~
Newsletter 3rd quarter 2021
Team News
International
Project Presentations
- PASSION Symposium Preparation at the International Society of
Teledermatology Conference. The program will be as follows:
- Presentation of the PASSION Project: Invited speaker - Myself
- AI in Dermatology: Somesh Gupta
- Federated Learning: Sharad Kumar
- Data Collection in Madagascar/Guinea: Dr. Ibrahima (accepted), Professor Rapelanoro
Presentations
- Annual Academy of Medicine in Antananarivo - August 10
- SSDV Annual Meeting - August 25-27 (Poster, Virtual): SGDV Congress
- EADV Annual Meeting in September (Poster, Virtual)
Future Presentations
- International Congress of Dermatology - November 12 by Ibrahima Traoré MD: ICD 2021 Program
Updates from Data Collection Centers and Technical Team
- SSDV Annual Meeting - August 25-27 (Poster, Virtual): SGDV Congress
- EADV Annual Meeting in September (Poster, Virtual)
Guinea
- Update on Case Collection (July-August-September): 32 cases (Total: 160
cases)
- Impetigo: 2, Tinea Capitis: 8, Scabies: 9, Tinea Corporis: 1, Intertrigo (Finger): 1, Tinea Pedis: 1, Onychomycosis: 2, Atopic Dermatitis: 4, Eczema: 1, Furuncle: 1, Chicken Pox: 1, Pitted Keratolysis: 1
Tanzania
- Update on Case Collection: Total number of cases 25
- Impetigo: 3, Tinea: 6, Scabies: 4, Atopic Dermatitis/Eczema: 12
Malawi
- Ethical Protocol Submitted in Mid-August and Approved
Madagascar
- Update on Case Collection:
- Total number of cases over the period: 158
- Atopic Dermatitis: 32, Eczema: 28, Scabies: 34, Fungal: 22, Intertrigo: 2, Seborrheic Dermatitis: 5, Nail Fungal: 6, Tinea Capitis: 3, Pityriasis Versicolor: 7, Pityriasis Alba: 1, Impetigo: 2, Impetiginization: 12, Insect Bites: 2
- 251 is the current total of cases in my.crf.one
Updates from Technical Team
-
Released an app for the PASSION Project for Android. The underlying model still uses some of the work done by the Indian Team as the model for PASSION-specific diseases requires more data. Discussion ongoing on hosting the Passion app - whether by the department in Basel or a separate AWS-like cloud server.
-
Preliminary work for a questionnaire to supplement the image model. Ideally, seeking higher accuracy in the image recognition model and then using the questionnaire to bring further improvements. More data is needed for this.
-
Experimentation started with Federated Learning model code from NVIDIA and Google TensorFlow. This substantial amount of work will continue for the next 6-9 months as the integration into the NVIDIA infrastructure is quite involved. Planning to obtain preliminary test results for a canned version of federated learning on TensorFlow by the end of the year.