A comparison of ImageJ and machine learning based image analysis methods to measure cassava bacterial blight disease severity

• DOI: 10.1186/s13007-022-00906-x
• Publication Date: 2022-06-21T09:03:11Z
• Authors: Kiona Elliott; Jeffrey C. Berry; Hobin Kim; Rebecca S. Bart
• Keywords: 0301 basic medicine; 2. Zero hunger; Disease symptoms; Cassava bacterial blight; QH301-705.5; Methodology; Plant culture; ImageJ; Image analysis; SB1-1110; 03 medical and health sciences; Machine learning; Biology (General)

Methods to accurately quantify disease severity are fundamental plant pathogen interaction studies. Commonly used methods include visual scoring of symptoms, tracking growth in planta over time, and various assays that detect defense responses. Several image-based for phenotyping symptoms have also been developed. Each these has different advantages limitations which should be carefully considered when choosing an approach interpreting the results.In this paper, we developed two image analysis tested their ability aspects lesions cassava-Xanthomonas pathosystem. The first method uses ImageJ, open-source platform widely biological sciences. second is a few-shot support vector machine learning tool classifier file trained with five representative infected leaf images lesion recognition. Cassava leaves were syringe infiltrated wildtype Xanthomonas, Xanthomonas mutant decreased virulence, mock treatments. Digital captured overtime using Raspberry Pi camera. analyzed compared segment from background capture measure differences between treatment types.Both presented paper allow accurate segmentation non-infected plant. Specifically, at 4-, 6-, 9-days post inoculation (DPI), both provided quantitative types. Thus, either could applied extract information about severity. Strengths weaknesses each discussed.