- Title：MAIB-Talk-022: Deep Learning-Enabled Morphometric Analysis for Toxicity Screening Using Zebrafish Larvae
- Date：10:00pm US East time, 07/15/2023
- Date：10:00am Beijing time, 07/06/2023
- Zoom ID：933 1613 9423
- Zoom PWD：416262
- Zoom: https://uwmadison.zoom.us/meeting/register/tJcudu-prTIuGNda1MsF8PKyRQlnGn06TP2E
MAIB: Manifold learning, Artificial Intelligence, Biology Forum (MAIB)
Presentation Record(Previous Presentation will be showed here if the video is not released for this talk)
Dr. Sijie Lin (林思劼), Professor at Tongji University, Shanghai, China
Dr. Lin joined the faculty of his Alma mater, Tongji University in 2016. His 90+ publications have acquired 11,000+ total citations, with an H-index of 48. Dr. Lin serves as an Associate Editor of Frontiers in Toxicology and editorial board member of NanoImpact, Human and Experimental Toxicology, The Innovation and Chinese Chemical Letters.
Research Interests: nano/bio interface, environmental toxicology, environmental nanotechnology, zebrafish model, high-throughput screening and methodologies
https://www.researchgate.net/profile/Sijie-Lin-4
https://scholar.google.com/citations?user=V0IPah4AAAAJ&hl=en
Title: Deep Learning-Enabled Morphometric Analysis for Toxicity Screening Using Zebrafish Larvae
Toxicology studies heavily rely on morphometric analysis to detect abnormalities and diagnose disease processes. The emergence of ever-increasing varieties of environmental pollutants makes it difficult to perform timely assessments, especially using in vivo models. Herein, we propose a deep learning-based morphometric analysis (DLMA) to quantitatively identify eight abnormal phenotypes (head hemorrhage, jaw malformation, uninflated swim bladder, pericardial edema, yolk edema, bent spine, dead, unhatched) and eight vital organ features (eye, head, jaw, heart, yolk, swim bladder, body length, and curvature) of zebrafish larvae. A data set composed of 2532 bright-field micrographs of zebrafish larvae at 120 h post fertilization was generated from toxicity screening of three categories of chemicals, i.e., endocrine disruptors (perfluorooctanesulfonate and bisphenol A), heavy metals (CdCl2 and PbI2), and emerging organic pollutants (acetaminophen, 2,7-dibromocarbazole, 3-monobromocarbazo, 3,6-dibromocarbazole, and 1,3,6,8-tetrabromocarbazo). Two typical deep learning models, one-stage and two-stage models (TensorMask, Mask R-CNN), were trained to implement phenotypic feature classification and segmentation. The accuracy was statistically validated with a mean average precision >0.93 in unlabeled data sets and a mean accuracy >0.86 in previously published data sets. Such a method effectively enables subjective morphometric analysis of zebrafish larvae to achieve efficient hazard identification of both chemicals and environmental pollutants.
Background to known before you coming to the presentation:
Toxicology: Understanding the basic principles of toxicology, including the assessment of chemical hazards and the potential risks posed by environmental pollutants.
Morphometric analysis: Familiarity with the concept of morphometric analysis, which involves the measurement and analysis of physical features and structures to detect abnormalities and diagnose diseases.
Zebrafish as a model organism: Knowledge about zebrafish as a commonly used model organism in toxicology studies. Understanding their developmental stages, anatomy, and relevance to human health research would provide a context for the proposed study.
Deep learning and computer vision: Awareness of deep learning techniques and their application in computer vision tasks, such as image classification and segmentation. Understanding the basics of neural networks and their ability to learn complex patterns from data would be beneficial.
Environmental pollutants: A general understanding of different categories of environmental pollutants, including endocrine disruptors, heavy metals, and emerging organic pollutants. Awareness of their potential adverse effects on organisms and the need for efficient hazard identification is important.
Data analysis and validation: Familiarity with statistical validation methods used in evaluating the accuracy and performance of machine learning models. Knowledge of metrics such as mean average precision (mAP) and mean accuracy would aid in understanding the results presented in the study.
What are challenges for this filed:
Dataset variability: Obtaining a diverse and representative dataset that covers a wide range of chemical exposures and environmental conditions is essential. However, collecting such datasets can be challenging due to the extensive variability in chemical structures, concentrations, exposure durations, and environmental factors.
Annotation and ground truth generation: Generating accurate annotations and ground truth data for training the deep learning models can be labor-intensive and time-consuming. Expert knowledge and manual labeling are often required to identify abnormal phenotypes and accurately segment vital organ features in zebrafish larvae images.
Generalization to new chemicals and pollutants: Deep learning models trained on specific chemicals and pollutants may struggle to generalize to new, unseen compounds or environmental pollutants. The performance of the models may vary depending on the characteristics and unique effects of each chemical or pollutant, requiring ongoing model adaptation and refinement.
Interpretability and explainability: Deep learning models are often considered “black boxes” due to their complex architectures and internal workings. Interpreting the models’ decisions and understanding the features they rely on to identify abnormalities and segment organ structures is a challenge. Ensuring the interpretability and explainability of the models’ outputs is crucial for gaining trust and acceptance in the scientific community.
Ethical considerations: The use of zebrafish larvae as a model organism raises ethical considerations regarding animal welfare. Efforts should be made to minimize the number of animals used, refine experimental protocols to minimize suffering, and adhere to ethical guidelines and regulations.
Integration with regulatory frameworks: To effectively translate deep learning-based morphometric analysis into regulatory practice, integration with existing regulatory frameworks is necessary. Developing guidelines and standards for evaluating the performance and reliability of deep learning models in hazard identification is essential to ensure their acceptance and adoption in regulatory decision-making processes.
Addressing these challenges requires interdisciplinary collaboration between toxicologists, computer scientists, and regulatory experts. Overcoming these hurdles will contribute to the advancement of efficient and reliable methods for identifying chemical hazards and environmental pollutants, ultimately supporting the protection of human and environmental health.
林博士于2016年加入母校同济大学，发表论文90余篇，总被引次数超过11,000次，H指数为48。林博士担任Frontiers in Toxicology副主编和编委。 NanoImpact、人类和实验毒理学、创新和中国化学快报成员。
