-this research provide approach for embryo cell detection from images based on convolutional neural networks. Deep neural network used for experiment consist of 15 layers and is trained using GPU for calculations. In research training data set size impact to model training duration is identified. R-CNN model embryo cell detection results are compared to human expert labeling data to evaluate its precision.
I.
INTRODUCTION
Identify and count objects in an image or sequence of images is challenging computer vision problem, which can be found in many applications and systems, ranging from traffic monitoring to biological research. This paper is focused on biological research embryo cells detection. However, developed methodology can be used in numerous medicine procedures that requires counting and detection, such as red or white blood cells count for patient’s health, clinical pathology or cell concentration investigation.
Manual embryo cell detection is very monotonous and timeconsuming work that is prone to errors. According to this automating, the detection process has many benefits, such as reducing time consumption, minimizing errors possibility and cost. In addition, it is improving consistency of results between individuals and clinics. Our goal is to simplify the task and improve its robustness.
One of the difficulties is to count non-stained cells in dark images, because of constraints, such as the light intensity, transparency or exposure time. All these factors cause image quality and result in faint cell boundaries. One more challenge is that embryos cells has wide variability in appearance and shape. Furthermore, every embryo grows in different individual manner, there cells overlap each over. Also between cells could be found extracellular material hand crafted algorithms.
In this paper, we develop a Convolutional Regression Networks (CNN) approach for regression of density map. Our main goal is to automatically detect and count the number of human cells in developing embryos. In addition, experimental Copyright © 2017 held by the authors results shows that CNN can be used to provide state-of-the-art cell counting, also detect overlapping cells.
II.
Object detection system R-CNN can be divided into three
main modules. The first module is responsible for generation
category-independent region proposals [
According to the Pan and Yang taxonomy [
Donahue et al. [
Regions with Convolutional Neural Networks consists of
two sibling output layers. The first one is used for discrete
probability distribution, p = (p0,...,pK), over K + 1 categories.
Always parameter p is computed by a softmax over outputs of
layer. The second one layer outputs bounding – box regression
offsets, tk = ( ) given in [
R-CNN regions of interest training is labeled with u (ground – truth class) and v (ground – truth bounding – box regression target). Then multitask loss L for each RoI classification (1). (1)
Here Lcls is equal to log loss for true class Lcls(p,u) = - log pu the hyper – parameter is used to control the balance between the two task losses. For normalization the ground – truth regression targets vi is equal to zero mean and unit variance.
In the second task is defined loss over a tuple of true bounding – box regression targets Lloc for class u and v and is equal to v = (vx,vy,vw,vh), predicted tuple for class u. For background regions of interest there is not used notion of ground – truth bounding box also Lloc is not involved. In this case for bounding – box regression is used (2) expression. (2)
(3)
If the regression targets are unbounded, there is probability, that training with L2 loss can require tuning of learning rates to prevent gradients explosion.
III.
Deep learning is package of different methods used in machine learning which attempts to present detail features in multiple-layer structure data. R-CNN is one of the most effective learning techniques and is able to minimize learnable parameters significantly by using the same basis function across different image locations.
In this research, we suggest an automatic learning based cell
detection framework, which is suitable for 3D and 2D
microscopy images. This framework can be used, for the
efficiency and accuracy improvement of training a CNN from
larger size images, an SVM classifier is applied to detect cell
regions for collecting the CNN training set [
The exposure time range is dynamical and may not be equal
for each session of recording through the light microscope,
according to this the color of each stack may be different. Also
in this research we apply Image Intensity Standardization (IIS),
which was considered in [
Fig. 1 The workflow of the embryo cell detection framework.
As well, there are more algorithms used for embryo cell
detection in machine vision and medical image analysis areas
[
In this research, cell regions R are determined to discard the irrelevant background regions. Selecting background patches is important for training a CNN. Wherefore, cell regions detection is more efficient and rough using an SVM classifier, after that cell and background training patches are gathered from R instead of the whole stack.
The Support Vector Machine (SVM) detector is used for cell region detection and for collecting CNN training patches, which are used to remove large part of background pixels. This part of process is like feature selection pre-process. In our case, accuracy of CNN could be increased using cell detection samples in the cell region. Similarly, in the test case, to increase accuracy, in first step we apply the SVM detector to identify those regions. In the training mode of conventional CNN, the cell samples are the same, however the non-cell samples are different.
Then the cell region R is detected using SVM-RGB Histogram detector, second step is to extract cell and patches in region R from all test stacks for training CNN which is also the same size of patches and neighborhood. Pixels in the cell region R have almost same colors. According to this color feature in the cell region is not reliable for distinguishing cell and background patches. On purpose to decrease time range for training all RGB patches are transferred into the YUV color space and only the Y channel patches are needed. Every Ychannel cell patch, is rotated 0, 90, 180, 270 degrees to ensure the detector rotation invariant and increase the amount of cell samples. Also there is probability that cell and background patches can have overlapping pixels. This is useful for increasing the probability of correct cell detection. Approximately half million cell patches are extracted from all training stacks, and the same amount of background patches from the cell region R.
After the last step, max-pooling CNN is ready for testing on the test stacks. The cell region is detected by the SVM RGB Histogram detector for each frame of every stack in the dataset used for testing. After that, the pre-trained CNN is used for identifying embryo cells by scanning each pixel in region and every pixel is given a probability value P.
IV.
R-CNN TRAINING
Experiment was done using MATLAB 2016b software in a personal computer with i5-4570 CPU clocked at 3.2 GHz, 8 GB memory 64-bit operating system and video card GeForce GTX 650 Ti. Training process was done with GPU processor instead of CPU to accelerate training procedure.
We train the R-CNN network demonstrated in Fig. 2. It consist of 1 input layer, 13 hidden layers (convolutional, Relu, Max Pooling, Fully Connected, Softmax) and classification output layer. Training run for 100 epoch, with base learning rate of 0.001 and Stochastic Gradient Descent training method.
For experiment there was randomly selected thousand embryo photos Fig. 3 which was labeled by human expert. To evaluate training data set size impact to detection precision there was trained 14 R-CNN networks with different size training data set. Data set size for training was increased from 5% to 70% with 5% steps. 30% of data set was used to evaluate network cell detection precision. To decrease training time there was used pre trained CIFAR-10 network. Pre trained network biases and neuron weights there adjust to detect embryos cells in photos.
Training data set size impact to training time is linear and it can be seen in Fig. 4. Training duration using biggest training dataset with 700 images was 38 minutes 40 seconds.
Trained R-CNN network was tested with new, do not used at training process, 300 embryo images Fig. 5. Predicted embryo position and size was compared with human expert labeled embryo size and position results.
After comparing specialist data labeling results with deep neural network result, received size mean squared error and standard deviation presented at Table. 1. Some trained neural networks do not detected one or two embryos cell at images. Models with 30% or bigger size training dataset detected all embryos in images.
From Fig. 6 it is seen that best results got with 20% and 70% size of training data set where models results compared with human expert gives 11.92% mean square error. It shows that not only training data set size impacts model accuracy, but images distribution in training data set influence model prediction accuracy.
Comparing model position predicting results with human expertise prediction from Table 2 it seen that error rate is smaller than size error rate. Smallest mean square error rate got using model trained with 30%, 40% and 65% training data set size. Close error rate got using 25% training data set, but this model do not detect one embryo cell. 25% 30% 35% 40% 45% 50% 55% 6,05 5,59 5,45 5,29 4,64 4,64 6,82 4,63 5,29 5,06 5,42 2,55 2,42 2,76 2,58 2,07 2,17 2,92 2,28 2,5 2,25 2,31 1 2 0 1 1 0 0 0 0 0 0 65% 70%
At Fig. 7 it seen whole error distribution. Inaccuracies appears using model with 35% training data set. That means few images could distort model parameters and decrease its accuracy.
CONCLUSIONS
From experiment results it is possible to confirm that deep neural network training time is linearly dependent to training data set size. After detected region size comparison with human expertise prediction best result with mean square error rate 11.92% without any undetected embryos cell got using biggest 70% training data set size. More precise result got comparing embryos cell position. Smallest error 4.62% got using 65% training data set size. This shows that offered model better works for position prediction.