Plant identification applications do not reliably identify toxic and edible plants in the American Midwest

Kevan Long, Andrew Townesmith,Alex Overmiller,Wendy Applequist, Anthony Scalzo,Paula Buchanan,Cindy C. Bitter

CLINICAL TOXICOLOGY(2023)

引用 0|浏览4
暂无评分
摘要
Introduction Exposure to potentially toxic plants is a global problem, resulting in thousands of calls to poison centers and emergency department visits annually and occasional deaths. Persons with limited botanical knowledge may be tempted to rely on smartphone applications to determine if plants are safe to forage. This study evaluated the reliability of several popular smartphone applications to identify foraged foods and distinguish them from potentially toxic plants in the Midwestern United States. Methods Sixteen plant species were selected based on local availability, attractiveness as foraged food, and potential for misidentification. Of the 16 species, five are edible, three are potentially toxic if improperly harvested or prepared, and eight are considered to be toxic. Plant specimens were identified by graduate-level botanists and photographed during multiple stages of their growth cycles. LeafSnap, PictureThis, Pl@ntNet and PlantSnap were used to identify the plants. Results Overall accuracy of the applications in identifying plant genus was 76% (95% confidence interval: 73-79, range 96% for PictureThis to 53% for PlantSnap). Accuracy for identification of plant species was 58% (95% confidence interval 55-62%, range 94% for PictureThis to 34% for PlantSnap). Five of eleven potentially toxic species were identified as an edible species by at least one application. Conclusion Accuracy of the smartphone applications varies, with PictureThis outperforming other apps. At this time, apps cannot be used to safely identify edible plants. Foragers must have adequate botanical knowledge to ensure safe harvesting of wild plants.
更多
查看译文
关键词
Toxic plants,foraged foods,smartphone applications,plant identification,>
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要