Unveiling the infectious morphological behaviour of banana crop pathogenic nematodes inhabited from soil medium to pseudostem using an artificial intelligence approach

Abstract

Soil-borne microorganisms target the rhizosphere by invading from soil to plants through pseudostem. Fusarium oxysporum f.sp. cubense, an infectious agent’s host, interacts with nematodes present in the single point regional area (SPRA), causing tissue necrosis, and physical disordering of banana plants poses high yield loss. Diagnosing the source of pathogenic microbes on a crop significantly prevents its transmission to other regions. Disease characteristics cannot be accurately assessed through physical observation alone. We proposed Nematode Detection and Morphological Analysis (NDMA-YOLO), a deep learning-based futuristic algorithm, and Tracking Live Parasites (TLP) to tackle this issue. Experiments demonstrated in Fusarium-affected fields with similar soil properties. The chemical composition of soils is characterized by FTIR spectroscopic analysis, pH, moisture, SEM, and fluorescence spectrophotometer content characteristics. Physically identifying the source of infection using the (x, y) Grid Ring Axis Pseudo Stem Holistic (GRAPH) method, obtained plant tissue samples, and generated large image datasets through phase contrast microscopic. Recorded structure of nematodes to understand physiological, behavioral, and biotic stress patterns. We utilized AI-based computer vision for live event monitoring and morphological analysis, employing an enhanced YOLO-v8 model trained on a custom dataset to detect nematodes with 86.66 % accuracy and an overall performance of 98.93%. Our model surpasses previous versions like YOLO-v3, YOLO-v5, and YOLO-v7, showcasing significant advancements in dataset preparation for accurate predictions in plant pathology.