An Automatic Behavioral Detection System For High Throughput Screening Of Auditory Function Of Zebrafish

Damage of hair cells resulted from ototoxic drug and gene mutations is the primary cause of hearing loss,especially sensorineural deafness.Because of the importance of early diagnosis and intervention of the sensorineural deafness,as well as lack of treatment at present,high-efficiency screening for the ototoxic drug and therapeutic agents is of great significance in both clinical and preclinical applications.Zebrafish is a common model animal used for screening ototoxicity drugs.The current techniques for the assessment of auditory function based on electrophysiological or morphological studies,especially in animal models,are prevented from being gold standards owing to their limited accuracy.As complementary but more direct and efficient methods,behavioral tests,such as the examination of startle responses,are used to evaluate the auditory function,mainly in zebrafish.The development of computer-vision and machine-learning techniques remarkably benefits the behavioral tests by enabling automatic detection,recognition,and tracking of zebrafish motions and behaviors.In the presented research project,an automatic detection system capable of high-throughput in vivo behavioral tests in zebrafish,as well as the corresponding software that is responsible for both the hardware control and video-image processing,has been designed and developed.The main hardware components of the detection system include a couple of plates within each of which 20-30 zebrafish can freely swim without interruptions.With a video camera and a drive belt,video recording can be conducted above every plate in sequence,which warrants the high efficiency of the system.To preclude the disturbance from visual inputs,the entire behavioral tests are designed to be performed within darkness.The illumination of the video recording is provided using invisible infrared light.Regarding the software,both deep-learning network,yolov3,and OpenCV library are employed and,as a result,high-accuracy recognition and tracking of the animal motions and behaviors can be achieved.For validating the efficacy of the developed system and the corresponding software,startle response and prepulse inhibition of the zebrafish that is exposed to ototoxic drug,including neomycin and cis-platinum,were recorded and analyzed.The acquired results demonstrated that the motions and behaviors of the zebrafish could be automatically recorded,tracked,and analyzed and,more essentially,the impairment of vestibular function could be successfully evaluated using the developed system and software.