Fig 1.
Influenza-induced cytopathic effects (CPEs) in MDCK cells.
MDCK cells were mock-infected (a), or infected with influenza viruses, which resulting in 20% (b), 50% (c) and 80% (d) of cytopathic effects. Scale bars: 200 μm. Black arrows indicate influenza-induced cytopathic effects. White arrows indicate the area where all infected cells were completely detached and floating in the culture medium.
Table 1.
Information table for the Training 1 training data.
Fig 2.
Influenza-induced cytopathic effects (CPEs) at different time points after infection.
Images of MDCK cells were taken at 25hpi (hours post infection) (a-c), 16hpi (d-f), 28hpi (g), and 40hpi (h). (a) and (d) represent MDCK cells which were mock-infected. (b)(e) and (c)(f) represent MDCK cells which were infected with influenza viruses at 0.05 M.O.I. and 0.5 M.O.I., respectively. (g) represents image of MDCK cells infected with 0.5 M.O.I. of influenza virus at 28hpi. (h) represents image of MDCK cells infected with 0.05 M.O.I. of influenza virus at 40hpi. Scale bars: 200 μm. Black arrows indicate influenza-induced cytopathic effects.
Table 2.
Comparison of Training 1 and Training 2 with 1200 epochs weights and the saved weights.
Fig 3.
Infection of MDCK cells by non-influenza viruses.
MDCK cells were infected with herpes simplex virus type 1 (HSV-1) (a), herpes simplex virus type 2 (HSV-2) (b), adenovirus (c), coxsackie B3 virus (d), parainfluenza 3 virus (e), and respiratory syncytium virus (RSV)(f). The images were taken at 40 hpi (hours post infection). Scale bars: 200 μm.
Table 3.
Information table for the influenza experiment data set.
Table 4.
Comparison of Training 1 and Training 2 with 1200 epochs weights and the saved weights on Experiment Data.
Fig 4.
Flow charts for Training 1 A and Training 2 B.
(A) The flow chart of Training 1. The blue, red, and black frames indicate the data condition of influenza infection in the training process, testing process, and limitation set, respectively. The viruses in the yellow frame were used for specificity test. (B) The flow chart of Training 2. The blue frame indicates the data condition of influenza infection in the training and testing process. The viruses in the yellow frame were used for specificity test. The black frame indicates the data condition of influenza infection in the limitation set.
Table 5.
Comparison of Training 1, Training 2 and visual examination.
Fig 5.
Architecture of the Convolutional Neural Network model.
There are three convolutional layers with max pooling and four fully-connected layers. The numbers beside the input indicate the size of the image. The numbers of the convolutions and pooling are the size of output from previous layer. The numbers in the cube indicate the size of convolutional filter and pooling kernel. The bold numbers are the number of the neurons in the fully-connected layers.