| 1. |
- Petrov, Dmitry, et al.
(författare)
-
Machine Learning for Large-Scale Quality Control of 3D Shape Models in Neuroimaging
- 2017
-
Ingår i: Machine learning in medical imaging. MLMI (Workshop). - Cham : Springer International Publishing. ; 10541, s. 371-378
-
Tidskriftsartikel (refereegranskat)abstract
- As very large studies of complex neuroimaging phenotypes become more common, human quality assessment of MRI-derived data remains one of the last major bottlenecks. Few attempts have so far been made to address this issue with machine learning. In this work, we optimize predictive models of quality for meshes representing deep brain structure shapes. We use standard vertex-wise and global shape features computed homologously across 19 cohorts and over 7500 human-rated subjects, training kernelized Support Vector Machine and Gradient Boosted Decision Trees classifiers to detect meshes of failing quality. Our models generalize across datasets and diseases, reducing human workload by 30-70%, or equivalently hundreds of human rater hours for datasets of comparable size, with recall rates approaching inter-rater reliability.
|
|
| 3. |
- Harding-Esch, Emma M., et al.
(författare)
-
Impact of a single round of mass drug administration with azithromycin on active trachoma and ocular Chlamydia trachomatis prevalence and circulating strains in The Gambia and Senegal
- 2019
-
Ingår i: Parasites & Vectors. - : BMC. - 1756-3305. ; 12
-
Tidskriftsartikel (refereegranskat)abstract
- Background: Mass drug administration (MDA) with azithromycin is a cornerstone of the trachoma elimination strategy. Although the global prevalence of active trachoma has declined considerably, prevalence persists or even increases in some communities and districts. To increase understanding of MDA impact, we investigated the prevalence of active trachoma and ocular C. trachomatis prevalence, organism load, and circulating strains at baseline and one-year post-MDA in The Gambia and Senegal.Methods: Pre- and one-year post-MDA, children aged 0-9 years were examined for clinical signs of trachoma in six Gambian and 12 Senegalese villages. Ocular swabs from each child's right conjunctiva were tested for evidence of ocular C. trachomatis infection and organism load (ompA copy number), and ompA and multi-locus sequence typing (MLST) was performed.Results: A total of 1171 children were examined at baseline and follow-up in The Gambia. Active trachoma prevalence decreased from 23.9% to 17.7%, whereas ocular C. trachomatis prevalence increased from 3.0% to 3.8%. In Senegal, 1613 and 1771 children were examined at baseline and follow-up, respectively. Active trachoma prevalence decreased from 14.9% to 8.0%, whereas ocular C. trachomatis prevalence increased from 1.8% to 3.6%. Higher organism load was associated with having active trachoma and severe inflammation. Sequence typing demonstrated that all Senegalese samples were genovar A, whereas Gambian samples were a mix of genovars A and B. MLST provided evidence of clustering at village and household levels and demonstrated differences of strain variant frequencies in Senegal, indicative of an "outbreak". MLST, including partial ompA typing, provided greater discriminatory power than complete ompA typing.Conclusions: We found that one round of MDA led to an overall decline in active trachoma prevalence but no impact on ocular C. trachomatis infection, with heterogeneity observed between villages studied. This could not be explained by MDA coverage or number of different circulating strains pre- and post-MDA. The poor correlation between active trachoma and infection prevalence supports the need for further work on alternative indicators to clinical signs for diagnosing ocular C. trachomatis infection. MLST typing has potential molecular epidemiology utility, including better understanding of transmission dynamics, although relationship to whole-genome sequence variability requires further exploration.
|
|
