10 SEED TESTING INTERNATIONAL www.seedtest.org FEATURE ARTICLE SABIMA: An Objective Method for Seed Vigour Assessment Using Surface Area Measurement Tahl Paran Global Seed Biology and Testing, Syngenta, Basel, Switzerland; tahl.paran@syngenta.com SEED VIGOUR ASSESSMENT IS CRITICAL IN SEED SAFETY EVALUATION, to ensure seed treatment does not impact physiological quality. Traditional manual evaluation methods are subjective and prone to human bias. We present SABIMA (Surface Area BIo MAss), a patent-protected image processing tool that quantifies seedling vigour through root and shoot surface area measurement. Strong correlations were established between SABIMA and manual germination evaluation (R² = 0.902 for wheat, N = 104; R² = 0.918 for corn, N = 16) and field emergence (R² = 0.82, N = 33). SABIMA offers an objective, quantitative alternative to traditional vigour assessment for seed industry quality control. Keywords: seed vigour, image analysis, germination testing, seed treatment, machine learning Introduction At Syngenta’s Seedcare and Biologicals Institute, seed safety ensures treatments do not compromise seed quality. We evaluate whether treatments maintain vigour and germination across crops and storage conditions. This is critical because effective treatments become unusable if they impair germination or establishment. Current assessment relies on rolled paper towel germination tests following the International Rules for Seed Testing (ISTA Rules) [1], involving visual classification of seedlings as normal, abnormal or dead. While standardised, this method has limitations: (i) subjectivity between technicians; (ii) binary classification without quantifying vigour degrees [2]; (iii) labour intensity; and (iv) limited sensitivity to subtle differences [3]. These limitations prompted the exploration of objective, quantitative alternatives. The SABIMA Concept SABIMA is a patent-protected automated image analysis tool using machine learning trained on hundreds of annotated germination images to identify root and shoot tissues across crops and backgrounds. Key Distinction: Unlike manual evaluation counting percentage of normal seedlings, SABIMA quantifies total root and shoot surface area (cm²). More vigorous seed lots produce greater tissue development, resulting in larger surface area measurements. Advantages: (i) Objectivity, it eliminates human bias; (ii) sensitivity, it detects subtle vigour differences such as germination rate; (iii) quantification, it provides numerical data for analysis; and (iv) automation, it reduces labour. Implementation: Web-based interface where users upload rolled paper towel test images. A model automatically segments images, measuring root and shoot areas for every image. Development and Validation Pouria Sadeghi-Teheran (Senior Digital Imaging and Phenotyping Expert, Syngenta CP R&D, Switzerland) created and trained the model. Helen Day (Laboratory Manager, NIAB, UK) facilitated germination data generation for comparison to manual evaluation. Brent Reschly (Global Innovation Lead, P&S Global Seed Quality, Syngenta Seeds USA) coordinated vigour tests to be compared with field emergence trials. Results Wheat Correlation to Manual Evaluation Wheat samples (N=104, 100 seeds/replicate) underwent cold germination testing. SABIMA measured total surface area while technicians performed ISTA evaluation. Statistics: R² = 0.902, F = 937, N = 104 90.2% of manual evaluation variation is explained by SABIMA measurements. Corn Manual Evaluation Correlation Corn samples (N = 16, 50 seeds/replicate) were tested under optimal conditions. Statistics: R² = 0.918, F = 156, N = 16 91.8% of manual evaluation variation is predicted by SABIMA. Corn Field Emergence Correlation Corn samples (N = 33) were evaluated in laboratory and field emergence trials at 150 growing degree units (GDU). Shoot surface area was examined as a predictor. Statistics: R² = 0.82, F = 149, N = 33 82% of field emergence variation is predicted from laboratory shoot area measurements. Practical Value: Early quality assessment before planting; problematic lot identification before distribution. Figure 1. Wheat correlation between SABIMA and manual evaluation: [left] regression plot showing SABIMA total surface area (cm2) vs % normal seedlings; [right] representative seedling images showing high vigour sample (large root and shoot area) and low vigour sample (reduced tissue development)
View this content as a flipbook by clicking here.