In order to determine the most appropriate treatment of a wound, the precise characteristics of the wound, including the size of the wounded area and its depth, and healing progress are the primary focuses of the physician [1, 2]. As the wound naturally recovers from its deepest part [3], the wound volume can, therefore, be used to indicate the wound recovery condition, as well as, how effective the treatment is. In Thailand, physicians mostly rely on manual approaches for measuring wound sizes, such as using rulers, transparent Visitrak films, and 2D camera [4]. Diagnosing the wound at its depth is generally avoided as direct contact between tools and the wound might cause inflammation. Unfortunately, there have been cases in which physicians decided to opt-out of the current treatment although it has been effective as the recovery was not visible on the surface.

Apart from not being able to diagnose the wound at depth, other challenges lie in the process of wound treatment. In terms of data management, data generated during the treatment processes are manually recorded on paper or separately saved as electronic files. Some operational parameters, such as the amount of lactated Ringer’s solution needed for the burns, are partly sight-estimated based on the wound surface area. Moreover, past treatment records are stored in various forms in different systems, thus making it difficult to compare past records to observe changes over time.

Our research and development attempts aim at inventing an integrated solution to facilitate the physician’s workflow of wound treatment. This covers the process of measuring wound volume, determining treatment, handling data electronically, and presenting associated data in proper forms. Our proposed solution integrates a newly invented machine that can accurately measure the wound size and volume, and a software that collects, analyses, and presents data essential in the treatment process. Specifically, the software also calculates certain operational parameters used in the treatment, instead of relying on sight estimation. The design of the wound measuring machine is presented in our previous work [5], while the software platform and user interface design can be found in a separate article [6].

This paper presents the manifestation of the proposed machine design, integrated successfully with the proposed software solution. To the best of our understanding, this is the first machine that allows contactless estimation of wound volume with satisfactory accuracy. The system was tested against simulated wounds of various shapes, both regular and irregular, in order to verify its measurement accuracy. The test showed that our system is approximately 90 percent accurate compared to Archimedes’ principle [7] for measuring object volume. The system was found to be useful for simplifying wound treatment workflow, especially in terms of contactless wound measurement, result accuracy, as well as data synthesis and visualization.

The paper is organized as follows. Section 2 outlines various traditional methods for wound measurement. Section 3 explains the digital image processing technique. Section 4 describes the design of our hardware. The next section discusses the software design and our approach to calculate the amount of lactated Ringer’s solution needed for the burn. Section 6 presents the workflow of “WoundCal” programs designed to estimate the area and volume of the wound followed by the experimental design to test the estimation accuracy and the results. The paper ends with conclusions and suggestions for future work.

To facilitate the effective application of the 3D scanner for estimating wound volume, a prototype machine was designed and improved, as described in this section.

3.2. The Improved Prototype

The second design removes the aforementioned constraint of the initial prototype. The scanner is attached to a half-circular rail, while the rail itself is attached to a servo motor [19] which allows it to move 360-degrees around the scanned object. Fig. (7) illustrates the hardware platform and the possible movement of the scanner. It is also possible to manually move the scanner through the rail to the desired position. The design has been built into a working prototype, as presented in Fig. (8).

Fig. (7). Second prototype machine, horizontal movement using motor control (15a) and vertical movement using manual adjustment (15b).

Fig. (8). The wound measuring machine.

The developed prototype is explained in the previous section. This section describes in detail the approach used to calculate the volume of the wound and how the test of accuracy was carried out. The 3-step procedure was as follows:

•Step 1: Developing wound models and creating their 3D-model

•Step 2: Creating very thin slices of the entire wound model

•Step 3: Wound size and volume estimation using a pixel reading method [22]

Step 1: Developing wound models and creating their 3D-model. Models of the wounds were developed to test the application of the prototype and its accuracy in measuring wound volume. Archimedes’ principle of liquid displacement was applied to calculate the base value of wound volume for comparison. An error rate of the machine-estimated values against the base values was identified.

Two types of wound models were developed. The first kind was made from molding clay, while the other was made from pork rind. To create a wound on the surface of pork rind, a small part of the flesh was simply removed from the unit of pork rind. Examples of wound models are displayed in Table 1, along with their 3D models acquired from the 3D scanner attached to the rail.

Step 2: Creating very thin slices of the entire wound model. The second step's input was the 3D model of the wound obtained from the small software module integrated with the prototype machine. The slicing technique was applied to prepare input for wound size volume estimation. Fig. (16) illustrates the collection of wound slices generated from the 3D models. The opensource program called Slic3r PE [23] was used to simulate the slicing model, which gave an image file for each wound slice. The slice thickness was set to 0.3 millimeters, which is the minimum value of thickness provided in Slic3r PE configuration. It is a fact that the flatter the slice, the more accurate the volume estimation. However, the thickness of 0.3 is considered sufficient, given the time required for the calculation and the level of accuracy acquired compared with the base practice. Nevertheless, the example of two consecutive wound slices is provided in Fig. (17) to illustrate the differences between the wounded areas.

Fig. (16). Wound slices generated from the 3D model.

Fig. (17). Two consecutive layers of wound slices with a slicing thickness of 0.3 millimeters.

Step 3: Wound size and volume estimation using a pixel reading method. The estimation was done using a program “WoundCal” which was specifically developed for this purpose. The program flowchart exhibiting 14 activities involved in this step is summarized in Fig. (19). The input of this step was the images of each sliced layer of the wound obtained from the previous step. The process is as follows:

(1) a user identifies the slice with the largest wound size and then roughly indicates the wounded area on the image file by replacing that area with a black pixel. This can be done by using any image editor software. Fig. (18) illustrates an example of a wound slice image marked with a black pixel that covers the wounded area.

Fig. (18). Wounded area marked with black pixels.

Fig. (19). Flowchart of the WoundCal program.

A critical analysis, interpretation, and evaluation of the obtained results are presented.

(2-3) upload the file from step (1) on WoundCal. The program transforms the image file to .bmp if it is not so.

(4) the program locates the area on the image that is marked with the black pixel from step (1). A two-dimensional (x, y) matrix is used to indicate the area. That is:

for every point of y,

identify the starting point and the ending point of x marked with the black pixel.

This marked area matrix is used as a reference as the program identifies the wounded area on each slice image of the wound.

(5) upload all the slice images of the wound on WoundCal

(6-7) the program calculates the size of the wounded area for each wound slice image. The wounded area can be identified by a white pixel as the result of the 3D scanner. The identified area is converted from pixels to centimeters (38 x 38 pixel is equal to 1 cm²⁾

(8-10) the program identifies the wound slice with the largest size of the wound and uses that value to represent the size of the wound. This value is entered manually into the wound treatment application, as presented in Fig. (20).

(11-14) The program calculates the wound volume as follows:

(5)

with i refers to the wound slice. Thickness is set to 0.3 as mentioned previously. The wound volume is also entered into the wound treatment application.

In the previous subsection, our approach to estimating the wound’s size and volume is explained in detail. The program “WoundCal” was developed to realize this approach. An experiment was then carried out for two primary objectives, firstly to measure how accurate the proposed method is compared to the estimation of wound volume based on the Archimedes’ principle of water displacement, and secondly, to assess usage limitation with respect to wound size.

Fig. (20). Interface of the WoundCal program.

6.2. Estimating Wound Size and Volume

Two methods were compared for estimating the size and volume of the wounds. The first one was our proposed method using the WoundCal program, while the other one was the base method using Archimedes’ principle of water displacement.

• Using our proposed WoundCal program. the process was as follows:

(1) Place the wound model inside the prototype machine for wound measurement, as shown in Fig. (16). The machine is connected to the computer where the 3D model and filename.stl can be saved as the input for the next step.

(2) Create slices of the wound 3D model using the Scli3e PE program. N image files of wound slices are generated.

(3) Identify the slice with the largest wounded area and roughly mark that area with a black pixel using any image editor program.

(4) Upload the marked image file to WoundCal.

(5) Upload the N image files of wound slices to WoundCal. This program calculates the volume of the wound.

•Using Archimedes’ principle. To estimate the wound volume based on Archimedes’ principle, the following steps were performed:

(1) weigh the wound model.

(2) fill water inside the wound.

(3) weigh the wound model again and then calculate the weight of filled water by water-filled model’s weight – initial weight. The weight of the water is equal to its volume since:

(6)

given that the density of the water is one. Therefore, the volume is the estimated wound’s volume.

The estimation of wound volume was carried out based on the two methods mentioned previously. The Archimedes’ principle was used as the base values for comparison. Fourteen models of wounds were made from molding clay and the three models made from pork rind were experimental subjects. Table 2 presents the results of wound volume estimation.

The results show that the volume estimation for small surfaced wounds, such as 1 x 1 cm², is more than 80% inaccurate. However, the WoundCal method provides acceptable volume estimation for bigger wounds. Wounds larger than a surface area of 1.5 x 1.5 cm² are estimated to have less than 15 percent error. It can also be noticed that the bigger the wound, the more accurate the estimation.

Our proposed system is one of the first that tried to estimate wound volume using a contactless method. This approach is applicable to the pork rinds model which, to a certain extent, simulates the nature of human skin. The error found in the estimation can be associated with two primary causes. The first one is related to the weigh scale used in the Archimedes’ principle that allows for a minimum scale of one gram. The second cause might be the precision of the 3D scanners in capturing wound details.

Table 1. Two types of wound models from (1) molding clay and (2) pork rind, and their 3D models obtained from the scanner.

Wound Mockup before Water Replacement	Wound Mockup after Water Replacement	3D Model(.stl)
Example1 1cm x 1cm x 1cm Weight 119 g	Example1 1cm x 1cm x 1cm Weight 120 g
Example2 1cm x 1cm x 2cm Weight 232 g	Example2 1cm x 1cm x 2cm Weight 233 g
Example3 1.5cm x 1.5cm x 1cm Weight 113 g	Example3 1.5cm x 1.5cm x 1cm Weight 115 g
Wound mockup before water replacement	Wound mockup after water replacement	3D Model(.stl)
Example4 2cm x 2cm x 1cm Weight 101 g	Example4 2cm x 2cm x 1cm Weight 103 g
Example5 2cm x 2cm x 2cm Weight 230 g	Example5 2cm x 2cm x 2cm Weight 233 g
Example6 3cm x 3cm x 1cm Weight 116 cm3	Example6 3cm x 3cm x 1cm Weight 121 cm3
Wound mockup before water replacement	Wound mockup after water replacement	3D Model(.stl)
Example7 3cm x 3cm x 2cm Weight 222 g	Example7 3cm x 3cm x 2cm Weight 231 g
Example 8 similar real wounds Weight 447 g	Example 8 similar real wounds Weight 497 g
Example 9 similar real wounds Weight 429 g	Example 9 similar real wounds Weight 486 g
Example 10 similar real wounds Weight 433 g	Example 10 similar real wounds Weight 490 g
Wound mockup before water replacement	Wound mockup after water replacement	3D Model(.stl)
Example 11 similar real wounds Weight 425 g	Example 11 similar real wounds Weight 489 g
Example 12 similar real wounds Weight 428 g	Example 12 similar real wounds Weight 478 g
Example 13 similar real wounds Weight 421 g	Example 13 similar real wounds Weight 570 g
Example 14 similar real wounds Weight 434 g	Example 14 similar real wounds Weight 546 g
Wound mockup before water replacement	Wound mockup after water replacement	3D Model(.stl)
Example 15 Pork rinds model 3cm x 3cm x 3 cm Weight 1085g	Example 15 Pork rinds model 3cm x 3cm x 3 cm Weight 1089g
Example 16 Pork rinds model 4cm x 4cm x 2cm Weight 1082 g	Example 16 Pork rinds model 4cm x 4cm x 2cm Weight 1088 g
Example 17 Pork rinds model 5cm x 5cm x 2cm Weight 1072 g	Example 17 Pork rinds model 5cm x 5cm x 2cm Weight 1082 g

With our goal to build a contactless system to support the wound treatment process, a prototype, integrating a machine and a software application that offers precise measurement of wound characteristics and handles wound healing progression, was created. Generally, a wound starts to heal from its bottom part, thus knowing the wound depth is vital for the treatment. Physicians typically insert a cotton-tip inside the wound to measure its depth. This may cause inflammation and increase the time to heal. A novel approach was proposed to estimate wound depth using a 3D model. A machine was built to control the stable movement of the 3D scanner over the wound using a servo motor. Then, the image processing technique was applied to calculate the wound’s area and volume. The method was tested with wound models of various sizes, made from molding clay and pork rinds. The accuracy of our approach was compared to the Archimedes’ principle of water displacement. The two methods were found to have less than 15 percent discrepancy. This level of accuracy is considered satisfactory, given that our work is among one of the first attempts to close this research gap.

There are certain limitations to be considered for future work. Firstly, different software components used in the image processing and estimation process can be integrated to enhance user experience. Secondly, it is possible to apply machine learning techniques to identify the wounded area on the wound image file. Lastly, it is suggested to create a cloud-based version of the software to allow easy access to the data and the application.