1 / 12

Addressing Antimicrobial Resistance Using AI in Healthcare

This presentation discusses the use of artificial intelligence to combat antimicrobial resistance within TG-Bacteria. It covers the challenges posed by AMR, the development of the Antibiogo app to support accurate antibiotic prescribing, and the application of machine learning in identifying complex resistance mechanisms. The presentation highlights the implementation of AI in healthcare settings and the potential to improve patient outcomes by enhancing antibiotic stewardship practices.

quintial
Télécharger la présentation

Addressing Antimicrobial Resistance Using AI in Healthcare

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. FGAI4H-L-008-A03 E-meeting, 19-21 May 2021 Source: TG-Bacteria Topic Driver Title: Att.3 - Presentation (TG-Bacteria) Purpose: Discussion Contact: Nada MALOU Clinical lead Fondation MSF E-mail: nada.malou@paris.msf.org Contact: Louis LAROCHE Product manager Fondation MSF E-mail: louis.laroche@paris.msf.org Abstract: This PPT summarizes the status of work within TG-Bacteria, for presentation and discussion during the meeting.

  2. Antibiogo Using AI to tackle antimicrobial resistance AI for Health - May 2021

  3. Hello AI for Health Nada MALOU Clinical lead 1. Fondation MSF intro Fondation MSF nada.malou@paris.msf.org 2. Antimicrobial resistance 3. Product value Louis LAROCHE Product manager 4. Machine learning specifics Fondation MSF louis.laroche@paris.msf.org

  4. Fondation Médecins Sans Frontières Improving MSF’s performance in the field through 3 main focus areas: Humanitarian knowledge Applied medical research T ech innovation: ● ● ● 3D PROGRAM COVID19 ANTIBIOGO 3D printing prosthesis/orthosis for burn and amputee patients Web app to support countries ripost to COVID19 pandemic A diagnostic tool to provide access to accurate antibiotics GAZA - HAITI NIGER - DRC JORDAN - MALI - SENEGAL

  5. Deaths attributable toAMR every year A global health issue Antibiotic or antimicrobial resistance (AMR) is defined as the resistance of bacterial to antibiotics that were previously effective for treatment of infections. AMR drivers by WHO: 1. 2. 3. Non rational use of antibiotics Lack of hygiene and infection control Non access to diagnostic tools Review onAMR's report (O'Neill,2016)

  6. How does a microbiology lab work? Incubation time Patient sample Lab tech sets up Doctor Patient Lab tech reads Microbio expert Prescription AST

  7. Antibiotic Susceptibility Testing (AST) Measurement: Short training Easy to achieve Low error margin ● ● ● Interpretation: 3K rules to know,updated yearly Requires highly qualified staff Alternative:SIRscan ● ● ●

  8. Value proposition Antibiogo is a free, open source and offlineAndroid app that supports non-expert laboratory technicians measuring and interpreting antibiotic susceptibilitytests (AST), to help doctors prescribe accurate antibiotics to their patients.

  9. Product demo Measure easily Complete interpretation Read results & take action ShareASTresults ● ● ● ● View on Youtube

  10. Reading antibiotic pellets on a Petri dish using machine learning There is a limited set of ~50 possible labels We used a Convolutional Neural Network (CNN) classifier to categorize labels We use an ensemble of 10 training models We embed the model in the app using Tensorflow liteto be available offline ● ● ● ● Results:99.97% test accuracy Limitation: there are multiple pellet brands, and model is only trained on two (i2a, liofilchem). We are currently struggling to train it on 2 other major brands. ● ●

  11. Complex resistance mechanisms identification postmortem Trained ML models that recognize: ○ D-Zone: 99.7% accuracy ○ Synergy: 98% accuracy ● Synergy > ESBL We decided not to use the model, because: ○ Westill need to ask confirmation to the user ○ Suggesting a result inducts bias in user confirmation ○ Identifying mechanisms technicians ● is straightforward for lab Low algorithm transferability ○ D-Zone > MRSA

  12. Antibiogo Thank you =) Any question?

More Related