Artificial Intelligence (AI) is boosting scientific research in complex fields such as neuroscience. A recent study conducted by researchers at Ecole Polytechnique Fédérale de Lausanne (EPFL) in Switzerland and Harvard University in the U.S. demonstrates the potential of AI in advancing neuroscience by identifying and tracking neurons in moving animals. Cognitive neuroscience can greatly benefit from machine learning as it relieves the need for the strenuous and time-consuming task of manual annotation, which involves labeling data within datasets to be used by machine learning algorithms.
- The EPFL and Harvard researchers utilized an AI convolutional neural network (CNN) model, capable of performing targeted augmentation to annotate complex brain-imaging data swiftly. They named this technique “Targettrack”.
- The study experimented on brain-wide imaging data of C. elegans (Caenorhabditis elegans), a nematode worm frequently used in research for neurodegenerative diseases.
- The utilization of Targettrack reported a significant improvement, reducing a 200-hour task of annotating 76 neurons in a 10-minute recording of moving C. elegans to just 65 hours – an enhancement of 67.5%.
- When a convolutional neural network is trained with a combination of images annotated both manually and synthetically, the need for proofreading decreased due to increased reliability.
This revolutionary innovation allows CNN to minimize the required amount of manually annotated training data, proving to be greatly efficient in annotating brain imaging problems. With the reduction in duration for manual annotations and proofreading, a 50-fold increase in throughput in comparison to full manual annotation is achieved.