Method for analyzing veterinary laboratory data

Abstract

Method for analyzing laboratory test data in support of a veterinary diagnostic process operated by a user on a patient, the method comprising the following steps: - extracting and collecting a data set of laboratory test, - storing the collected data set on physical or virtual devices, - analyzing the data set to estimate the compatibility between the data set and one or more diseases, - drawing up a written report for each disease, - drawing up a visual summary of each disease, - drawing up a curve of historical patient data, and - developing a diagnosis on the basis of the set of data analyzed.

Description

[0001] METODO PER ANALISI DI DATI DI LABORATORIO VETERINARIO D E S C R I Z I O N E Settore Tecnico dell’Invenzione La presente invenzione è relativa ad un metodo per l'analisi dei dati di laboratorio a supporto del processo diagnostico veterinario. In particolare, il metodo ricomprende un modello di interazione tra veterinario e strumento di analisi dati con caricamento di dati di laboratorio per il supporto alla diagnosi, una elaborazione di files di laboratorio per estrarre i dati e un’analisi dei dati laboratorio per supporto alla diagnosi, basato su intelligenza artificiale (AI). Tecnica Nota Sono noti metodi automatizzati per processare analisi in ambito veterinario. Ad esempio, sono noti software in ambito medico e veterinario che operano su diverse fasi dell'interazione tra medico e paziente: marketing, vendite o assistenza, consentendo la gestione del carico di lavoro, distribuendo le attività e automatizzando i processi medici. Sono anche noti sistemi di telemedicina veterinaria e metodi per la diagnosi remota degli animali da compagnia. Tali metodi generalmente includono una libreria di comportamenti di animali domestici associati a un singolo animale domestico da archiviare in database cloud ed utilizzabile per diagnosticare la salute degli animali domestici. Sono infine noti metodi di gestione delle informazioni relative a interaction between veterinarian and patient to be completely and exhaustively automated. For example, known methodologies do not allow the processing of veterinary analysis data in raw form, i.e. paper printouts of laboratory tests presenting information of a different nature (name of the laboratory, location of the laboratory, etc.) not all of which are inherent to clinical data useful for diagnosis support.

[0002] There is therefore a need to define an innovative methodology for the analysis of veterinary laboratory data that is exhaustive and therefore free from the drawbacks mentioned above.

[0003] Summary of the invention

[0004] To substantially solve the technical problems highlighted above, an aim of the present invention is to define a method for the analysis of laboratory data to support the veterinary diagnostic process.

[0005] According to a first aspect, the method includes an interaction model between veterinarian and data analysis tool with loading of laboratory data to support diagnosis.

[0006] According to another aspect, the method involves the processing of laboratory files to extract the significant data to support the diagnosis.

[0007] According to a further aspect, the method includes an analysis of laboratory data to support diagnosis, based on artificial intelligence (Al). Therefore, according to the present invention, a method is provided for the analysis of laboratory data to support the veterinary diagnostic process, based on artificial intelligence and having the characteristics set out in the process claim, attached to this description.

[0008] Further embodiments of the invention, preferred and / or particularly advantageous, are described according to the characteristics set out in the attached dependent claims.

[0009] Brief description of the drawings

[0010] The invention will now be described with reference to the attached drawings, which illustrate some non-limiting implementation examples, in which:

[0011] - figure 1 is a block diagram of the method for the analysis of laboratory data to support the veterinary diagnostic process, according to a preferred embodiment of the present invention,

[0012] - figure 2 is a block diagram of a phase of the method of figure 1,

[0013] - figure 3 graphically illustrates a first sequence of signs,

[0014] - figure 4 graphically illustrates a second sequence of signs,

[0015] - figure 5 graphically illustrates a third sequence of signs, e

[0016] - figure 6 is a block diagram of a phase of the method in figure 1, focused on the analysis of data to estimate the compatibility of diseases.

[0017] Detailed Description

[0018] By way of example and not by way of limitation, the method for analyzing laboratory data to support the veterinary diagnostic process, according to the present invention, will now be described with reference to the aforementioned figures. For simplicity, reference is made to the veterinary field and therefore to a veterinary laboratory and to users who may be veterinarians. However, as will become clear below, this method also finds more general application in other medical fields.

[0019] The method is suitable for implementation on personal computers or other similar devices, using specific computer software and / or software application for mobile devices.

[0020] With reference to figure 1, the method essentially includes the following phases:

[0021] - extracting and collecting 110, by one or more users, a data set of laboratory test. As already mentioned, according to a preferred embodiment of the invention, the laboratory can be a veterinary laboratory and the users can be veterinarians. Data extraction requires recognition of the same from any source document, as will be better explained below. Data collection can also be done using an interactive automated chat interface (e.g., a chatbot).

[0022] - storing the collected data on physical or virtual devices (for example, on a cloud),

[0023] - analyzing 130 the data set to estimate the compatibility between the aforementioned data set and the diseases,

[0024] - drawing up 140 a written report for each illness,

[0025] - drawing up 150 a visual summary of diseases,

[0026] - drawing up 160 a curve of historical data of the patient, e

[0027] - developing 170, by one or more users, a diagnosis on the basis of the analyzed data.

[0028] Therefore, the results of the analysis are provided in written form as a report and in graphical form (visualization).

[0029] Conveniently, the method also includes a data visualization phase that supports the user in investigating data trends (e.g., when the test value approaches permissible limits).

[0030] Advantageously, by means of phase 140, the written report can identify for each disease the group to which it belongs, the parts of the body involved, the symptoms that may occur, a detailed description of the pathology. It may also suggest obtaining further data necessary to confirm or correct the diagnosis. Finally, the report may also include the interactive automated chat interface (e.g., chatbot).

[0031] Preferably, by means of the step 150, the visual summary of the diseases can be organized by grouping the diseases into homogeneous groups (for example, metabolic, hepatic, hormonal, etc.) and display the groups compatible with each other together, via histogram or equivalent display.

[0032] Preferably, by means of step 160, the patient's historical data, i.e. the patient's laboratory test values can be represented to identify trends, within or outside normal ranges. The historical representation can also include automatic analyzes such as:

[0033] - value prediction (e.g. extrapolation)

[0034] - control of anomalies based on trends obtained by means of artificial intelligence.

[0035] The data extraction and collection phase 110 is a very important phase of the present method, as it allows the automatic extraction of laboratory test values from any structured document, even with an unknown template. Phase 110, therefore, is divided into three further steps:

[0036] - classifying 112 each datum according to its syntactic characteristics. For example, possible classes are: the value (intended as a numerical value), the word (intended as a denomination), the interval separator, the interval threshold (intended as the minimum and maximum limit values allowed) and the unit of measurement,

[0037] - detecting and removing 114 data not associated with technical, numerical or veterinary terms, - recognizing 116 data sequences through a dictionary tree, as will be better explained below.

[0038] In particular, the first step 112 consists in identifying each data item in the text as a word, numerical value, limit or threshold numerical value, separator or unit of measurement. The process follows a series of steps to recognize the syntactic characteristics of each word, such as the presence of numbers or a unit of measurement. The aim of the process is to identify plausible patterns that can allow the identification of specimens with their fields. This step also deals with identifying the date of the document. The following table provides a practical example of such recognition: Table 1

[0039] The second step 114 focuses on the removal of veterinary inconsistent words. In this way, the extraction of only the information useful for the purposes of subsequent analyzes and, in particular, for the purposes of applying artificial intelligence algorithms is facilitated.

[0040] This information to be removed includes both words used in everyday language, such as the names of days or months, adverbs and conjunctions, and symbols or letters without semantic meaning.

[0041] Removing this information allows you to detect the correct patterns within your document.

[0042] By way of example, some of the removed words could be: report, clinic, low, client, identifier, report.

[0043] At the end of the second step, the result is a series of sequences where at least one word and one value are identified in each of them: these are the components without which the sequence would not be accepted.

[0044] Furthermore, the sequence can contain, although not necessarily, other values which identify, for example, a minimum value and a maximum value or threshold value.

[0045] The order of the elements within the sequence determines whether the sequence is valid or not. For example, it is very unlikely that a value or unit of measurement will be found before the word that identifies the sample name.

[0046] The third step 116 is dedicated to the recognition of sign sequences. Figures 3 to 5 illustrate some examples of sequences. In these figures, the large arrows indicate the path of the sequence while the thin arrows indicate possible sequence options.

[0047] In particular, figure 3 illustrates the following sequence: Sequence: HCT 28 30.3 52.3 %

[0048] Word Value Value Separator Value Unit in which appears a word (HCT), a first value (28), a second value (30.3), a separator (-), a third value (52.3) and finally a unit of measurement (%). The sequence is interpreted as follows: a test (HCT) was carried out whose result was 28 and whose threshold values were between 30.3 (minimum value) and 52.3 (maximum value). The unit of measurement is a pure number expressed as a percentage.

[0049] Figure 4 illustrates another example of a sign sequence:

[0050] Sequence: HGB 8 Clinica di Torino

[0051] Word Value Word Word Word in which a word appears (HGB), a value (8), a second word (Clinic), a third word (di) and finally a fourth word (Turin). The sequence is interpreted as follows: a test (HGB) was carried out with the result being 8. The further words are not significant and related to the sequence which will therefore be blocked and not taken into further consideration. For diagnostic purposes, in fact, it is not possible to interpret the result of a test without correlating it to a range of normal values.

[0052] Finally, Figure 5 illustrates the following sequence:

[0053] Sequence: Urobilinogeno Norm <= 1

[0054] Word Value Threshold Value in which a word (Urobilinogen), a value (Norm), a threshold (<=) and a second value (1) appears. The sequence is interpreted as follows: a test was carried out (Urobilinogen) whose result was a normal value lower than a maximum threshold (1) and, therefore, between 0 and 1.

[0055] Still according to the present invention, the data analysis phase 130 for estimating the compatibility of diseases is now described with reference to figure 6. This figure illustrates a block diagram of the laboratory data analysis steps to support diagnoses. In other words, it involves determining the compatibility of a disease with veterinary laboratory data, using artificial intelligence (Al) algorithms.

[0056] Starting from generic input laboratory data 130_I, for example blood analysis, complete blood count, urinalysis, etc., the analysis phase 130 involves the following steps:

[0057] - for each laboratory data, checking the normal range 131 and marking the data as normal, higher or lower than the normal value, or the normal range,

[0058] - retrieving 132 disease data profiles from database,

[0059] - for each profile, evaluating 133 the compatibility with the data provided,

[0060] - for each disease present in the database, estimating 134 a compatibility index ranging from 0% to 100%,

[0061] - ordering 135 diseases from the most compatible, i.e. the one with the highest compatibility index, to the least compatible, i.e. the one with the lowest compatibility index. Evidently only diseases applicable to the specific case are considered: for example, in the case of a cat, cat diseases.

[0062] In particular, through the disease data profile retrieval step 132, it is possible to determine which tests should present normal, higher than normal or lower than normal values in the case of the specific disease. Furthermore, the tests are subdivided into primary, secondary and tertiary.

[0063] By step 133, each test is controlled according to the following table: where the control symbols (check) represent:

[0064] - HIT: correspondence between the value of the laboratory test data (read value) and the value obtained from the disease data profile (expected value),

[0065] - MISS: mismatch between read value and expected value

[0066] - HARD MISS: wide incompatibility between read value and expected value.

[0067] Depending on the type of disease, the disease is discarded if:

[0068] - all primary tests are "HARD MISS", or

[0069] - none of the primary tests are "HIT" (centered).

[0070] Rejected diseases are assigned a compatibility of 0%. Furthermore, depending on the results of the check and the class the tests belong to (primary, secondary, tertiary), the values are assigned as per table 3.

[0071] The values relating to the specific disease are added over all the tests carried out and a score (SCORE) is obtained. Furthermore, these values are added considering only the tests that had the HIT symbol as control.

[0072] This way you get the maximum possible score (MAX_SCORE).

[0073] Finally, the compatibility index (index) is calculated from the following ratio:

[0074] Therefore, the method according to the present invention allows the processing of any veterinary analysis data, even those in raw form and not inherent to clinical data useful for supporting diagnoses.

[0075] It therefore allows the extraction of only significant data to support the diagnosis.

[0076] Finally, it allows the analysis of laboratory data to support diagnosis, based on artificial intelligence (Al).

[0077] In addition to the ways of implementing the invention, as described above, it should be understood that numerous further variations exist. It must also be understood that said ways of implementation are only exemplary and do not limit neither the object of the invention, nor its applications, nor its possible configurations. On the contrary, although the above description makes it possible for the skilled man to implement the present invention at least according to one of its exemplary configurations, it must be understood that numerous variations of the described components are conceivable, without thereby departing from the object of the invention, as defined in the attached claims.

Claims

C LA I M S1. Method for analyzing laboratory test data in support of a veterinary diagnostic process operated by a user on a patient, the method comprising the following steps:- extracting and collecting a data set of laboratory test,- storing the collected data set on physical or virtual devices,- analyzing the data set to estimate the compatibility between the data set and one or more diseases,- drawing up a written report for each disease,- drawing up a visual summary of each disease,- drawing up a curve of historical patient data, and- developing a diagnosis on the basis of the set of data analyzed, the method being characterized in that the step of extracting and collecting a data set of laboratory test comprises the following sub-steps:- classifying each datum according to its syntactic characteristics, which comprises identifying the datum as a word, numeric value, separator, numeric limit value or unit of measure,- detecting and removing data not associated with technical, numerical or veterinary terms, which allows veterinary inconsistent words to be removed,- recognizing sequences of data through a dictionary tree, which allows to remove data sequences that are not significant for diagnostic purposes.

2. Method according to claim 1, wherein the step of analyzing the data set to estimate the compatibility between the data set and one or more diseases comprises the following steps:- for each entry in the dataset checking a normal range and marking the entry as normal, above or below the normal range,- retrieving disease data profiles,- for each disease data profile evaluating the compatibility with data from the laboratory test dataset,- for each disease estimating a compatibility index ranging from 0% to 100%,- ordering the diseases from the one with the highest compatibility index to the one with the lowest compatibility index.

3. Method according to claim 2, wherein the step of retrieving disease data profiles comprises:- determining which tests should have normal, above normal or below normal values in the case of the specific disease,- subdividing the tests into classes according to their relevance to the data profiles of the disease.

4. Method according to claim 3, wherein the step of evaluating, for each disease data profile, the compatibility with the data set of laboratory test comprises:- defining control symbols that represent: a match between the lab test data value and the value obtained from the disease data profile, a mismatch between the lab test data value and the value obtained from thedisease data profile, and a 'wide incompatibility between value of laboratory test data and value obtained from disease data profile,- discarding a disease if all the primary tests have a large incompatibility between the value of the laboratory test data and the value obtained from the disease data profile, or none of the primary tests have a match between the value of the laboratory test data and the value obtained from the profile of disease data.

5. Method according to claim 4, wherein the step of estimating for each disease a compatibility index variable from 0% to 100% comprises:- assigning to all data of the data set of laboratories test a numerical value depending on the control symbol and the class to which the laboratory test belongs,- adding up all the numerical values of all data of the data set of laboratory test relating to the specific disease, obtaining a first score(SCORE)- adding the numerical values of the laboratory test data relating to the specific disease in which the correspondence between the value of the laboratory test data and the value obtained from the disease data profile occurs, obtaining a second score (MAX_SCORE),- calculating the compatibility index (index) using the formula:

6. Method according to claim 1, wherein the step of processing the historical data of the patient is a representation which allows to identify the trends, inside or outside the normal ranges.

7. Method according to claim 1, wherein the steps of extracting andcollecting a set of laboratory test data and producing a written report for each disease are configured to include an automatic interactive chat interface.

Images