Classify AI
Classify AI is a software package designed to train, optimise and evaluate machine learning classifiers for automated reading of HIV Rapid Diagnostic Tests (RDTs). It compares the performance of 4 different approaches (MobileNet_v2, MobileNet_v3, ResNet50 and SVM), using a dataset of images captured in the field in rural South Africa.
Product Specification
Classify AI is a series of interdependent scripts written in Python 3.6.
Description
This repository contains the code used to produce the results presented in the following publication:
Deep learning of HIV field-based rapid tests DOI:10.1038/s41591-021-01384-9
The work presented is the result of a collaboration between i-sense (https://www.i-sense.org.uk/) and the Africa Health Research Institute (AHRI, https://www.ahri.org/).
The datasets are available upon request, please see details in the publication.
The repository is divided in 2 subfolders corresponding to Figures 3 and 4 of the publication, respectively.
In Figure 3, the others used a dataset of 11, 374 images of HIV RDTs captured in the field, around AHRI. They trained and optimised classifers, performed cross-validation to check for reproducibility. They produced 2 optimal classifers, one for each type of HIV RDT used in the study.
In Figure 4, the authors conducted a pilot study to assess 1) the level of agreement between participants when intepreting HIV RDTs by eye; 2) the performance of their optimal classifers (one for each type of HIV RDT used in the study) compared to traditional visual interpretation by humans.
-
swap_vertical_circlemode_editAuthors (1)Prof Rachel McKendry
-
swap_vertical_circlelibrary_booksReferences (1)
- V Turbé, C Herbst, T Mngomezulu, S Meshkinfamfard, N Dlamini, T Mhlongo, T Smit, V Cherepanova, K Shimada, J Budd, N Arsenov, S Gray, D Pillay, K Herbst, M Shahmanesh & R A. McKendry (17 June 2021), Deep learning of HIV field-based rapid tests, Nat Med
-
swap_vertical_circlecloud_downloadSupporting documents (0)Additional files may be available once you've completed the transaction for this product. If you've already done so, please log into your account and visit My account / Downloads section to view them.