The objective of this work was to develop and to validate an Expert System (ES) in order to facilitate the fungi identification in the seed health analysis. The ES will be able to help in the identification of 46 major economical importance fungi that occur in seeds of species such as bean, carrot, corn, cotton, rice, sorghum, soybean, sunflower and wheat submitted to the blotter test. The ES contains pictures of the pathogens on the seed surface and in glass slides mounts, under different magnification of the stereomicroscope and compound microscope. To increase the level of certainty by the user, the ES has a glossary of technical terms and texts related to the pictures. When the user arrives at a diagnosis, the system shows a reliability level in the reply. The system was validated by 14 users of three different levels of knowledge: high (group 1), medium (group 2) and low (group 3) levels of experience/knowledge in seed health analysis. Based on the percentage of success obtained before and after the use of the ES, the following results were observed: group 1: before accessing the program, the average was 62,30%, and after use, 95,24%; for the groups 2 and 3: 0% of success before using the program for both groups. After the use, the success percentages were 88,10% and 95,24%, respectively. Considering all the fungi tested in the validation phase, independent of their hosts, the ES in seed health analysis provided for the group 1 an average percentage of success increment of 35,33% after the use of the system, of 86% for the group 2, and, of 94% for the group 3. By means of the ÷² test, considering expected certainty frequencies of 90%, the results before the use of the system were significant for the groups 2 and 3, and not significant for the group 1. After the use of the system, the results were significant for all groups, and it confirms that the expected certainty frequencies of 90% were obtained. In that way, it can be verified that the program increases considerably the accuracy and precision of the fungi identification in the seed health analysis and it makes possible that professionals without previous knowledge in the area can access specific information as in the seed health analysis by the blotter test method.
pathology; artificial inteligence; fungi identification
