Hello,
The instructions mention that make sure the output csv has patient-wise prediction instead of individual cough prediction. I was wondering if this means that we need to average the probabilities based on each participant and then calculate the AUROC. I also checked the https://github.com/Sage-Bionetworks/tb_challenge_demo/blob/main/inference_tb/inference.py but I could not find how the data is merged based on participants from individual cough prediction.
Please advise.
Created by Sina Akbarian Sakb @Sakb-
You are responsible for generating the patient-specific TB predictions in whatever way you want to to aggregate across the multiple coughs per patient.
Per the instructions, your code should generate a file that contains two columns:
* participant: Patient id
* probability: TB Positive class confidence score (probabilities are between [0, 1])