On the objectivity, reliability, and validity of deep learning enabled bioimage analyses

eLife

Christoph M Flath¹  Matthias Griebel¹  Nikolai Stein¹  Alexander Dürr¹  Victoria Schoeffler²  Christina Lillesaar²  Teresa Lüffe²  Anupam Sah³  Nicolas Singewald³  Lucas B Comeras⁴  Ramon O Tasan⁴  Robert Blum⁵  Dennis Segebarth⁵  Rohini Gupta⁵  Cora R von Collenberg⁵  Manju Sasi⁵  Corinna Martin⁵  Dominik Fiedler⁶  Hans-Christian Pape⁶  Maren D Lange⁶ 

[1] Department of Business and Economics, University of Würzburg, Würzburg, Germany;Department of Child and Adolescent Psychiatry, Center of Mental Health, University Hospital Würzburg, Würzburg, Germany;Department of Pharmacology and Toxicology, Institute of Pharmacy and Center for Molecular Biosciences Innsbruck, University of Innsbruck, Innsbruck, Austria;Department of Pharmacology, Medical University of Innsbruck, Innsbruck, Austria;Institute of Clinical Neurobiology, University Hospital Würzburg, Würzburg, Germany;Institute of Physiology I, Westfälische Wilhlems-Universität, Münster, Germany;

关键词: bioimage informatics;    deep learning;    reproducibility;    objectivity;    validity;    fluorescence microscopy;   

DOI  :  10.7554/eLife.59780

来源: DOAJ

【 摘 要 】

Bioimage analysis of fluorescent labels is widely used in the life sciences. Recent advances in deep learning (DL) allow automating time-consuming manual image analysis processes based on annotated training data. However, manual annotation of fluorescent features with a low signal-to-noise ratio is somewhat subjective. Training DL models on subjective annotations may be instable or yield biased models. In turn, these models may be unable to reliably detect biological effects. An analysis pipeline integrating data annotation, ground truth estimation, and model training can mitigate this risk. To evaluate this integrated process, we compared different DL-based analysis approaches. With data from two model organisms (mice, zebrafish) and five laboratories, we show that ground truth estimation from multiple human annotators helps to establish objectivity in fluorescent feature annotations. Furthermore, ensembles of multiple models trained on the estimated ground truth establish reliability and validity. Our research provides guidelines for reproducible DL-based bioimage analyses.

【 授权许可】

Unknown