DSSApple: A hybrid expert system for the diagnosis of post-harvest diseases of apple

Abstract

Post-harvest diseases of apple can cause considerable economic losses. Thus, we developed DSSApple, an interactive web-based decision support system, that helps users to diagnose post-harvest diseases of domesticated apple based on observed macroscopic symptoms on fruit. Specifically, DSSApple is designed as a two-stream hybrid diagnostic tool, that can be effectively used by both expert and non-expert users to diagnose diseased instances of apple. The image-based stream allows the user to interact simply by selecting pictures, representing the variety of symptoms of diseases at different stages of the infection and on different cultivars. Instead, the expert-based stream of the system incrementally collects user feedback about the target disease by asking questions related to the macroscopic characteristics of the observed symptoms on a target apple. The expert-based reasoning mechanism of DSSApple is developed by leveraging the framework of Bayesian Networks (BNs). We detail the process of building this knowledge base with the support of a domain expert. We further exploit the BN to process incomplete or conflicting user feedback within the inference mechanism as well as to provide human-understandable explanations on the suggested diagnoses. The proposed hybrid approach has been thoroughly evaluated in two studies, involving simulated (by photos) as well as real infected apples. Thus, the proposed hybrid version of DSSApple is able to outperform both the single streams and the user intuition in terms of diagnostic accuracy.