongoing since this evaluation. Those improvements have generated a significant increase in diagnostic performance as discussed in the poster presented by Aurum at the International Union Conference held in November 2012.
As noted in our blog of May 10, 2011, Dr. Fleming Lure presented a paper Application of Stepwise Binary Decision Classification For Reduction of False Positives in Tuberculosis Detection from Smeared Slides at the 2011 Artificial Intelligence and Applications Conference. Dr. Lure presented the methods and processes he and colleagues were employing in our development process to better classify objects segmented in TB images.
On January 31, 2012 our blog noted a second paper accepted for presentation during the February 4-9, 2012 International Society for Optics and Photonics. The authors, including Ajay Divekar and Fleming Lure, detailed their further refinements in their classification methods in a paper entitled Automated Detection of Tuberculosis On Sputum Smeared Slides Using Stepwise Classification.
The initial presentation of the study data appeared in our March 8, 2012 and March 15, 2012 blogs. At that time we mentioned that Gavin Churchyard, CEO of The Aurum Institute, presented a paper Automated AFB Microscopy Substantially Reduces Microscopists Work Load and participated in a panel presentation at the Conference on Retrovirals and Opportunistic Infections (CROI). The conference posted a video of his remarks and the panel discussions. In the blogs we acknowledged that no “silver bullet” technology existed today that could diagnosis TB by itself and that combinations of technologies such as TBDx TM and GeneXpert TM together could produce a cost effective layered diagnostic protocol.
Since then we have continued to make improvements in the technology and have continually measured our progress through internal testing. That was reported in our April 9, 2012 blog when we presented the results of testing 181 cases from the National Health Laboratory Services and modeled how the performance of TBDx TM could screen cases and send them to GeneXpert TM for confirmation and drug susceptibility. Economic modeling of this layered diagnostic approach clearly indicates that more positive cases of TB are detected over current diagnostic methods, in less time and with a lower cost per-positive-case detected.
The improvements in the technology were recently presented by Dr. Dave Clark during the 43rd Union World Conference on Lung Health. Entitled Automated TB Microscopy – Recent results and a model to increase pre-test probability to gene-based diagnostics Dr. Clark presented the results of our internal testing and presented a cost / benefit model for combining TBDx TM and GeneXpert TM.
We are delighted to see the results of our initial TBDx TM evaluation published in the PLoS One open access website. It is a well-regarded source on the latest TB research. No doubt it will be seen by TB professionals and will generate interest and questions. This article presents the initial steps towards the evidence-based research necessary to fully evaluate our TBDx TM technology. It is a vital step towards deployment.
Since this study was completed, we have continued to develop our technology as we prepare for the next evaluation this upcoming January. At that time, the evaluation not only will compare the performance of TBDx TM with conventional microscopy, using culture as the reference standard, but also will model how TBDx TM can act as a screening technology for GeneXpert TM. The expected outcome is that the study will support our previous study indicating improved TB detection, in less time and at a lower cost per positive case detection.