Participants are tasked with developing algorithms that can effectively detect and classify Mycetoma from histopathological images.

Task 1: Detection of Mycetoma Grains

  • Objective: Develop an algorithm capable of accurately detecting the presence of Mycetoma grains within the provided histopathological images.
  • Background: Mycetoma grains are indicative of the disease, and their accurate detection is crucial for correct diagnosis and subsequent treatment planning.
  • Output Specification: Algorithms should output the boundary definitions of each detected Mycetoma grain within the images.

Task 2: Classification of Mycetoma Type

  • Objective: Classify detected Mycetoma grains into one of two categories: Actinomycetoma (caused by bacteria) or Eumycetoma (caused by fungi).
  • Background: The type of Mycetoma significantly affects treatment decisions. Actinomycetoma is generally treated with antibiotics, while Eumycetoma often requires antifungal medications and sometimes surgical intervention.
  • Output Specification: For each detected Mycetoma grain, the algorithm should output a classification label indicating whether it is Actinomycetoma or Eumycetoma.

Data Use and Integration

  • Participants are encouraged to utilize the training data provided to develop their detection and classification models. While additional external data and/or pre-trained models can be incorporated, transparency in their use is required for submission evaluation.

Algorithm Development

  • Development Environment: Participants may choose their development environment and tools but must ensure that their submission is compatible with the evaluation platform specified by the challenge organizers.
  • Interpretability: While not a requirement, the ability to explain decisions made by the algorithm will be viewed favourably, as it enhances trust in automated systems by clinicians.
  • Participants are encouraged to share their codes in the public repositories.