Smartphone-based Detection of Group A Streptococcal Pharyngitis in Ugandan Children: A Pilot Study

Figures

• 

  FIGURE 1.. Illustration of the Light AI model for diagnosis of group A Streptococcus pharyngitis on a smartphone that is connected to the server, providing noninvasive, point-of-care diagnosis almost instantly. The smartphones used in the current study were not connected to the server. They simply took short videos that were transferred to Light AI. The Cape Town clinical decision rule was applied in the field for clinical decision-making regarding the need for antibiotics in every study participant. AI indicates artificial intelligence.

• 

  FIGURE 2.. Study flow diagram. The parent study, RESET (under review elsewhere), was conducted in Eastern Uganda, Tororo District, with an estimated total population of 636,629 people, including ~115,000 children 5–15 years of age. *Program interventions were implemented within the public primary healthcare system, focused on 3 of the more easily rectifiable impediments to RHD primary prevention, namely, limited expertise of healthcare providers in diagnosing and treating GAS pharyngitis, poor uptake of guideline-based treatments for GAS pharyngitis, and low level of public awareness of the importance of sore throat and its link to RHD. **These additional 187 culture/image pairs were independent of the Ugandan samples, but images were taken using similar devices. ¶These 250 culture/image pairs included 158 culture-negative and 92 culture-positive samples. †These 82 Ugandan samples included 70 negative and 12 positive cultures, tested with technicians inputting the images blinded to culture results. These cases were all independent of those used in training and validation. AI indicates artificial intelligence; CNN, convolutional neural network; ROC, receiver operating curve.