Method for the screening of child undernutrition

Abstract

Method for the screening of child undernutrition comprising a photograph taking step (101),a digital image generation step (102), a digital image reception step (103), a landmark configuration identification step (104) wherein a configuration of bi-dimensional landmarks and semi-landmarks defining the shape of a body part depicted in the digital image is identified by means of geometric morphometric techniques, a coordinate acquisition step (105) wherein the x, y coordinates of each landmark and semi-landmark in the configuration are obtained, a normalization step (106) wherein the x, y coordinates of each landmark and semi-landmark in the configuration are normalized thus obtaining a normalized landmark configuration, and a classification step (107) wherein the normalized landmark configuration is classified by using a classifier trained with training data previously collected from a training sample of children.

Description

[0001] DESCRIPTION

[0002] Method for the screening of child undernutrition

[0003] TECHNICAL FIELD

[0004] The present invention relates to methods and processing systems for the screening of child undernutrition.

[0005] PRIOR ART

[0006] According to the World Health Organization (WHO), malnutrition refers to "deficiencies, excesses, or imbalances in a person's intake of energy and / or nutrients”. Furthermore, there are 2 different types of malnutrition. One is “undernutrition”, which includes stunting (low height or length for age), wasting (low weight for height or length), underweight (low weight for age), and micronutrient deficiencies or insufficiencies (lack of essential vitamins and minerals). The other is “overweight”, obesity and diet-related noncommunicable diseases (such as heart disease, stroke, diabetes and cancer).

[0007] Considering that the nutritional status is a factor related to the development in the human body, then it must be paid extreme attention to the diet every child takes so that they are assured a proper growth process. However, according to the World Health Organization (WHO) “in 2020, 149 million children under 5 were estimated to be stunted, 45 million were estimated to be wasted and 38.9 million were overweight or obese”. And, even worse, “around 45% of deaths among children under 5 years are linked to undernutrition”. In this scenario, it is clear that many improvements need to be made in order to fight malnutrition worldwide.

[0008] The present patent application focuses on the wasting and stunting sub-forms of undernutrition in children. The first one, also called Acute Malnutrition, “is associated with higher risk of death if not treated properly”. Furthermore, stunting or Chronic Malnutrition, occurs when a person “has not had food of adequate quality and quantity and / or has had frequent or prolonged illnesses”. As a result, it is critical to identify a child with this condition early enough so that he or she can receive appropriate health care. Three different anthropometric techniques are commonly used to screen wasting undernutrition:

[0009] - detecting nutritional bilateral oedema in limbs by pressing the child's feet or hands with a finger for 10 seconds and checking whether the pressed surfaces retract or not. If oedema is found, the child needs immediate medical attention since it is considered wasted with medical complications,

[0010] - calculating the weight-for-height or length z-score (WHZ) by considering, for a given height or length (cm), the sample mean and sample standard deviation in weight (kg) from a reference sample (generally supplied by WHO). The following nutritional statuses are obtained: o optimal nutritional condition (ONC): -1 < WHZ < 1 , o risk of undernutrition (RIS): -2 <WHZ< -1, o moderate wasting or / acute malnutrition (MAM): -3 < WHZ < -2, and o severe wasting or / acute malnutrition (SAM): WHZ< -3, and

[0011] - measuring the mid-upper arm circumference or MLIAC (mm) mid-way between the tip of the elbow and the tip of the shoulder. To facilitate this task, there are paper bands that are typically coded into 3 or 4 colors (green, yellow, orange, red) and depending on which color the child’s arm falls within, the nutritional status is obtained: o green - ONC: MLIAC > 135 mm, o yellow - RIS: 125 mm < MLIAC < 135 mm, o orange - MAM: 115 mm < MLIAC < 125 mm, and o red - SAM: 115 mm > MLIAC.

[0012] The second type of undernutrition, chronic malnutrition, refers to prolonged episodes of inadequate nutrition leading to stunting. It is estimated by height or length-for-age z-score (HAZ) < -2 SD.

[0013] However, techniques based on anthropometrical measurements have some drawbacks when used in humanitarian operative contexts. On the one hand, the examinations are timeconsuming and rely on specific instruments that need to be properly calibrated (such as stadiometers, measuring tapes, calipers or scales). On the other hand, they require the patient's cooperation (by remaining still) and professional supervision in order to avoid measurement errors and thus inaccurate results. It is very important to keep in mind that any misclassification of nutritional status could be fatal for many (malnourished) children who do will not receive proper health care if their nutritional status is not identified properly. However, health workers working in communities in low- or middle-income countries are sometimes not adequately trained. It is therefore necessary to provide communities with easy-to-use alternative methods for screening undernutrition in children.

[0014] DISCLOSURE OF THE INVENTION

[0015] The object of the invention is to provide a method for the screening of child undernutrition, a processing system for the screening of child undernutrition, and a computer readable storage medium, as defined in the claims.

[0016] The method of the invention comprises a photograph taking step wherein a photograph of a body part of a child is taken by means of a camera, a digital image generation step wherein a digital image of the body part being photographed is generated, and a digital image reception step wherein a processing system receives the digital image of the body part. Then, the method comprises a landmark configuration identification step wherein a configuration of bi- dimensional landmarks and semi-landmarks defining the shape of the body part depicted in the digital image is identified by means of geometric morphometric techniques. Next, the method comprises a coordinate acquisition step wherein the x, y coordinates of each landmark and semi-landmark in the configuration are obtained, and a normalization step wherein the x, y coordinates of each landmark and semi-landmark in the configuration are normalized thus obtaining a normalized landmark configuration. Next, the method comprises a classification step wherein the normalized landmark configuration is classified by using a classifier trained with training data previously collected from a training sample of children in order to classify the child whose body part is depicted in the digital image as having undernutrition or not.

[0017] A second aspect of the invention relates to a processing system for the screening of child undernutrition from a digital image of a body part of the child. The processing system comprises an image receiving unit for receiving the digital image of the body part of the child, a landmark configuration identification unit for identifying a configuration of bi-dimensional landmarks and semi-landmarks defining the shape of the body part depicted in the digital image by means of geometric morphometric techniques, a coordinate acquisition unit for obtaining the x, y coordinates of each landmark and semi-landmark in the configuration identified by the landmark configuration identification unit, a normalization unit for normalizing the x, y coordinates of each landmark and semi-landmark in the configuration, thereby obtaining a normalized landmark configuration, and a classifier unit for classifying the normalized landmark configuration by using a classifier trained with training data previously collected from a training sample of children in order to classify the child whose body part is depicted in the digital image as having undernutrition or not.

[0018] A third aspect of the invention relates to a computer readable storage medium comprising computer-executable instructions that, when executed by a processing system, cause the processing system to perform the following steps of a method for the screening of child undernutrition from an image of a body part of the child: a digital image reception step wherein the processing system receives the digital image of the body part, a landmark configuration identification step wherein the processing system identifies a configuration of bi-dimensional landmarks and semi-landmarks defining the shape of the body part depicted in the digital image by means of geometric morphometric techniques, a coordinate acquisition step wherein the processing system obtains the x, y coordinates of each landmark and semi-landmark in the configuration, a normalization step wherein the processing system normalizes the x, y coordinates of each landmark and semi-landmark in the configuration thus obtaining a normalized landmark configuration, and a classification step wherein the processing system classifies the normalized landmark configuration by using a classifier trained with training data previously collected from a training sample of children in order to classify the child whose body part is depicted in the digital image as having undernutrition or not.

[0019] By means of the method, the processing system, and the computer readable storage medium of the invention, a way of accurately screening child undernutrition is provided in which the use of specific instruments that need to be properly calibrated is dispensed with. At the same time, the method of the invention does not require of medical professionals to carry out the screening. These facts are especially advantageous in remote hard-to-reach villages in developing countries, where undernutrition screening is particularly useful.

[0020] DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 shows a flow chart of the steps of an embodiment of the method for the screening of child undernutrition of the invention.

[0022] Figure 2 shows a person taking a photograph of a body part of a child to perform the method of the invention.

[0023] Figure 3a shows a landmark configuration of an arm of a child, and Figure 3b shows a landmark configuration of an arm of a different child obtained in an embodiment of the method of the invention.

[0024] Figure 4a shows the landmark configuration of Figure 3a, and Figure 4b shows the landmark configuration of Figure 4a once it has been scaled.

[0025] Figure 5a shows two landmark configurations, and Figure 5b shows the landmark configurations of Figure 5a once they have been translated.

[0026] Figure 6a shows two landmark configurations centered about the coordinate origin, and Figure 6b shows the landmarks configurations of Figure 6a once one of the landmark configurations has been normalized to the other by using full Ordinary Procrustes Analysis.

[0027] Figure 7a shows a plurality of landmarks configurations centered about the coordinate origin, and Figure 7b shows the landmark configurations of Figure 7a once they have been normalized by using a full Generalized Procrustes Analysis.

[0028] Figure 8 shows an embodiment of the processing system 500 of the invention.

[0029] Figure 9 shows the processing system 500 of Figure 8, wherein a template 218 is displayed.

[0030] Figure 10 shows a schematic depiction of an embodiment of the processing system 500.

[0031] Figure 11 shows a detailed schematic depiction of an embodiment of the processing system DETAILED DISCLOSURE OF THE INVENTION

[0032] Figure 1 shows in a flow chart the steps of the method 100 for the screening of child undernutrition of the invention.

[0033] The method 100 for the screening of child undernutrition of the invention comprises a photograph taking step 101 wherein a photograph of a body part of a child is taken by means of a camera, a digital image generation step 102 wherein a digital image of the body part being photographed is generated, and a digital image reception step 103 wherein a processing system 500 receives the digital image of the body part. Then, the method 100 comprises a landmark configuration identification step 104 wherein a configuration of bi-dimensional landmarks and semi-landmarks defining the shape of the body part depicted in the digital image is identified by means of geometric morphometric techniques. Next, the method 100 comprises a coordinate acquisition step 105 wherein the x, y coordinates of each landmark and semi-landmark in the configuration are obtained, and a normalization step 106 wherein the x, y coordinates of each landmark and semi-landmark in the configuration are normalized thus obtaining a normalized landmark configuration. Next, the method 100 comprises a classification step 107 wherein the normalized landmark configuration is classified by using a classifier trained with training data previously collected from a training sample of children, in order to classify the child whose body part is depicted in the digital image as having undernutrition or not.

[0034] The method 100 of the invention screens child undernutrition based on the existing differences in body shape among children with different nutritional statuses. In the context of the invention a child is considered to be a person under the age of twelve, more particularly between 6 and 59 monts of age (both included). In order to identify said differences in body shape, in the photograph taking step 101 , a photograph of at least one body part of the child to be screened is taken by means of a camera, said camera being preferably a digital camera. Figure 2 shows a person taking a photograph of a body part of a child according to the invention. The child whose body part is being photographed is considered an out-of-sample observation, this is, a sample out from the training sample of children used to train the classifier. Then, in the digital image generation step 102, the camera digitizes the photograph taken and generates the digital image of the body part photographed so that the image can be processed digitally in subsequent steps of the method 100. Then, in the digital image reception step 103, the processing system 500 receives said digital image. In at least one embodiment, the processing system 500 may include one or more processors 502, one or more memory elements 504, storage 506, a bus 508, one or more network processing units 510 interconnected with one or more network input / output (I / O) interfaces 512, one or more I / O interfaces 514, and a computer program. Figure 10 shows a schematic depiction of an embodiment of the processing system 500.

[0035] In an embodiment of the invention, the processing system 500 comprises the camera, and is configured to generate the digital image of the body part of the child being photographed and to store the digital image in the memory element 504, or in a file stored in the storage 506. In said embodiment, the computer program comprises computer-executable instructions that, when executed by the processor 502, cause the processor 502 to perform the steps of the method 100 of the invention.

[0036] However, in another embodiment of the invention, the processing system 500 and the camera are two different devices. In such an embodiment, the camera is configured to generate the digital image of the body part of the child being photographed. The digital image is stored in a file by the camera so that when the processing system 500 receives the digital image via the I / O interface, it stores the file in the storage 506. In said embodiment, the computer program 520 comprises computer-executable instructions that, when executed by the processor 502, cause the processor 502 to perform the steps of the method 100 of the invention from the digital image reception step 103 onward.

[0037] Geometric morphometric techniques are a collection of tools widely used for visualization and quantification of shape changes among biological organisms, wherein these organisms can be represented by bi- or three-dimensional cartesian points that summarize their shape information. In the method 100 of the invention, geometric morphometric techniques are used to analyze the shape of different parts of the body of children. Thus, in the context of the invention, landmarks are points in a bi-dimensional space that correspond to discrete anatomical locations that match between and within children. Once an initial list of landmarks are located, a second list of landmarks called semi-landmarks are located, relative to the initial landmarks’ positions, on areas which are hard to identify (like smooth curves and surfaces) with the purpose of completing the geometrical information given by the initial landmarks. In the context of the invention, the semi-landmarks are points in a bi-dimensional space that correspond to points located in the contour of the body part being analyzed, and which define, together with the landmarks, the shape of said body part. A landmark configuration or configuration is a group of landmarks and semi-landmarks summarizing the shape of a form which in the context of the invention is a body part of a child. A landmark configuration matrix e Mkxm(K) (the vector space of real matrices) is a real matrix containing the shape information of a form in k landmarks and semi-landmarks (rows) of dimension m (columns), this is, a real matrix containing the x, y coordinates of the landmarks and semi-landmarks of the configuration. In the context of the invention, the landmarks and semi-landmarks are bi- dimensional, therefore m is equal to 2. In the context of the invention, the term shape refers to the geometric properties of an object that are invariant with respect to location, scale, or orientation.

[0038] In an embodiment of the method 100, after receiving the digital image, the processing system 500 is configured to perform the landmark configuration identification step 104, wherein the processing system 500 identifies a configuration of bi-dimensional landmarks and semilandmarks defining the shape of the body part depicted in the digital image by means of geometric morphometric techniques.

[0039] Once the landmark configuration identification step 104 is performed, the processing system 500 is configured to perform the coordinate acquisition step 105 wherein the processing system 500 obtains the x, y coordinates of each landmark and semi-landmark in the configuration, i.e. , obtains the x coordinate and the y coordinate of each landmark and semilandmark in the configuration.

[0040] Once the coordinate acquisition step 105 is performed, the processing system 500 is configured to perform the normalization step 106 wherein the processing system 500 normalizes the x, y coordinates of each landmark and semi-landmark in the configuration thereby obtaining normalized x, y coordinates for each landmark and semi-landmark in the configuration. The normalized x, y coordinates of the landmarks and semi-landmarks of the configuration constitute a normalized landmark configuration. Photographs may have been taken from different distances from the body part, from different positions, and from different angles. As a result, the landmark configuration must be re-scaled, translated and rotated (as a whole) so that the body part represented by the landmark configuration is aligned with the landmark configuration summarizing the shape of the same body part of another child. By normalizing a landmark configuration, that is, by normalizing the x, y coordinates of the landmarks and semi-landmarks of the configuration, it becomes comparable to other normalized landmark configurations, and thus classifiable.

[0041] Finally, once the normalization step 106 is performed, the processing system 500 is configured to perform the classification step 107 wherein the processing system 500 classifies the normalized landmark configuration by using a classifier trained with training data previously collected from a training sample of children, in order to classify the child whose body part is depicted in the digital image as having undernutrition or not.

[0042] For training the classifier, a training sample of children whose nutritional status is known is selected. For each child, a photograph of a body part of the child is taken by using a camera, such that the body part being photographed is the same in all the children, and a digital image of said body part is generated. Said body part is the same as the body part photographed to the child being screened, this is, the out-of-sample observation that will be screened by means of the classifier. Each digital image is received by a training processing system 600. From each digital image received, the training processing system 600 identifies a configuration of bi- dimensional landmarks and semi-landmarks defining the shape of the body part depicted in the digital image by means of geometric morphometric techniques. Furthermore, the anatomical locations associated with the bi-dimensional landmarks of the configurations identified from the training sample of children and the anatomical locations associated with the bi-dimensional landmarks of the configuration identified for the out-of-sample observation, correspond to each other. In the same way, the bi-dimensional semi-landmarks of the configurations identified from the training sample of children correspond to the bi-dimensional semi-landmarks of the configuration identified for the child being screened, i.e., the out-of- sample observation being classified. Next, the training processing system 600 obtains the x, y coordinates of each landmark and semi-landmark in each configuration, i.e., the x coordinate and the y coordinate associated with each landmark and semi-landmark in each configuration. Finally, the training processing system 600 normalizes the x, y coordinates of each landmark and semi-landmark in each configuration thus obtaining normalized x, y coordinates for each landmark and semi-landmark in each configuration. Thus, for each child in the training sample, normalized x, y coordinates for each landmark and semi-landmark in the configuration identified from the digital image of his / her body part are obtained. The normalized x, y coordinates of the landmarks and semi-landmarks of a configuration constitute a normalized landmark configuration. Thus, the normalized landmark configurations, together with the information regarding the nutritional status of each child in said training sample constitute the training data or dataset with which the classifier is trained.

[0043] As it can be observed, the steps to obtain the normalized landmark configuration from an out- of-sample observation to be classified, or from a sample observation from the training sample of children used to train the classifier, are the same.

[0044] In at least one embodiment, the training processing system 600 may include one or more processors 502, one or more memory elements 504, storage 506, a bus 508, one or more network processing units 510 interconnected with one or more network input / output (I / O) interfaces 512, one or more I / O interfaces 514, and a computer program. In said embodiment, the computer program comprises computer-executable instructions that, when executed by the processor 502, cause the processor 502 to perform the steps necessary to train the classifier. Figure 10 shows a schematic depiction of an embodiment of the training processing system 600.

[0045] Figure 3a and 3b show each a landmark configuration of a body part of a different child obtained in an embodiment of the method 100 of the invention. Specifically, in said embodiment the body part is the left arm. However, in other embodiments of the method 100 of the invention, the body part depicted in the digital image can be a different limb, such as the right arm, the left leg, the right leg, or the trunk.

[0046] The landmark configurations depicted in Figure 3a and 3b comprise landmarks 1 , 2, 3, 4 and semi-landmarks 5, 6, 7, 8, 9, 10, 11 , 12, 13, 14, 15, 16, 17, 18, 19, 20. Specifically, landmark 1 corresponds to the lateral acromion process, landmark 2 corresponds to the elbow, more specifically to the joint between the lateral epicondyle of humerus bone and the head of radius, landmark 3 corresponds to the wrist, more specifically to the palmar midcarpal joint between the lunate and the capitate bones, and landmark 4 to the superior axillary fossa. In other embodiments of the method 100, a different number of landmarks and a different number of semi-landmarks ca be used. In another embodiment of the method 100 wherein the part of the body being photographed is an arm, five landmarks are used in the landmark configuration.

[0047] Figure 3a and 3b show the x, y coordinates of each landmark and semi-landmark in both landmark configurations in a bi-dimensional space. As it can be observed, depending on the distance from the body part from which the photograph has been taken, and also depending on the real size of the body part, the scale in the coordinates may vary. Additionally, photographs might have been taken from different positions and angles, which makes it necessary to normalize or standardize said x, y coordinates.

[0048] There are many known methods for normalizing a landmark configuration in the state of the art. An embodiment of a method for normalizing a landmark configuration is described below. In an embodiment of the method 100, in the normalization step 106, the x, y coordinates of each landmark and semi-landmark in the configuration are scaled, rotated and translated. As previously disclosed, a landmark configuration is represented by a landmark configuration matrix X e Mkxm(K) containing the shape information of a body part in k landmarks and semilandmarks (rows) (k being the number of landmarks and semi-landmarks in total) of dimension m (columns) (in the present invention, m being equal to 2).

[0049] A size measure of X is any positive real valued function g:X IR such that, for all a > 0: g(aX) = ag(X). There exist many possible functions that might be used as size measure. In an enbodiment of the method 100, the centroid size is used as the size measure. Let X e Mkxm(IK) be a landmark configuration matrix with k rows (k being the number of landmarks and semi-landmarks) and m dimensions (m=2 for bi-dimensional landmarks and semilandmarks), the centroid size s of X is given by: where X;* is the / -th row of X, i.e. the / -th landmark / semi-landmark is the vector of column means, this is, the centroid. Therefore, for scaling a landmark configuration matrix, it is divided by its centroid size in order to standarize it to size 1. Figure 4a shows a landmark configuration corresponding to an arm of a child, and Figure 4b the same landmark configuration once it has been scaled. As can be observed, the scales of x and y coordinates change due to the scalation process, however the shape remais the same. The centroid 21 of the landmark configuration is represented in Figure 4a, wherein the distance from the centroid 21 to landmark 2, and semi-landmarks 9,10,18 has also been represented by means of dashed lines.

[0050] Once landmark configurations are scaled, the next step is translating them. In an embodiment of the method 100, for translating landmark configurations a constant vector y = ( / i, ..., ym)Te IRmis added to each row of the landmark configuration matrixkxm(I ) ■ Figure 5a shows two landmark configurations, and Figure 5b shows the same landmark configurations once they have been translated such that the centroid of both landmark configurations is located in the coordinate origin, this is in coordinate (x,y) = (0,0).

[0051] Once landmark configurations are translated, the next step is rotating them such that the landmarks 1 are all close to each other, landmarks 2 are all close to each other, etc. In an embodiment of the method 100, for rotating landmark configurations, each landmark configuration matrix X e Mkxm(K) is multiplied by an appropriate rotation matrix r e Mmxm(IPi), i.e., XT. Rotation matrix r is an orthogonal matrix (i.e., and square matrix such that rTr = rrT= Im) satisfying that det(T) = +1. Such a rotation matrix, when m=2, which is the case for bi-dimensional landmarks and semi-landmarks, can be parametrized as: where -n < 9 < it, such that given a landmark configuration matrix X e Mkxm(R) centered about the origin of coordinates, then XT is the anticlockwise rotation of X 9 radians about such origin.

[0052] In the embodiment just disclosed, landmark configurations are normalized by the previously disclosed steps of scaling, translating, and rotating. In other embodiments of the method 100 of the invention, the normalization step 108 can be performed by executing the same steps of scaling, translating and rotating in a different way, or by any other techniques known in the state of the art.

[0053] By means of the method 100 of the invention, a way of accurately screening child undernutrition is provided in which the use of specific instruments that need to be properly calibrated is dispensed with. At the same time, the method 100 of the invention does not require of medical professionals to carry out the screening. These facts are especially advantageous in remote hard-to-reach villages in developing countries, where undernutrition screening is particularly necessary.

[0054] In an embodiment of the method 100 of the invention, the children to be screened are children between the ages of 6 and 59 months, both ages included. The method 100 of the invention is particularly advantageous when the children to be screened have an age within the aforementioned range. In such embodiment, the classifier used is a classifier previously trained with training data previously collected from a training sample of children between 6 and 59 months of age. In an embodiment of the method 100 of the invention, children are grouped in two age ranges, a first range for children between 6 and 24 months of age, and a second range for children from 24 to 59 months of age. In such an embodiment two different classifiers are used, a first classifier previously trained with training data previously collected from a training sample of children between 6 and 24 months of age, and a second classifier previously trained with training data previously collected from a training sample of children from 24 to 59 months of age, in such a way that children whose age is between 6 and 24 months are screened by using the first classifier, and those whose age is from 24 to 59 months are screened with the second one. Grouping children into these two groups makes the method 100 of the invention be particularly accurate.

[0055] In an embodiment of the method 100 of the invention children are grouped by biological sex. There exist differences in body shape between children of the same age but from different biological sex. The method 100 of the invention is particularly advantageous when different classifiers are used, one for one biological sex and another one for the other, such that each child is classified by using the classifier previously trained with training data previously collected from a training sample of children of his / her same biological sex.

[0056] In an embodiment of the method 100 of the invention children are grouped by population origin. There are differences in body shape between children of the same age but from different population origins, i.e., the body shape of children of the same age, and even of the same biological sex is different between Ethiopian children and Indian children, for example. The method 100 of the invention is particularly advantageous when different classifiers are used for different population origins, such that each child is classified by using the classifier previously trained with training data previously collected from a training sample of children from the same population origin.

[0057] In an embodiment of the method 100 of the invention, the method 100 is for screening wasting in children, preferably severe acute malnutrition. In another embodiment, the method 100 is for screening chronic malnutrition in children. By severe acute malnutrition is meant the nutritional status in which the weight-for-height z-score (WHZ) is < -3 and / or MLIAC < 115mm and / or a bilateral nutritional oedema is identified either in hands or feet. On the other hand, chronic malnutrition is estimated by height or length-for-age z-score (HAZ) < -2 SD. In an embodiment of the method 100, the x, y coordinates of each landmark and semilandmark in the configuration are normalized by means of a Procrustes analysis technique. In said embodiment, both, the x, y coordinates of the landmarks and semi-landmarks in the configuration associated with the out-of-sample observation, and also the x, y coordinates of the landmarks and semi-landmarks in the configurations associated with the training sample of children are normalized by means of a Procrustes analysis technique. Procrustes analysis techniques are known for providing different methods for shape alignment.

[0058] In an embodiment of the method 100, the Procrustes analysis technique is a full Ordinary Procrustes Analysis (full OPA), or a full Generalized Procrustes Analysis (full GPA).

[0059] In an embodiment of the method 100, a full Ordinary Procrustes Analysis (full OPA) is used to normalize the x, y coordinates of the landmarks and semi-landmarks of a landmark configuration. Full OPA provides a methodology to estimate the optimal scaling, translation and rotation parameters that must be applied to a landmark configuration to align or superimpose it to another landmark configuration (Dryden, l.L, Mardia, K. V., Statistical Shape Analysis: with Applications in R (2nd. Ed.) Wiley, 2016). When in the context of the invention it is said that a landmark configuration is registered to another landmark configuration, it means that said landmark configuration is aligned with, standardized with, or normalized with the other landmark configuration. Let1; 2e Mkxm(R) be two landmark configuration matrices, the method of full OPA performs least squares to estimate the similarity parameters y, F, and ft that minimize the distance: where ||. || is the Euclide lfc = (1,fe)Te > 0 is a scale parameter, F is an mxm rotation matrix and y is an mx1 location vector. The full OPA registration to X2is given by:

[0060] Xf ■■= px + lkyT, (eq. 4) where / ?, F, y are the parameters minimizing equation 3.

[0061] Figure 6a and 6b show an exemplary application of full OPA to align a landmark configuration to another landmark configuration. In Figure 6a, both landmark configurations are shown, both of them being centered about the coordinate origin, and in Figure 6b both landmark configurations are shown once one of the landmark configurations has been aligned to the other by using full OPA. In an embodiment of the method 100, a full Generalized Procrustes Analysis (full GPA) is used to normalize the x, y coordinates of the landmarks and semi-landmarks of a landmark configuration. Full GPA provides a methodology to estimate the optimal scaling, translation and rotation parameters that must be applied to a plurality of landmark configurations to standardize or normalize them (Rohlf, F.J., Slice, D., Extensions of the Procrustes method for the optimal superimposition of landmarks. Systematic Zoology, 1990). Let

[0062] Mkxm(K) be n landmark configuration matrices, the method of full GPA performs least squares in order to estimate the similarity parameters yh, th,fih, h = 1, ... , n, and ft that minimize the total sum of squares: where ft is the average shape obtained from the sample of n landmark configurations, and p the shape of the sample or population mean. The other parameters have the same meaning as in full OPA, with the difference that in full GPA as many sets of parameters as landmark configurations in the sample must be estimated. There are several constraints that can be applied. One of them is: where s is the centroid size. The full Procrustes coordinates of the h-th sample Xhfrom the sample of n landmark configurations are given by:

[0063] Xhp= fthXhrh+ lkyf, h = 1. n, (eq. 7) where fth> 0, th, yf, are the minimizing parameters in equation 5. For simplification purposes, ft can be represented as: i.e., the arithmetic mean of the Procrustes landmark configuration matrices, this is, the landmark configuration matrices once they have been normalized by means of a Procrustes analysis technique, where ftt= for all i,j. In the context of the invention, the term Procrustes coordinates is used to refer to the normalized x, y coordinates of the landmarks and semi-landmarks of a configuration that has been normalized by means of a Procrustes analysis technique.

[0064] Figure 7 shows an exemplary application of full GPA to align a plurality of landmark configurations. In Figure 7a the plurality of landmark configurations are shown, all of them centered about the coordinate origin, and in Figure 7b said landmark configurations are shown after full GPA has been performed, with the sample mean estimated in grey color. However, by using full GPA to normalize a group of landmark configurations, each time a new landmark configuration is incorporated to the group, it makes the normalized landmark configurations obtained after performing full GPA on them be different from the normalized landmark configurations that would have been obtained if the new landmark configuration would had not been incorporated to the group. Thus, in case the landmark configuration associated with the out-of-sample observation and the landmark configurations associated with the training sample of children were normalized by means of full GPA all together, the normalized landmark configurations obtained for the training sample of children would change every time a new landmark configuration associated to a new out-of-sample observation has to be normalized. This implies that, since the normalized landmark configurations associated with the training sample of children are part of the training data used to train the classifier, each time a new child is screened the classifier should be trained with the newly computed normalized landmark configurations, which results in high processing cost.

[0065] In an embodiment of the method 100, in the normalization step 106, the x, y coordinates of each landmark and semi-landmark in the configuration are normalized based on themselves and also based on a plurality of normalized landmark configurations associated with the training sample of children, such that the plurality of normalized landmark configurations associated with the training sample of children remain unchanged during the execution of the method 100.

[0066] In this embodiment of the method 100, when normalizing the x, y coordinates of each landmark and semi-landmark in the configurations associated with the training sample of children, the x, y coordinates of the landmarks and semi-landmarks in the configuration associated with the out-of-sample observation are not taken into account. Thus, the normalized x, y coordinates of the landmarks and semi-landmarks in the normalized landmark configurations associated with the training sample of children, once calculated, do not change during the performance of the method 100. However, when the normalization step 106 is performed on the landmark configuration associated with the out-of-sample observation, both, the x, y coordinates of the landmarks and semi-landmarks in said configuration together with the normalized x, y coordinates of the landmarks and semi-landmarks in the normalized landmark configurations associated with the training sample of children are taken into account. This makes the method 100 very advantageous, since it greatly reduces the capacity that the processing system 500 must have to carry out the method 100 of the invention, both in terms of memory, and processing capacity, as normalized landmark configurations associated with the training sample of children do not change once they are obtained.

[0067] In an embodiment of the method 100, in the normalization step 106, the x, y coordinates of each landmark and semi-landmark in the configuration are normalized to the mean of the plurality of normalized landmark configurations associated with the training sample of children.

[0068] In an embodiment of the method 100, wherein X , ...,Xn are the Procrustes coordinates of Xlt...,Xne Mkxm(R~), calculated by means of a full GPA, and Xn+1e Mkxm(R~) is an out-of- sample observation, the registration of Xn+1to the Procrustes sample mean shape is given by: are the parameters minimizing equation 3 when registering Xn+1to with full Ordinary Procrustes Analysis (full OPA), ft being the arithmetic mean of Xf, ...,x , as previously explaind before in this patent application when explaining the full GPA.

[0069] Xf, ...,Xn are the matrixes containing the normalized landmark configurations calculated by means of a full GPA on the landmark configurations associated with the training sample of children, which have been used to train the classifier, and Xn+1is the matrix containing the normalized landmark configuration associated with the out-of-sample observation to be screened. As previously explained, Xf, ..., x do not change while the method 100 is performed for the out-of-sample observation.

[0070] Thus, in this embodiment, the normalized landmark configuration associated with the out-of- sample observation is obtained by applying full OPA to the landmark configuration associated with the out-of-sample observation in respect of the mean of the normalized landmark configurations associated with the training sample of children, which have been obtained by applying full GPA to the landmark configurations associated with said training sample of children.

[0071] In an embodiment of the method 100, in the normalization step 106, the x, y coordinates of each landmark and semi-landmark in the configuration are normalized to the median of the plurality of normalized landmark configurations associated with the training sample of children.

[0072] In an embodiment of the method 100, Xf, ...,x are the Procrustes coordinates of X1;... ,Xne Mfexm(]R), calculated by means of a full GPA, and Xn+1£ Mtam(R) is an out-of-sample observation. In said embodiment / / ^ refers to the Procrustes coordinates of one of the samples in the training sample of children, such that the landmarks in the normalized landmark configuration (as a whole) associated with said Procrustes coordinates X / / o, this is, once they have been normalized by means of a full GPA, are the landmarks most surrounded within all normalized landmark configurations belonging to the training sample, i.e., X / / orefers to the sample median Procrustes configuration matrix. In other embodiments a different definition for the median Procrustes configuration is used. The registration of Xn+1to the Procrustes sample median shape is given by: are the parameters minimizing equation 3 when registering Xn+1to X / owith full OPA. As previously explained, Xf, ...,X / / do not change while the method 100 is performing the normalization step 106 for the out-of-sample observation.

[0073] Thus, in this embodiment, the normalized landmark configuration associated with the out-of- sample observation is obtained by applying full OPA to the landmark configuration associated with the out-of-sample observation in respect of the median of the normalized landmark configurations associated with the training sample of children, which have been obtained by applying full GPA to the landmark configurations associated with said training sample of children.

[0074] In an embodiment, the method 100 comprises a digital image validation step 1031 after the digital image reception step 103 and before the landmark configuration identification step 104, wherein after the processing system 500 receives the digital image of the body part being photographed, the digital image is validated by the processing system 500, such that in case of an invalid digital image, another photograph of the body part of the child is taken. By validating the digital image, the method 100 ensures that the received digital image is valid in that it fully displays a child’s body part, and that the body part is correctly positioned, which is of utmost importance so that the landmarks configuration associated with said body part is comparable to the landmark configuration associated with the body parts of other children.

[0075] In an embodiment of the method 100, in the photograph taking step 101 , the photograph of the body part of the child is taken with the child being in supine position. In said position, the children lie on their back, in a plane parallel to the ground, with the face and torso facing up. In an embodiment of the method 100, in the photograph taking step 101 , the body part of the child to be photographed is the left arm. After testing the method 100 with different body parts, it is the left arm that makes the method 100 particularly accurate.

[0076] In an embodiment of the method 100, in the photograph taking step 101 , the photograph of the left arm is taken with the palm of the hand facing the camera. After testing the method 100 with the left arm being photographed in different positions, by positioning the palm of the hand of the arm being photographed facing the camera, the screening provided by the method 100 of the invention is particularly accurate.

[0077] In an embodiment of the method 100, in the digital image validation step 1031 , it is checked that the angle between the arm and the forearm is in the range of 0° to 15°. In case a straight line is drawn aligned with the lower contour of the arm in the direction of the forearm, when the angle formed between said straight line and the lower contour of the forearm is in the range of 0° to 15°, a correct position of the left arm is ensured, which contributes to the screening provided by the method 100 of the invention being particularly accurate.

[0078] In an embodiment of the method 100 the processing system 500 comprises the camera, the processing system 500 being a mobile device, preferably a smartphone. The method 100 of the invention requires reduced processing and memory capacities, which makes it possible for mobile devices, such as smartphones, tablets, etc., to be suitable for carrying out the method 100 of the invention without connecting to remote servers to carry out some steps of the method 100. That makes the method 100 be particularly advantageous when it is carried out in underdeveloped places where connectivity barely exists, as the method 100 can be carried out in its entirety in a mobile device, preferably a smartphone or a tablet, without internet connection.

[0079] In an embodiment of the method 100, in the classification step 107, a probability is obtained that the child whose body part is depicted in the digital image has undernutrition. In an embodiment of the method 100 the classifier is trained to classify the child whose body part is depicted as having undernutrition or not. In other embodiments of the method 100, the classifier is trained to provide a probability that the child whose body part was depicted has undernutrition. In an embodiment of the method 100, the classifier is trained to classify the nutritional status of the child whose body part is depicted in the digital image in four different nutritional statuses: optimal nutritional condition (ONC), risk of undernutrition (RIS), moderate wasting or / acute malnutrition (MAM), and severe wasting or / acute malnutrition (SAM). In other embodiments, the classifier is trained to provide the probability that the child has each of the aforementioned four nutritional statuses.

[0080] In another embodiment of the method 100, the classifier is trained to classify the nutritional status of the child whose body part is depicted in the digital image in two different nutritional statuses: MAM, and SAM. In another embodiment, the classifier is trained to provide the probability that the child has each of the aforementioned two nutritional statuses.

[0081] In an embodiment of the method 100, in the landmark configuration identification step 104, the identification of the bi-dimensional landmarks and semi-landmarks is performed manually by a user. After the digital image is received by the processing system 500, the digital image is displayed on a screen of the processing system 500 such that the user can manually select the position in the digital image where the landmarks and semi-landmarks are located. In an embodiment of the method 100, the screen is a touch screen such that the user can identify the position of each landmark and semi-landmark by touching the screen. In another embodiment, the processing system 500 comprises a mouse such that the user can identify the position of each landmark and semi-landmark on the screen by using the mouse. In another embodiment, the processing system 500 comprises an input interface connected to an input device configured to allow a user to identify the points in the digital image associated with each landmark and semi-landmark on the screen.

[0082] In an embodiment of the method 100, in the landmark configuration identification step 104, the identification of the bi-dimensional landmarks and semi-landmarks is automatically done by the processing system 500. Methods for automatic identification of landmarks and semilandmarks are well known in the state of the art, and the method 100 of the invention is configured to perform such identification steps by using one of said methods.

[0083] In an embodiment, the method 100 of the invention comprises an allometry correcting step 1081 after the normalization step 106 and before the classification step 107, wherein the effect associated with allometry in the normalized landmark configuration is corrected. In the normalization step 106 location, scale and rotation of the landmark configurations are corrected, however, said normalization step 106 does not correct the possible effect that the actual size of for example an arm (in the reality) might have on its shape, i.e., the effect associated with allometry. More precisely, as children’s age varies, so does their body size and, hence, their body proportions. As a result, different arm shapes (in the case where an arm is being photographed) might be obtained (even for children belonging to the same nutritional status) owing to the growth process of children. For this reason, and to improve the performance of the classifier, the effect of allometry is corrected before the classification step. Methods for allometry correction are well known in the state of the art, and the method 100 of the invention is configured to perform the allometry correction effect 1081 by using one of said methods.

[0084] In an embodiment of the method 100, in the classification step 107, the classifier is a Linear Discriminant Analysis (LDA). The use of LDA as the classifier is particularly advantageous because it requires reduced processing and memory capacities, which allows the method 100 to be performed in its entirety in a mobile device, preferably a smartphone or a tablet, without and internet connection.

[0085] However, in other embodiments of the method 100 other supervised classification models can be used as the classifier in the classification step 107. In an embodiment, K-nearest neighbors (K-NN) classification algorithm is used. In another embodiment of the method 100, a decision tree classification algorithm is used.

[0086] A second aspect of the invention relates to a processing system 500 for the screening of child undernutrition from a digital image of a body part of the child, the processing system 500 comprising:

[0087] - an image receiving unit 522 for receiving the digital image of the body part of the child,

[0088] - a landmark configuration identification unit 524 for identifying a configuration of of bi- dimensional landmarks and semi-landmarks defining the shape of the body part depicted in the digital image by means of geometric morphometric techniques,

[0089] - a coordinate acquisition unit 526 for obtaining the x, y coordinates of each landmark and semi-landmark in the configuration identified by the landmark configuration identification unit 524,

[0090] - a normalization unit 528 for normalizing the x, y coordinates of each landmark and semilandmark in the configuration, thereby obtaining a normalized landmark configuration, and - a classifier unit 530 for classifying the normalized landmark configuration by using a classifier trained with training data previously collected from a training sample of children in order to classify the child whose body part is depicted in the digital image as having undernutrition or not.

[0091] In at least one embodiment, the processing system 500 may include one or more processors 502, one or more memory elements 504, storage 506, a bus 508, one or more network processing units 510 interconnected with one or more network input / output (I / O) interfaces 512, one or more I / O interfaces 514, and a computer program. Figure 11 shows a detailed schematic depiction of an embodiment of the processing system 500.

[0092] The computer program comprises a plurality of instructions which, when executed by the processor 502, cause the processor 502 to execute the steps of the method 100 of the invention. In at least one embodiment, the processor or processors 502 are at least a hardware processor configured to execute various tasks, operations and / or functions for the processing system 500 according to the software and / or the instructions configured for the processing system 500, for example, in the computer program.

[0093] In at least one embodiment, the memory element 504 and / or storage 506 are configured to store data, information, software, and / or instructions associated with the processing system 500, and / or the logic configured for the memory element 504 and / or storage 506. In an embodiment of the processing system 500, the computer program is stored in any combination of memory element(s) 504 and / or storage 506.

[0094] In at least one embodiment, the bus 508 can be configured as an interface that enables one or more elements of the processing system 500 to communicate with each other so as to exchange information and / or data. The bus 508 can be implemented with any architecture designed for exchanging control, data, and / or information between processors, memory elements / storage, peripheral devices, and / or any other hardware and / or software component that may be configured for the processing system 500. In at least one embodiment, the bus 508 may be implemented as a fast kernel-hosted interconnect, potentially using shared memory between processes (for example, logic), which can enable efficient communication paths between processes.

[0095] In several embodiments, the network processor unit(s) 510 may enable communication between the processing system 500 and other systems, entities, etc., through the network I / O interface(s) 512 (wired and / or wireless). In several embodiments, the network processor unit(s) 510 can be configured as a combination of hardware and / or software, as one or more Ethernet drivers and / or controllers or interface cards, fibre channel (e.g., optical) driver(s) and / or controller(s), wireless receivers / transmitters / transceivers, baseband processor(s) / modem(s) and / or other similar network interface driver(s) and / or controller(s) that are known now or may be developed hereinafter so as to enable communications between the processing system 500 and other systems, entities, etc., to facilitate the operations for the various embodiments of the method 100 described herein. In several embodiments, the network I / O interface(s) 512 can be configured as one or more Ethernet ports, fibre channel ports, any other I / O port(s) and / or antennas / antenna array that are known now or may be developed in the future. Therefore, the network processor unit(s) 510 and / or the network I / O interface(s) 512 may include suitable interfaces for receiving, transmitting, and / or otherwise communicating data and / or information in a network environment.

[0096] I / O interfaces 514 allow the input and output of data and / or information with other entities which may be connected to the processing system 500. For example, the I / O interfaces 514 may provide a connection to external devices such as a keyboard, numerical keypad, a touch screen, and / or any other suitable input and / or output device that is known now or may be developed in the future. In some instances, the external devices can also include (non- transitory) computer-readable storage media such as database systems, USB memories, portable optical or magnetic discs and memory cards. In still some instances, the external devices can be a mechanism for displaying data to a user, such as a computer monitor, a display screen, or the like.

[0097] In several embodiments, the computer program can include instructions which, when executed, cause the processor or processors 502 to perform operations, which can include, among others, providing overall control operations of the processing system 500, interacting with other entities, systems, etc. described herein, maintaining and / or interacting with stored data, information, parameters, etc. (for example, memory element(s), storage, data structures, databases, tables, etc.); combinations thereof; and / or the like so as to allow the execution of the operations necessary for the implementation of the method 100 of the invention.

[0098] In some cases, the computer program of the present embodiments can be available via a non- transitory computer-usable storage medium (for example, magnetic or optical media, magneto- optical media, CD-ROM, DVD, memory devices, etc.). In some cases, the non-transitory computer-readable storage media can also be removable. Other examples may include optical and magnetic discs, USB memories, and smart cards which can be inserted into and / or otherwise connected to a computer to be transferred to another computer-readable storage medium.

[0099] In a first embodiment of the processing system 500, the image receiving unit 522 is configured for receiving the digital image of the body part of the child by means of a network I / O interface 512. In said first embodiment, the processing system 500 is connected locally or remotely to a computer, such that the digital image, which has been generated or is stored at the computer, is provided by a network to the image receiving unit 522. In an embodiment the computer is locally connected to the processing system 500, but in another embodiment said computer is remotely connected to the processing system 500.

[0100] In a second embodiment of the processing system 500, the image receiving unit 522 is configured for receiving the digital image of the body part of the child by means of an I / O interface 514. In said second embodiment, a device is connected to the processing system 500 by means of the I / O interface 514. Said device can be a digital camera, which stores the digital image and provides it to the image receiving unit 522, or it can be a non-transitory computer-usable storage medium that stores said digital image and provides it to the image receiving unit 522.

[0101] The processing system 500 of the invention is configured to perform the steps of the method 100 for the screening of child undernutrition of the invention. However, as it can be observed, in some embodiments of the processing system 500, as in the first embodiment and in the second embodiment of the invention, said processing system 500 is only configured to perform the steps of the method 100 of the invention from the digital image reception step 103 on. This is the case where the photograph taking step 101 and the digital image generation step 102 are not performed by the processing system 500, but in a remote computer or in a device, as disclosed in the first and in the second embodiments of the processing system 500. In said embodiments, the processing system 500 is a computer.

[0102] The image receiving unit 522 is configured for receiving the digital image of the body part of the child and providing it to the landmark configuration identification unit 524, which is configured for identifying a configuration of bi-dimensional landmarks and semi-landmarks defining the shape of the body part depicted in the digital image received from the image receiving unit 522. The landmark configuration identification unit 524 is configured for providing the identified landmark configuration to the coordinate acquisition unit 526, which is configured for obtaining the x, y coordinates of each landmark and semi-landmark in the configuration provided by the landmark configuration identification unit 524, and providing said x, y coordinates of each landmark and semi-landmark in the configuration to the normalization unit 528. The normalization unit 528 is configured for normalizing the x, y coordinates of the landmarks and semi-landmarks in the configuration provided by the coordinate acquisition unit 526 and obtaining a normalized landmark configuration. The normalization unit 528 is configured for providing the normalized landmark configuration to the classifier unit 530 which is configured for classifying the normalized landmark configuration by using a classifier previously trained with training data previously collected from a training sample of children, in order to classify the child whose body part is depicted in the digital image as having undernutrition or not.

[0103] The computer program can include instructions which, when executed, cause the processor or processors 502 to perform operations, which can include, among others, the image receiving unit 522 receiving the digital image of the body part of the child and providing it to the landmark configuration identification unit 524, the landmark configuration identification unit 524 identifying a configuration of bi-dimensional landmarks and semi-landmarks defining the shape of the body part depicted in the digital image and providing the identified landmark configuration to the coordinate acquisition unit 526, the coordinate acquisition unit 526 obtaining the x, y coordinates of each landmark and semi-landmark in the configuration and providing them to the normalization unit 528, the normalization unit 528 normalizing the landmark configuration so as to obtain a normalized landmark configuration and provide it to the classifier unit 530, and the classifier unit 530 classifying the normalized landmark configuration by using the classifier, among others. This is, said operations provide the different units 522, 524, 526, 528, 530 in the processing system 500 with the ability to interact among them, or with other entities described herein, such as interacting with stored data, information, parameters, etc. stored in memory element(s) 504, storage 506, etc., and / or the like so as to allow the execution of the operations necessary for the implementation of the method 100 of the invention.

[0104] In an embodiment of the processing system 500 for the screening of child undernutrition the children to be screened are between 6 and 59 months of age. In an embodiment, the processing system 500 is for screening wasting in children, preferably severe acute malnutrition. In another embodiment, the processing system 500 is for screening chronic malnutrition.

[0105] In an embodiment of the processing system 500, the normalization unit 528 is configured for normalizing the x, y coordinates of each landmark and semi-landmark in the configuration by means of a Procrustes analysis technique.

[0106] In an embodiment of the processing system 500, the Procrustes analysis technique is a full Ordinary Procrustes Analysis, or a full Generalized Procrustes Analysis.

[0107] In an embodiment of the processing system 500, the normalization unit 528 is configured for normalizing the x, y coordinates of each landmark and semi-landmark in the configuration based on themselves and also based on a plurality of normalized landmark configurations associated with the training sample of children, such that the plurality of normalized landmark configurations associated with the training sample of children remain unchanged.

[0108] In an embodiment of the processing system 500, the normalization unit 528 is configured for normalizing the x, y coordinates of each landmark and semi-landmark in the configuration to the mean of the plurality of normalized landmark configurations associated with the training sample of children.

[0109] In an embodiment of the processing system 500, the normalization unit 528 is configured for normalizing the x, y coordinates of each landmark and semi-landmark in the configuration to the median of the plurality of normalized landmark configurations associated with the training sample of children.

[0110] In an embodiment of the processing system 500, the processing system 500 comprises a validation unit 523 configured for validating the digital image.

[0111] In an embodiment of the processing system 500, the processing system 500 is configured for screening child undernutrition from the digital image of the body part of the child, the child being in supine position when the photograph is taken. In an embodiment of the processing system 500, the processing system 500 is configured for screening child undernutrition from the digital image of the left arm of the child.

[0112] In an embodiment of the processing system 500, the processing system 500 is configured for screening child undernutrition from the digital image of the left arm of the child, the palm of the hand facing the camera when the photograph is taken.

[0113] In an embodiment of the processing system 500, the validation unit 523 is configured for checking that the angle between the arm and the forearm is in the range of 0° to 15°.

[0114] In an embodiment of the processing system 500, the processing system 500 comprises a camera configured for taking a photograph of a body part of a child and generating a digital image of the body part being photographed, said digital image being accessible to the image receiving unit 522. In said embodiment, the processing system 500 is configured for performing the photograph taking step 101 and the digital image generation step 102 of the method 100 of the invention, unlike in the first and second embodiments of the processing device 500, where the processing device 500 is configured for performing the steps of the method 100 starting from the digital image reception step 103.

[0115] In an embodiment of the processing system 500, the processing system 500 is a mobile device, preferably a smartphone. This makes the processing system 500 of the invention be particularly advantageous, as the method 100 can be carried out in its entirety in a mobile device, preferably a smartphone or a tablet, without internet connection. A mobile device can easily be carried to underdeveloped places where connectivity barely exists, as the method 100 can be carried out in its entirety in a mobile device, preferably a smartphone or a tablet, without internet connection. The method 100 of the invention requires reduced processing and memory capacities, which makes it possible for mobile devices, such as smartphones, tablets, etc., to be suitable for carrying out the method 100 of the invention without connecting to remote servers to carry out some steps of the method 100.

[0116] In an embodiment of the processing system 500, the processing system 500 comprises camera positioning means 210. The camera positioning means 210 are configured to provide the camera user with the necessary information to position the camera correctly, so that the digital image obtained from the photograph is correct and allows a correct screening to be made. The camera positioning means 210 enable the screening of undernutrition in children to be carried out using a single mobile device. This eliminates the need to carry other physical camera positioning devices to remote, hard-to-reach villages in developing countries, which is often impossible due to their size and weight, to ensure that the image taken is valid for the invention and allows a correct screening to be made.

[0117] In an embodiment of the processing system 500, the camera positioning means 210 comprise a screen 211. In said embodiment, the necessary information to position the camera correctly is provided to the user on the screen 211 .

[0118] In an embodiment of the processing system 500, the camera positioning means 210 comprise a first ruler 212 and a second ruler 214, and a first tilt pointer 213 and a second tilt pointer 215 displayed on the screen 211 , the first tilt pointer 213 being configured to slide up and down on the first ruler 212 as the tilt of the camera in the direction of the first ruler 212 changes, and the second tilt pointer 215 being configured to slide to the right and to the left on the second ruler 214 as the tilt of the camera in the direction of the second ruler 214 changes. As an illustrative example, when the user of the camera takes a photograph of a child in a supine position, lying on their back on the floor, with their face and torso facing upwards, it is necessary for the camera to face the floor perpendicularly in order for the photograph to be valid. Figure 8 shows an embodiment where the processing system 500 is a smartphone. In said embodiment, in order for the photograph taken with the smartphone to be valid, it is necessary for the smartphone to be positioned parallel to the ground. In such a situation, when the user tilts the smartphone to the right or to the left, the second tilt pointer 215 moves to the right or to the left on the second ruler 214, such that when the second tilt pointer 215 is located at the center of the second ruler 214, it indicates that the smartphone is correctly positioned in that direction. By means of the first tilt pointer 213, which moves up and down on the first ruler 212 as the user rotates the smartphone forward or backward, the user knows whether the smartphone is correctly positioned in that direction, such that when the smartphone is just parallel to the ground the first tilt pointer 213 and the second tilt pointer 215 are both in the center of the first ruler 212 and the second ruler 214 respectively.

[0119] In an embodiment of the processing system 500, the camera positioning means 210 comprises a first position reference 216, and a second position reference 217 displayed on the screen 211 , the first position reference 216 being configured to display on the screen 211 the location where a first reference of the child’s body part is to be placed, and the second position reference 217 being configured to display on the screen 211 the location where a second reference of the child’s body part is to be placed, so that the digital image obtained when taking the photograph is valid for screening the child’s undernutrition. Figure 8 shows an embodiment where the processing system 500 is a smartphone. When the user is using the smartphone, the smartphone is configured to display on the screen what is in front of the camera, and what is to be photographed when the user takes the photograph. By means of the first position reference 216 and the second position reference 217, the camera positioning means 210 assist the user by indicating on the screen where the anatomical locations of the child’s body part associated with the first position reference 216 and to the second position reference 217 are to be placed. Depending on the part of the body being photographed to screen child undernutrition, the first position reference 216 and to the second position reference 217 can be display at different places on the screen 11.

[0120] In an embodiment of the processing system 500, the first position reference 216 is a hand position reference 216, and the second position reference 217 is an armpit position reference 217. Figure 8 shows an embodiment where the processing system 500 is a smartphone, and the body part being photographed to screen child’s undernutrition is an arm. In this embodiment the hand position indicator 216 is configured to display on the screen 211 the location where the child’s hand is to be placed, and the armpit position indicator 217 is configured to display on the screen 211 the location at which the armpit is to be placed, so that the digital image obtained when taking the photograph is valid for screening the child’s undernutrition. In this embodiment the hand position indicator216 is a circle in which the child’s hand is to be placed, and the armpit position indicator 217 is two straight and perpendicular lines intersecting at a point, said point indicating the position in which the child’s armpit is to be placed. In other embodiments, the camera positioning means may be implemented in a different manner.

[0121] In an embodiment of the processing system 500, the camera positioning means 210 comprises a template 218 configured to display on the screen 211 the location where the body part to be photographed is to be placed. In the embodiment of Figure 9, the template 218 limits an area of the screen 211 where the arm is to be placed so that the digital image obtained when photographing is valid for the invention.

[0122] In an embodiment of the processing system 500, the processing system 500 comprises a correcting unit 529 configured for correcting the effect associated with Allometry in the normalized landmark configuration. In an embodiment of the processing system 500, the classifier unit 530 is configured for providing a probability of the child whose body part is depicted in the digital image having undernutrition.

[0123] In an embodiment of the processing system 500, the classifier is a Linear Discriminant Analysis.

[0124] A third aspect of the invention relates to a computer readable storage medium comprising computer-executable instructions that, when executed by a processing system 500, cause the processing system 500 to perform the following steps of a method for the screening of child undernutrition from an image of a body part of the child: a digital image reception step 103 wherein the processing system 500 receives the digital image of the body part, a landmark configuration identification step 104 wherein the processing system 500 identifies a configuration of bi-dimensional landmarks and semi-landmarks defining the shape of the body part depicted in the digital image by means of geometric morphometric techniques, a coordinate acquisition step 105 wherein the processing system 500 obtains the x, y coordinates of each landmark and semi-landmark in the configuration, a normalization step 106 wherein the processing system 500 normalizes the x, y coordinates of each landmark and semi-landmark in the configuration thus obtaining a normalized landmark configuration, and a classification step 107 wherein the processing system 500 classifies the normalized landmark configuration by using a classifier trained with training data previously collected from a training sample of children in order to classify the child whose body part is depicted in the digital image as having undernutrition or not.

[0125] The computer readable storage medium of the invention comprises computer-executable instructions that, when executed by the processing system 500 of the invention, cause the processing system 500 to perform the steps of the method 100 of the invention.

[0126] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to perform the steps of the method 100 of the invention from the digital image reception step 103 on. This is the case where the photograph taking step 101 and the digital image generation step 102 of the method 100 of the invention are not performed by the processing system 500, but in a remote computer or in a device, as disclosed in the first and in the second embodiments of the processing system 500 of the invention.

[0127] In an embodiment of the computer readable storage medium wherein the processing system 500 is a mobile device comprising a camera, preferably a smartphone, the computerexecutable instructions, when executed by the processing system 500, cause the processing system 500 to perform a photograph taking step 101 wherein a photograph of a body part of a child is taken by means of the camera, and a digital image generation step 102 wherein a digital image of the body part being photographed is generated, before the digital image reception step 103.

[0128] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to screen child undernutrition in children between 6 and 59 months of age.

[0129] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to screen wasting, preferably severe acute malnutrition, or chronic malnutrition.

[0130] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to normalize the x, y coordinates of each landmark and semi-landmark in the configuration by means of a Procrustes analysis technique in the normalization step 106.

[0131] In an embodiment of the computer readable storage medium, the Procrustes analysis technique is a full Ordinary Procrustes Analysis, or a full Generalized Procrustes Analysis.

[0132] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to normalize the x, y coordinates of each landmark and semi-landmark in the configuration based on themselves and also based on a plurality of normalized landmark configurations associated with the training sample of children, such that the plurality of normalized landmark configurations associated with the training sample of children remain unchanged during the normalization step 106d.

[0133] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to normalize the x, y coordinates of each landmark and semi-landmark in the configuration to the mean of the plurality of normalized landmark configurations associated with the training sample of children.

[0134] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to normalize the x, y coordinates of each landmark and semi-landmark in the configuration to the median of the plurality of normalized landmark configurations associated with the training sample of children.

[0135] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to perform a validation step 1031 after the digital image reception step 103 and before the landmark configuration identification step 104, wherein the processing system 500 validates the digital image, such that in case of an invalid digital image, another photograph of the body part of the child is taken.

[0136] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to screen child undernutrition from the digital image of the body part of the child, the child being in supine position when the photograph is taken.

[0137] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to screen child undernutrition from the digital image of the left arm of the child.

[0138] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to screen child undernutrition from the digital image of the left arm of the child, the palm of the hand facing the camera when the photograph is taken. In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to perform a digital image validation step 1031 after the digital image reception step 103 and before the landmark configuration identification step 104, wherein the processing system 500 validates the digital image of the body part.

[0139] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to check that the angle between the arm and the forearm is in the range of 0° to 15° in the digital image validation step 104.

[0140] In an embodiment of the computer readable storage medium, the computer-executable instructions, when executed by the processing system 500, cause the processing system 500 to provide a probability that the child whose body part is depicted in the digital image has child undernutrition in the classification step 107.

[0141] In an embodiment of the computer readable storage medium, wherein the computerexecutable instructions, when executed by the processing system 500, cause the processing system 500 to perform an allometry correcting step 1061 after the normalization step 106 and before the classification step 107, wherein the processing system 500 corrects the effect associated with allometry in the normalized landmark configuration.

[0142] In an embodiment of the computer readable storage medium, the classifier is a Linear Discriminant Analysis.

Claims

CLAIMS1 . Method for the screening of child undernutrition comprising: a photograph taking step (101) wherein a photograph of a body part of a child is taken by means of a camera, a digital image generation step (102) wherein a digital image of the body part being photographed is generated, a digital image reception step (103) wherein a processing system (500) receives the digital image of the body part, a landmark configuration identification step (104) wherein a configuration of bi- dimensional landmarks and semi-landmarks defining the shape of the body part depicted in the digital image is identified by means of geometric morphometric techniques, a coordinate acquisition step (105) wherein the x, y coordinates of each landmark and semi-landmark in the configuration are obtained, a normalization step (106) wherein the x, y coordinates of each landmark and semilandmark in the configuration are normalized thus obtaining a normalized landmark configuration, and a classification step (107) wherein the normalized landmark configuration is classified by using a classifier trained with training data previously collected from a training sample of children, in order to classify the child whose body part is depicted in the digital image as having undernutrition or not.

2. Method according to claim 1 , wherein the children are between 6 and 59 months of age.

3. Method according to claim 1 or 2, wherein the method (100) is for the screening of severe acute malnutrition or for the screening of chronic malnutrition.

4. Method according to any of the preceding claims, wherein in the normalization step (106), the x, y coordinates of each landmark and semi-landmark in the configuration are normalized by means of a Procrustes analysis technique.

5. Method according to claim 4, wherein the Procrustes analysis technique is a full Ordinary Procrustes Analysis, or a full Generalized Procrustes Analysis.

6. Method according to any of the preceding claims, wherein in the normalization step (106), the x, y coordinates of each landmark and semi-landmark in the configuration are normalized based on themselves and also based on a plurality of normalized landmark configurations associated with the training sample of children, such that the plurality of normalized landmark configurations associated with the training sample of children remain unchanged during the execution of the method (100).

7. Method according to claim 6, wherein in the normalization step (106) the x, y coordinates of each landmark and semi-landmark in the configuration are normalized to the mean of the plurality of normalized landmark configurations associated with the training sample of children.

8. Method according to claim 6, wherein in the normalization step (106) the x, y coordinates of each landmark and semi-landmark in the configuration are normalized to the median of the plurality of normalized landmark configurations associated with the training sample of children.

9. Method according to any of the preceding claims, comprising a digital image validation step (1031) after the digital image reception step (103) and before the landmark configuration identification step (104), wherein the digital image is validated by the processing system (500), such that in case of an invalid digital image, another photograph of the body part of the child is taken.

10. Method according to any of the preceding claims, wherein in the photograph taking step (101) the photograph of the body part of the child is taken with the child being in supine position.

11. Method according to any of the preceding claims, wherein in the photograph taking step (101), the body part of the child to be photographed is the left arm.

12. Method according to claim 11 , wherein in the photograph taking step (101), the photograph of the left arm is taken with the palm of the hand facing the camera.

13. Method according to claim 11 or 12, wherein in the digital image validation step (1031) it is checked that the angle between the arm and the forearm is in the range of 0° to 15°.

14. Method according to any of the preceding claims, wherein the processing system (500) comprises the camera, the processing system (500) being a mobile device, preferably a smartphone.

15. Method according to any of the preceding claims, wherein in the classification step (107), a probability is obtained that the child whose body part is depicted in the digital image has undernutrition.

16. Method according to any of the preceding claims, wherein in the landmark configuration identification step (104) the identification of the bi-dimensional landmarks and semilandmarks is performed manually by a user.

17. Method according to any of claims 1 to 15, wherein in the landmark configuration identification step (104), the identification of the bi-dimensional landmarks and semilandmarks is automatically done by the processing system (500).

18. Method according to any of the preceding claims, comprising an allometry correcting step (1061) after the normalization step (106) and before the classification step (107), wherein the effect associated with allometry in the normalized landmark configuration is corrected.

19. Method according to any of the preceding claims, wherein in the classification step (107) the classifier is a Linear Discriminant Analysis.

20. Processing system for the screening of child undernutrition from a digital image of a body part of the child, the processing system (500) comprising:- an image receiving unit (522) for receiving the digital image of the body part of the child,- a landmark configuration identification unit (524) for identifying a configuration of bi- dimensional landmarks and semi-landmarks defining the shape of the body part depicted in the digital image by means of geometric morphometric techniques,- a coordinate acquisition unit (526) for obtaining the x, y coordinates of each landmark and semi-landmark in the configuration identified by the landmark configuration identification unit (524),- a normalization unit (528) for normalizing the x, y coordinates of each landmark and semi-landmark in the configuration, thereby obtaining a normalized landmarkconfiguration, and- a classifier unit (530) configured for classifying the normalized landmark configuration by using a classifier trained with training data previously collected from a training sample of children in order to classify the child whose body part is depicted in the digital image as having undernutrition or not.

21. Processing system according to claim 20, wherein the children are between 6 and 59 months of age.

22. Processing system according to claim 20 or 21 , wherein the processing system (500) is for screening severe acute malnutrition or chronic malnutrition.

23. Processing system according to any of claims 20 to 22, wherein the normalization unit (528) is configured for normalizing the x, y coordinates of each landmark and semilandmark in the configuration by means of a Procrustes analysis technique.

24. Processing system according to claim 23, wherein the Procrustes analysis technique is a full Ordinary Procrustes Analysis, or a full Generalized Procrustes Analysis.

25. Processing system according to any of claims 20 to 24, wherein the normalization unit (528) is configured for normalizing the x, y coordinates of each landmark and semilandmark in the configurations based on themselves and also based on a plurality of normalized landmark configurations associated with the training sample of children, such that the plurality of normalized landmark configurations associated with the training sample of children remain unchanged.

26. Processing system according to claim 25, wherein the normalization unit (528) is configured for normalizing the x, y coordinates of each landmark and semi-landmark in the configuration to the mean of the plurality of normalized landmark configurations associated with the training sample of children.

27. Processing system according to claim 25, wherein the normalization unit (528) is configured for normalizing the x, y coordinates of each landmark and semi-landmark in the configuration to the median of the plurality of normalized landmark configurations associated with the training sample of children.

28. Processing system according to any of claims 20 to 27, comprising a validation unit (523) configured for validating the digital image.

29. Processing system according to any of claims 20 to 28, wherein the processing system (500) is configured for screening child undernutrition from the digital image of the body part of the child, the child being in supine position when the photograph is taken.

30. Processing system according to any of claims 20 to 29, wherein the processing system (500) is configured for screening child undernutrition from the digital image of the left arm of the child.

31. Processing system according to claim 30, wherein the processing system (500) is configured for screening child undernutrition from the digital image of the left arm of the child, the palm of the hand facing the camera when the photograph is taken.

32. Processing system according to claim 31 , wherein the validation unit (523) is configured for checking that the angle between the arm and the forearm is in the range of 0° to 15°.

33. Processing system according to any of claims 20 to 32, comprising a camera configured for taking a photograph of a body part of a child and generating a digital image of the body part being photographed, said digital image being accessible to the image reception unit (522).

34. Processing system according to claim 33, wherein the processing system (500) is a mobile device, preferably a smartphone.

35. Processing system according to claim 33 or 34, comprising camera positioning means (210).

36. Processing system according to claim 35, wherein the camera positioning means (210) comprises a screen (211).

37. Processing system according to claim 36, wherein the camera positioning means (210) comprises a first ruler (212) and a second ruler (214), and a first tilt pointer (213) and asecond tilt pointer (215) displayed on the screen (211), the first tilt pointer (213) being configured to slide up and down on the first ruler (212) as the tilt of the camera in the direction of the first ruler (212) changes, and the second tilt pointer (215) being configured to slide to the right and to the left on the second ruler (214) as the tilt of the camera in the direction of the second ruler (214) changes.

38. Processing system according to claims 36 or 37, wherein the camera positioning means (210) comprises a first position reference (216), and a second position reference (217) displayed on the screen (211), the first position reference (216) being configured to display on the screen (211) the location where a first reference of the child’s body part is to be placed, and the second reference (217) being configured to display on the screen (211) the location where a second reference of the child’s body part is to be placed.

39. Processing system according to claim 38, wherein the first position reference (216) is a hand position reference (216), and the second position reference (217) is an armpit position reference (217).

40. Processing system according to any of claims 36 to 39, wherein the camera positioning means (210) comprise a template (218) configured to display on the screen (211) the location where the body part to be photographed is to be placed.41 . Processing system according to any of claims 20 to 40, comprising a correcting unit (529) configured for correcting the effect associated with Allometry in the normalized landmark configuration.

42. Processing system according to any of claims 20 to 41 , wherein the classifier unit (530) is configured for providing a probability of the child whose body part is depicted in the digital image having undernutrition.

43. Processing system according to any of claims 20 to 42, wherein the classifier is a Linear Discriminant Analysis.

44. Computer readable storage medium comprising computer-executable instructions that, when executed by a processing system (500), cause the processing system (500) to perform the following steps of a method (100) for the screening of child undernutrition froman image of a body part of the child: a digital image reception step (103) wherein the processing system (500) receives the digital image of the body part, a landmark configuration identification step (104) wherein the processing system (500) identifies a configuration of bi-dimensional landmarks and semi-landmarks defining the shape of the body part depicted in the digital image by means of geometric morphometric techniques, a coordinate acquisition step (105) wherein the processing system (500) obtains the x, y coordinates of each landmark and semi-landmark in the configuration, a normalization step (106) wherein the processing system (500) normalizes the x, y coordinates of each landmark and semi-landmark in the configuration thus obtaining a normalized landmark configuration, and a classification step (107) wherein the processing system (500) classifies the normalized landmark configuration by using a classifier trained with training data previously collected from a training sample of children in order to classify the child whose body part is depicted in the digital image as having undernutrition or not.

45. Computer readable storage medium according to claim 44, wherein the computerexecutable instructions, when executed by the processing system (500), cause the processing system (500) to normalize the x, y coordinates of each landmark and semilandmark in the configuration by means of a Procrustes analysis technique in the normalization step (106).

46. Computer readable storage medium according to claim 45, wherein the Procrustes analysis technique is a full Ordinary Procrustes Analysis, or a full Generalized Procrustes Analysis.

47. Computer readable storage medium according to any of claims 44 to 46, wherein the computer-executable instructions, when executed by the processing system (500), cause the processing system (500) to normalize the x, y coordinates of each landmark and semilandmark in the configuration based on themselves and also based on a plurality of normalized landmark configurations associated with the training sample of children, such that the plurality of normalized landmark configurations associated with the training sample of children remain unchanged during the normalization step (106).

48. Computer readable storage medium according to claim 47, wherein the computerexecutable instructions, when executed by the processing system (500), cause the processing system (500) to normalize the x, y coordinates of each landmark and semilandmark in the configuration to the mean of the plurality of normalized landmark configurations associated with the training sample of children.

49. Computer readable storage medium according to claim 47, wherein the computerexecutable instructions, when executed by the processing system (500), cause the processing system (500) to normalize the x, y coordinates of each landmark and semilandmark in the configuration to the median of the plurality of normalized landmark configurations associated with the training sample of children.

50. Computer readable storage medium according to any of claims 44 to 49, wherein the computer-executable instructions, when executed by the processing system (500), cause the processing system (500) to perform a validation step (1031) after the digital image reception step (103) and before the landmark configuration identification step (104), wherein the processing system (500) validates the digital image, such that in case of an invalid digital image, another photograph of the body part of the child is taken.

51. Computer readable storage medium according to any of claims 44 to 50, wherein the computer-executable instructions, when executed by the processing system (500), cause the processing system (500) to screen child undernutrition from the digital image of the body part of the child, the child being in supine position when the photograph is taken.

52. Computer readable storage medium according to any of claims 44 to 51 , wherein the computer-executable instructions, when executed by the processing system (500), cause the processing system (500) to screen child undernutrition from the digital image of the left arm of the child.

53. Computer readable storage medium according to claim 52, wherein the computerexecutable instructions, when executed by the processing system (500), cause the processing system (500) to screen child undernutrition from the digital image of the left arm of the child, the palm of the hand facing the camera when the photograph is taken.

54. Computer readable storage medium according to any of claims 44 to 53, wherein thecomputer-executable instructions, when executed by the processing system (500), cause the processing system (500) to perform a digital image validation step (1031) after the digital image reception step (103) and before the landmark configuration identification step (104), wherein the processing system (500) validates the digital image of the body part.

55. Computer readable storage medium according to any of claims 44 to 54, wherein the computer-executable instructions, when executed by the processing system (500), cause the processing system (500) to provide a probability that the child whose body part is depicted in the digital image has child undernutrition in the classification step (107).

56. Computer readable storage medium according to any of claims 44 to 55, wherein the computer-executable instructions, when executed by the processing system (500), cause the processing system (500) to perform an allometry correcting step (1061) after the normalization step (106) and before the classification step (107), wherein the processing system (500) corrects the effect associated with allometry in the normalized landmark configurations.

57. Computer readable storage medium according to any of claims 44 to 56, wherein the classifier is a Linear Discriminant Analysis.

Images