US20190142334A1

US20190142334A1 - Diagnosis system

Info

Publication number: US20190142334A1
Application number: US16/300,483
Authority: US
Inventors: Hans-Ulrich von Sobbe
Original assignee: Hans-Ulrich von Sobbe
Current assignee: Von Sobbe Hans Ulrich
Priority date: 2016-05-10
Filing date: 2017-05-09
Publication date: 2019-05-16
Also published as: AU2017264468B2; CN109561824B; US11029678B2; US20190227535A1; AU2017264468A1; WO2017194514A1; EP3244278A1; EP3244278B1; CN109661625B; CN109561824A; CN109661625A; WO2017194513A1; ES2710338T3

Abstract

A diagnosis system includes an image supplying or recognition device for inputting skin condition parameters into a model correlating skin condition parameters to skin diseases. Values for an initial set of skin condition parameters are established and compared with stored skin condition parameters and correlated skin diseases. For correlated skin diseases matching the initial set, a probability of coincidence is calculated. For a matching correlated skin disease with the highest probability of coincidence, the system checks for additional skin diseases with a similar probability of coincidence or a probability of coincidence below a threshold, and if positive, an additional condition parameter is correlated with a skin disease of similar probability of coincidence, and an additional skin condition parameter value is requested to generate an expanded condition parameter set, and if negative, the skin disease with the highest associated probability of coincidence is displayed and transmitted to a computer.

Description

CROSS-REFERENCE TO RELATED APPLICATION

This application claims priority from European patent application No. 16168902.1 filed May 10, 2016, the disclosure of which is hereby incorporated herein in its entirety by reference.

BACKGROUND

The present invention relates to a diagnosis system for the machine-aided diagnosis of skin diseases which is currently performed by the doctor based on experience.
The object of the present invention is to improve or verify a diagnosis on skin diseases.

SUMMARY

The invention is achieved by means of a diagnosis system according to claim 1 and by means of a diagnosis method according to claim 15. Advantageous further embodiments of the invention are the subject of the attached dependent claims. Preferred embodiments are likewise described in the description and also in the drawing.
According to the invention the diagnosis system for a machine-aided diagnosis of skin diseases comprises an input device for retrieving/inputting skin condition parameters of the skin to be diagnosed. The input device can be a terminal or keyboard but also may be an image forming and/or image recognition device so that it is possible even to make a diagnosis based on a photo of the skin to be diagnosed.
The diagnosis system further comprises a processing logic with a model comprising mutual correlations of different skin condition parameters and correlated skin diseases. The processing logic may be a conventional microprocessor or a neural network or it may comprise a genetic algorithm.
The model may, for example, be an associative memory comprising dependencies between different skin condition parameters and skin diseases.
Optionally the value at least of one condition parameter is calculated from the retrieved/inputted skin condition parameters by means of data from the model. This step is optional and not necessary to carry out the invention.
From the retrieved/inputted skin condition parameters and optionally from the calculated operation parameter an initial set of skin condition parameters is formed, from which the diagnosis is started.

The Diagnosis System

a) compares the values of the initial set of skin condition parameters with stored skin condition parameters and correlated skin diseases from the model, and now the following succession of steps is performed by the diagnosis system.
b) the processing logic calculates for each correlated skin disease matching the initial set of skin condition parameters or an expanded set of skin condition parameters a probability of coincidence,
c) the processing logic checks for the skin disease with the highest probability of coincidence (matching correlated skin disease) whether there are additional skin diseases with a similar probability of coincidence and/or whether the probability of coincidence of the matching correlated skin disease is below a pre-defined threshold value,
d) if the check in step c) is positive, the processing logic establishes by means of the model at least one additional condition parameter being correlated with at least one of the skin diseases of similar probability of coincidence, and
e) requests the value of said additional condition parameter from the input device and returns to step b) with an expanded set of condition parameters, said expanded set of condition parameters comprising the initial set of condition parameters and said additional condition parameter,
f) if the check in step c) is negative or after an abort criterion has been achieved, the processing logic indicates the skin disease with the highest associated probability of coincidence on a display and/or transmits said information to a computer system.
With this inventive diagnosis system, it can be checked whether a diagnosis is made based on the skin condition parameters with a sufficiently high probability of coincidence, so that the diagnosis has a certain level or reliability. On the other hand, it is ensured that the matching skin disease which matches the skin condition parameters with the highest probability of coincidence has a probability of coincidence which differs from the next probable one by a value which also ensures that the diagnoses is quite reliable. Thus, the system can effectively be used to diagnose skin diseases or to assist a doctor in his diagnosis.
Preferably, the processing logic is designed to repeat steps b) to e) until an abort criterion is achieved that relates either to a period of time or to the difference between the probability of coincidence of the matching skin disease and the probability of coincidence of the next-probable skin disease, wherein, once the abort criterion has been achieved, in step g) the skin diseases with similar probability of coincidence are indicated on the display. Via this measure it is ensured that if the diagnosed skin diseases with the highest probability of coincidence have a similar level of probability of coincidence (e.g. differ more than 10% of the probability of coincidence of the matching skin disease) all these skin diseases are displayed together with their probability of coincidence so that the doctor may verify the results with his experience.
Preferably, in step g) the skin disease with the highest probability of coincidence is displayed together with its probability of coincidence which gives the doctor feedback about the reliability of the diagnosis.
Preferably, the processing logic has a decision making network, e.g. a neural network, which is self-organizing and self-learning to improve its function with continuing operating time.
Preferably, the model is an associative memory comprising dependencies between different skin condition parameters and skin diseases. This associative memory forms an efficient linked database as a model for the processing logic.
In this case, the associative memory preferably is self-organized as to be able to include new parameters and to establish dependencies based on decision making history. Thus, the reliability and accuracy of the diagnoses will improve over the operating time of the diagnosis system.
Preferably, the processing logic in connection with the model is configured to retrieve the additional condition parameter of a type which excludes as many competitive condition parameters as possible. This means that the processing logic and the model co-act so as to retrieve an additional skin condition parameter which excludes as many competitive skin condition parameters as possible so that the diagnosis is terminated in a shorter time period.
Preferably, the processing logic defines a probability of coincidence as similar if the difference between the probability of coincidence of the matching skin disease and the probability of coincidence of the next probable skin disease is within a threshold value, for example at most 30%, more preferably at most 20%, most preferably at most 10% of the probability of coincidence of the matching skin disease.
Preferably, the model is embodied as a neural network that automatically creates correlations between skin condition parameters and/or between skin condition parameters and skin diseases on the basis of the diagnosis activity of the diagnosis system. Also, this embodiment includes a self-learning function which increases the accuracy of the diagnosis over the operation time of the system.
Preferably, the model comprises a fuzzy logic, which enables the use of various different parameters to be considered in the finding of the correct diagnosis.
Preferably the ID or name of the further condition parameter to be inputted is indicated on a display of the input terminal, so that the doctor or the assistant knows which further parameter value is to be inputted. Thus, operating mistakes of the system can be minimized.
In one embodiment of the invention the input device is an image supplying or image forming device, which is combined with an image recognition device, whereby the output of the image recognition device or a storage with the output data of the image recognition device are fed to the processing logic. Thus, the diagnosis can even be made based on an image or photo of a diseased skin portion.
The invention also relates to a diagnosis method for machine-aided diagnosis of skin diseases,
comprising an input device for retrieving/inputting skin condition parameters of the skin to be diagnosed, which diagnosis system comprises a processing logic with a model comprising mutual correlations of different skin condition parameters and correlated skin diseases, optionally the value at least of one condition parameter is calculated from the retrieved/inputted skin condition parameters by means of data from the model, from the retrieved/inputted skin condition parameters and optionally from the calculated operation parameter, an initial set of skin condition parameters is formed, in which diagnosis system
a) the values of the initial set of skin condition parameters are compared with stored skin condition parameters and correlated skin diseases from the model, whereby
b) for each correlated skin disease matching the initial set of skin condition parameters or an expanded set of skin condition parameters a probability of coincidence is calculated,
c) for the skin disease with the highest probability of coincidence (matching correlated skin disease) it is checked whether there are further skin diseases with a similar probability of coincidence and/or whether the probability of coincidence of the matching correlated skin disease is below a pre-defined threshold value,
d) if the check in step c) is positive, by means of the model, at least one further condition parameter is established which is correlated with at least one of the skin diseases of similar probability of coincidence, and
e) the value of said further condition parameter is requested from the input device and it is returned to step b) with an expanded set of condition parameters comprising the initial set of condition parameters and said further condition parameter,
f) if the check in step c) is negative or after an abort criterion has been achieved, the processing logic indicates the skin disease with the highest associated probability of coincidence on a display and/or transmits said information to a computer system. With respect to the features and advantages of the inventive method it is referred to the above description of the inventive system.
Preferably, steps b) to e) are repeated until an abort criterion is achieved that relates either to a period of time or to the difference between the probability of coincidence of the matching skin disease and the probability of coincidence of the next-probable skin disease, wherein, once the abort criterion has been achieved, in step g), the skin diseases with similar probability of coincidence are indicated on the display.
Preferably, a probability of coincidence is defined as similar if the difference between the probability of coincidence of the matching skin disease and the probability of coincidence of the next probable skin disease is within a threshold value, for example at most 30%, more preferably at most 20%, most preferably at most 10% of the probability of the matching skin disease.
In one preferred embodiment of the inventive method, the skin condition parameters are retrieved from an image forming device and/or image recognizing device. Thus, the diagnosis can be directly based on an image of the skin or a photo thereof.
In one advantageous embodiment of the invention, the model is embodied as a neural network that, on the basis of the analytic activity of the diagnosis system, automatically creates correlations between the skin condition parameters and/or between skin condition parameters and skin diseases. The processing logic is therefore designed so as to feed into the model correlations of values of the skin condition parameters and skin diseases, said correlations being ascertained during the course of the diagnosis procedure. In other words, the model has a self-learning function and is therefore able to continuously improve the correlations between the stored skin condition parameters and skin diseases. The obtained diagnosis results are therefore constantly improving as the use of the system increases. Suitable models for this are for example a fuzzy logic or a genetic algorithm so as to improve the logic links of the neural network.
Preferably, the process logic comprises a decision making network, e.g. a neural network, as well as an associative memory for dependencies between different skin condition parameters. This associative memory is advantageously self-organized as to be able to include new skin condition parameters and to establish dependencies based on decision making history. The process logic further comprises a signal output device in which the “best decision” of the decision making network is outputted in form of the requested system parameter or parameter set. The signal output device may be connected to a screen to display the diagnosed skin disease(s).
The table below shows a table with different skin condition parameters in the left column and different skin diseases in the first row. The values in the table fields are correlation values between the skin condition parameters and the skin diseases as they could be stored in the model of the inventive system. By increasing the number of skin condition parameters, the reliability of the diagnosis result is improved. The system itself asks the assistant or doctor for further skin condition parameters if it is not able to provide sufficiently exact results. If the input device is a photograph or image the system may even be able to retrieve skin condition parameters by itself without interposing a doctor or an assistant.


		Psoriasis	Atopisches			Lichen
xxx	Frage/Antwort xxxxx	vulgaris	Ekzem	Rosacea	Erysipel	ruber

	1. Have you got red and
	scaly skin areas?
1	never	0.1	0.2	0.7	0.5	0.7
2	sometimes	0.4	0.8	0.5	0.6	0.1
3	often	0.7	0.7	0.1	0.2	0.1
4	always	0.7	0.5	0.1	0.3	0.3
	2. Is the skin itching?
1	never	0.5	0.0	0.8	0.6	0.0
2	seldom	0.7	0.2	0.6	0.8	0.2
3	sometimes	0.9	0.6	0.5	0.7	0.3
4	often	0.2	0.9	0.1	0.1	0.9
5	almost daily	0.4	0.7	0.2	0.5	0.9
	3. Have you got allergies?
1	yes	0.3	0.8	0.5	0.5	0.5
2	no	0.6	0.2	0.5	0.5	0.5

Further relevant skin condition parameters may be age and sex of the person, location, color and texture of the skin alteration(s).
This diagnosis might be supported by a smartphone photo by which a pre-diagnosis of the skin can be performed, so that already by the image recognition the number of possible skin diseases can be reduced.
In case of the use of a fuzzy data base as a model, the skin diseases may be classified into classes. The classes are linked to skin condition parameters which are evaluated in a fuzzy format, as shown in the above table. Via communication of the system with the patient, assistant and/or doctor when retrieving the values of further skin condition parameters, the most probable class is found so that a certain solution (matching skin disease) can be outputted with a high probability of coincidence.
The following terms are used as synonyms: condition—state—status; value—value range; coincidence—existence; correlation—interdependency; parameter—data
It should be understood that the inventive system comprises software and hardware components having the above mentioned combined functionality and that the inventive method describes the functionality which can be performed by the inventive system. Thus, features from the inventive system may be employed for the inventive method and vice versa.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention is described herein by way of example with reference to an exemplary embodiment in conjunction with the schematic drawing, in which:

FIG. 1 illustrates a view of the components of a diagnosis system in accordance with the invention, and

FIG. 2 illustrates a flow chart relating to the inventive diagnosis procedure.

DETAILED DESCRIPTION

FIG. 1 illustrates a diagnosis system 10 comprising processing logic 12, for example a microprocessor, in which a model 14 of a skin knowledge base 16 is mapped. The processing unit 12. The diagnosis system 10 comprises furthermore a display 20 and an input keyboard 22 for displaying or inputting data. The processing unit 12 is furthermore connected with an image forming device, e.g. a camera 24 whose output is connected to an image recognition device 26, so that the output of the image recognition device 26 is able to provide skin condition parameters for the diagnosis system 10.
The method sequence for diagnosing a skin disease is illustrated in FIG. 2. The procedure starts at Point A and in step a) of the procedure wherein the values of at least two different skin condition parameters of the skin to be diagnosed are retrieved via keyboard 22 and/or the image recognition device 26. In the procedural step b) the processing logic 12 of the diagnosis system 10 correlates the values of the retrieved skin condition parameters, the initial set of skin condition parameters, by means of the model 14, with at least one skin disease. Once this has occurred, a check is performed in step c) as to whether possibly multiple skin diseases with similar probabilities are correlated with the initial skin condition parameter set and whether the probability of coincidence of the matching skin disease (the skin disease with the highest probability of coincidence) is too low, i.e. is below a predefined threshold value, e.g. 60%. If this is not the case, the procedure continues to step f) in which the matching skin disease with the associated probability of coincidence is indicated on the display 20 as a diagnosis result.
If it is found in step c) that multiple skin diseases with a similar probability of coincidence are correlated with the initial skin condition parameter set, the processing logic 12 ascertains by means of the model 14 at least one further skin condition parameter that is correlated with at least one of the found skin diseases. Thus, an expanded set of skin condition parameters is formed of the initial skin condition parameter set and the further skin condition parameter.
Subsequently, in step d), the further condition parameter is retrieved either directly from the operating system via the image recognition device 26 or by way of requesting the input of a further skin condition parameter via the input device 22 in conjunction with the display 20. As a consequence of which in step e) an expanded set of skin condition parameter values is formed, which is now returned back to step b). It should now be possible by means of the further skin condition parameter to provide an improved diagnosis, i.e. matching a skin disease with a better probability of coincidence and/or with a better distance of probability of coincidence to the next probable skin disease found in step b). In the event that this should not yet be the case, in step c) the procedure branches again into the steps d) to e) in which then further condition parameters are retrieved from the image recognition device 26 and/or inputted via the input device 22 until finally the matching skin disease has been ascertained with sufficient accuracy or an abort criterion is achieved that includes for example a specific period of time, for example 0.1 s, or a sufficiently small change in the probability value of the most probable skin disease to the next probable one.
The invention is not limited to the described exemplary embodiment but rather can be varied within the protective scope of the attached claims.

Claims

What is claimed is:

1. A diagnosis system for a machine-aided diagnosis of skin diseases,

comprising an input device for retrieving/inputting skin condition parameters of the skin to be diagnosed, which diagnosis system comprises a processing logic with a model comprising mutual correlations of different skin condition parameters and correlated skin diseases,

optionally the value at least of one condition parameter is calculated from the retrieved/inputted skin condition parameters by means of data from the model, from the retrieved/inputted skin condition parameters and optionally from the calculated operation parameter an initial set of skin condition parameters is formed,

which diagnosis system

a) compares the values of the initial set of skin condition parameters with stored skin condition parameters and correlated skin diseases from the model,

b) the processing logic calculates for each correlated skin disease matching the initial set of skin condition parameters or an expanded set of skin condition parameters a probability of coincidence,

c) the processing logic checks, for a matching correlated skin disease comprising the skin disease with the highest probability of coincidence, whether there are additional skin diseases with a similar probability of coincidence and/or whether the probability of coincidence of the matching correlated skin disease is below a pre-defined threshold value,

d) if the check in step c) is positive, the processing logic establishes by means of the model at least one additional skin condition parameter being correlated with at least one of the skin diseases of similar probability of coincidence, and

e) requests the value of said further additional skin condition parameter from the input device and returns to step b) with the expanded set of skin condition parameters, said expanded set of skin condition parameters comprising the initial set of condition parameters and said additional skin condition parameter,

f) if the check in step c) is negative or after an abort criterion has been achieved, the processing logic indicates the skin disease with the highest associated probability of coincidence on a display and/or transmits said information to a computer system, wherein the processing logic in connection with the model is configured to retrieve the additional skin condition parameter of a type which excludes as many competitive skin condition parameters as possible.

2. The diagnosis system as claimed in claim 1, wherein the processing logic is designed to repeat steps b) to e) until an abort criterion is achieved that relates either to a period of time or to the difference between the probability of coincidence of the matching skin disease and the probability of coincidence of the next-probable skin disease, wherein, once the abort criterion has been achieved, in a step g) the skin diseases with similar probability of coincidence are indicated on the display.

3. The diagnosis system as claimed in claim 1, wherein in step g) the skin disease with the highest probability of coincidence is displayed together with its probability of coincidence.

4. The diagnosis system as claimed in claim 1, wherein the processing logic has a decision making network, such as a neural network.

5. The diagnosis system as claimed in claim 1, wherein the model is an associative memory comprising dependencies between different skin condition parameters and skin diseases.

6. The diagnosis system as claimed in claim 5, wherein the associative memory is self-organized as to be able to include new parameters and to establish dependencies based on a decision making history.

7. The diagnosis system as claimed in claim 1, wherein the processing logic defines a probability of coincidence as similar if the difference between the probability of coincidence of the matching skin disease and the probability of coincidence of the next probable skin disease is within a threshold value, for example at most 30%, more preferably at most 20%, most preferably at most 10%.

8. The diagnosis system as claimed in claim 1, wherein the model is embodied as a neural network that automatically creates correlations between skin condition parameters and/or between skin condition parameters and skin diseases on the basis of the diagnosis activity of the diagnosis system.

9. The diagnosis system as claimed in claim 1, wherein the model comprises a fuzzy logic.

10. The diagnosis system as claimed in claim 1, wherein the ID or name of the additional condition parameter to be inputted is indicated on a display of the input terminal.

11. The diagnosis system as claimed in claim 1, wherein the input device is an image supplying or image forming device, and/or an image recognition device, whereby the output of the image forming/recognition device or a storage with the output data of the image forming/recognition device is connected to the processing logic.

12. A diagnosis method for machine-aided diagnosis of skin diseases,

optionally the value of at least one condition parameter is calculated from the retrieved/inputted skin condition parameters by means of data from the model, wherein from the retrieved/inputted skin condition parameters and optionally from the calculated operation parameter, an initial set of skin condition parameters is formed,

in which diagnosis system

a) the values of the initial set of skin condition parameters are compared with stored skin condition parameters and correlated skin diseases from the model,

b) for each correlated skin disease matching the initial set of skin condition parameters or an expanded set of skin condition parameters a probability of coincidence is calculated,

c) for a matching correlated skin disease, comprising the skin disease with the highest probability of coincidence, it is checked whether there are further skin diseases with a similar probability of coincidence and/or whether the probability of coincidence of the matching correlated skin disease is below a pre-defined threshold value,

d) if the check in step c) is positive, by means of the model, at least one additional skin condition parameter is established which is correlated with at least one of the skin diseases of similar probability of coincidence, and

e) the value of said additional skin condition parameter is requested from the input device and it is returned to step b) with an expanded set of skin condition parameters comprising the initial set of skin condition parameters and said additional skin condition parameter,

f) if the check in step c) is negative or after an abort criterion has been achieved, the processing logic indicates the skin disease with the highest associated probability of coincidence on a display and/or transmits said information to a computer system, wherein an additional skin condition parameter of a type is retrieved which excludes as many competitive skin condition parameters as possible.

13. The method as claimed in claim 12, wherein steps b) to e) are repeated until an abort criterion is achieved that relates either to a period of time or to the difference between the probability of coincidence of the matching skin disease and the probability of coincidence of the next-probable skin disease, wherein, once the abort criterion has been achieved, in a step g) the skin diseases with similar probability of coincidence are indicated on the display.

14. The method as claimed in claim 12, a probability of coincidence is defined as similar if the difference between the probability of coincidence of the matching skin disease and the probability of coincidence of the next probable skin disease is within a threshold value, for example at most 30%, more preferably at most 20%, most preferably at most 10%.

15. The method as claimed in claim 12, wherein the skin condition parameters are retrieved from an image forming device and/or image recognizing device.

16. A diagnosis system for diagnosing skin diseases, comprising:

a processing logic running on a computer processor, the processing logic comprising a model having mutual correlations of different skin condition parameters and correlated skin diseases;

an image supplying device or image forming device in combination with an image recognition device for retrieving from or inputting into the processing logic skin condition parameters of the skin to be diagnosed, wherein, optionally, the value at least of one skin condition parameter is calculated from the retrieved/inputted skin condition parameters by means of data from the model, and wherein, from the retrieved/inputted skin condition parameters and optionally from the calculated skin condition parameter, an initial set of skin condition parameters is formed,

wherein the processing logic:

b) calculates for each correlated skin disease matching the initial set of skin condition parameters or an expanded set of skin condition parameters a probability of coincidence,

c) checks for the skin disease with the highest probability of coincidence which is designated as a matching correlated skin disease and checks whether there are any additional skin diseases with a similar probability of coincidence or whether the probability of coincidence of the matching correlated skin disease is below a pre-defined threshold value,

and wherein:

if the check in step c) is positive, the processing logic establishes by means of the model at least one additional condition parameter being correlated with at least one of the skin diseases of similar probability of coincidence, and requests the value of said additional condition parameter from the input device and returns to step b) with an expanded set of condition parameters, said expanded set of condition parameters comprising the initial set of condition parameters and said additional condition parameter,

if the check in step c) is negative or after an abort criterion has been achieved, the processing logic indicates the skin disease with the highest associated probability of coincidence on a display and transmits said information to the computer processor, wherein the processing logic in connection with the model is configured to retrieve the additional condition parameter of a type which excludes as many competitive condition parameters as possible.