Colony Counting_Innovative Technology (2): Filter and 3M Test Strip Detection
2022-09-14 03:03:43
Filter membranes and 3M test strips are increasingly used for microbial detection in the fields of biology, medicine, food, and the environment. A lot of filters or test strips have a feature that they have grid lines of different colors and sizes on their surface. These grids help to manually count the counts, but they present difficulties for the current use of more and more automatic colony counters. Because the color of the grid is often very deep, traditional methods often detect grids rather than colonies. The following figures 1, 2, and 3 show three commonly used filters and 3M test pieces, and the segmentation results using the traditional image segmentation method.
Among them, Figure 1-a is the original image of the Sedoris filter, with a black grid on the surface and light yellow colonies growing. Figure 1-b shows the segmentation effect using the traditional threshold segmentation method. Figure 1-c shows the segmentation effect using the color gradient method. Since the mesh color of the filter is deeper than that of the colony, the conventional image processing method divides the mesh instead of the colony.
Figure 1. Sedoris filter
Figure 2-a is a 3M S. aureus test piece with a yellow grid and purple colonies on the surface. Figure 1-b shows the segmentation effect using the traditional threshold segmentation method. Figure 1-c shows the segmentation effect using the color gradient method. These two segmentation methods divide the mesh while segmenting the colonies, so that the colonies cannot be counted.
Figure 2. 3M Staphylococcus aureus test piece
Figure 3-a is a 3M E. coli test piece with a red grid, red colonies, and a red background. This situation is the most complicated, and it is impossible to obtain the desired effect regardless of the traditional threshold segmentation method (Fig. 3-b) or the color gradient segmentation method (Fig. 3-c).
Figure 3. 3M E. coli test piece
It turns out that the traditional image processing technology can not solve the colony detection of the above filter membrane or 3M test strip, and it is necessary to study and establish a new detection method. Shineso _ Technology team, two-year research, based on the world's most advanced level set active contour model, combining the characteristics of filters and 3M test sheet to form a new segmentation algorithm suitable membrane or 3M test piece, successfully resolved The above problem.
1. Level set active contour model based on morphological constraints
The principle of image segmentation based on the horizontal set active contour model is to make the active contour approach the segmentation target in the process of minimizing the energy functional. If a constraint based on prior knowledge is introduced into the energy functional, the active contour is approximated to the target specified by the constraint, and the desired target can be segmented. The two ideas proposed earlier in the world are mainly the following two ideas.
One is a model based on prior knowledge constraints proposed by Cremers. The shape determined by the prior knowledge is represented by the level set Ф 0. The active contour model based on the shape prior knowledge adds a shape-constrained energy term to the energy functional to guide the curve to converge to this shape:
Where L defines the extent to which shape prior knowledge occurs, and the region where L = -1 is excluded from the integral. This method strictly defines the position and size of the shape information and is limited in practical applications.
The other is Tony Chan's active contour model based on shape prior knowledge, which allows the translation, scaling and rotation of shapes. If the level set Ф 2 is obtained by translating, rotating, and scaling the level set , 1 , and the translation coordinates are a, b , the scaling multiple is r , and the rotation angle is θ, then the relationship expressions of the two levels are:
If ψ0 is a level set function of a fixed shape, the level set function is obtained by solving the symbol distance function. ψ is the horizontal set function corresponding to the original shape after translation, rotation or scaling. Then the level set model energy function based on shape prior knowledge is:
The numerical solution of the above two methods involves solving the gradient descent flow of multiple variables of the energy function. Each curve iteration needs to update multiple variables, so the approximation speed of the active contour model is very slow and cannot be practically adopted.
For mesh filters or 3M test pieces, the colonies to be detected are usually circular, and the prior knowledge of the shape of the level set can be set to a circle. For a circle, the translation and scaling in its geometric transformation will only result in a shape change, and rotation will not affect the circular shape, so a circular-based level set model requires only three additional variables ( a, b , r ), the energy functional of the horizontal set active contour model based on circular constraints can be simplified as:
In the formula, the second term is the constraint term, which acts to cause the final contour to converge to a circle. In addition to the gradient descent flow required for the level set function, only three variables ( a, b, r ) need to be updated, the number of iterations is greatly reduced, and the segmentation speed is improved.
In order to realize the synchronous detection of multiple colonies on the plate, it is necessary to further introduce the multi-phase level set active contour model. At the same time, in order to further improve the detection speed, a fast solution method of the level set active contour model is needed. In these two aspects, the research team of the Xun Science and Technology Department has achieved important results and practical applications. For more information, please refer to "Cluster Counting_Innovative Technology (I): Level Set Activity Contour Model" published by "XunQian Technology Co., Ltd.".
2 , the detection of surface wrinkles, edge blurred colonies
The level set active contour model based on morphological constraints not only preserves the strong anti-noise performance, smooth and continuous segmentation boundary, and can deal with the complex surface structure, but also can very well approximate the circle. aims. Especially for some colonies or cells with blurred contours and serious surface wrinkles, they show extremely superior segmentation effects.
Figure 4 shows the detection effect of a marginally blurred protoplast cell. Figure 4-a shows the protoplast cell original map; Figure 4-b uses the general horizontal set active contour model. Due to the lack of circular constraints, a non-circle is detected; Figure 4-c uses the fast According to the “morphological constraint-based level set active contour model†algorithm developed by Digital Science and Technology, due to the circular constraint, the final approximation must be a circle, which can restore the original state of the cell well.
Figure 4. Detection of edge-blurred protoplast cells
Figure 5 shows the detection effect of protoplast cells with a very severe surface wrinkle. Figure 4-a shows the protoplast cell original map; Figure 4-b uses the general horizontal set active contour model. Due to the lack of circular constraints, a pile of debris is detected; Figure 4-c uses the fast According to the “morphological constraint-based level set active contour model†algorithm developed by Digital Science and Technology, due to the circular constraint, the final approximation is a complete protoplast cell.
Figure 5. Detection of surface wrinkled protoplast cells
3 , the detection effect of the filter membrane and 3M test piece
Figure 6 shows the “shape-constrained level set active contour model†developed by Fast Technology, which detects the mesh filter and 3M test piece. Among them, Figures 6-a, 6-b, and 6-c are original images of the Sedoris filter, the 3M S. aureus test piece, and the 3 M E. coli test piece. Figures 6-d, 6-e, and 6-f are the segmentation effects after using the “level constraint active contour model based on morphological constraintsâ€. Due to the circular constraint, the active contour avoids the approximation of the mesh and eventually detects all round colonies.
Figure 6. Effect of level set active contour model based on morphological constraints
4 , outlook
The image segmentation method of the horizontal set active contour model has the advantages of strong anti-noise performance, good numerical solution stability, smooth and continuous segmentation boundary, and can handle complex topology. It has become one of the most advanced image segmentation techniques in the world.
After more than two years of research, the R&D team of science and technology not only mastered this advanced technology, but also based on the characteristics of microbial colonies, based on the traditional level set active contour model, creative research and development for the rapid counting of complex colony segmentation Active contour models, multi-phase level set active contour models, and level set active contour models based on morphological constraints. These models not only achieve accurate and effective statistics of complex colonies and difficult plates, but also are suitable for the detection of cells and the like.
Xunda Technology Co., Ltd. R & D Department Agrochemical And Intermediates
Refers to the inputs in agricultural production such as chemical fertilizers, Pesticides, veterinary drugs and growth regulators. Their use can promote the production of edible agricultural products and play an important role in the sustained and rapid development of agriculture.
Top Quality Baricitinib Baricitinib CAS 1187594-09-7,CAS 1187594-09-7
Shijiazhuang Dingmin pharmaceutical Sciences Co.,Ltd , https://www.dingminpharma.com