WO2024104704A1 - Système fondé sur l'ia pour la surveillance de patients souffrant d'une maladie vasculaire périphérique - Google Patents

Système fondé sur l'ia pour la surveillance de patients souffrant d'une maladie vasculaire périphérique Download PDF

Info

Publication number
WO2024104704A1
WO2024104704A1 PCT/EP2023/079161 EP2023079161W WO2024104704A1 WO 2024104704 A1 WO2024104704 A1 WO 2024104704A1 EP 2023079161 W EP2023079161 W EP 2023079161W WO 2024104704 A1 WO2024104704 A1 WO 2024104704A1
Authority
WO
WIPO (PCT)
Prior art keywords
patient
visual information
anomaly
scoring value
skin
Prior art date
Application number
PCT/EP2023/079161
Other languages
English (en)
Inventor
Azadeh MEHRABI
Original Assignee
Biotronik Ag
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Biotronik Ag filed Critical Biotronik Ag
Publication of WO2024104704A1 publication Critical patent/WO2024104704A1/fr

Links

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/44—Detecting, measuring or recording for evaluating the integumentary system, e.g. skin, hair or nails
    • A61B5/441—Skin evaluation, e.g. for skin disorder diagnosis
    • A61B5/445—Evaluating skin irritation or skin trauma, e.g. rash, eczema, wound, bed sore
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/44—Detecting, measuring or recording for evaluating the integumentary system, e.g. skin, hair or nails
    • A61B5/441—Skin evaluation, e.g. for skin disorder diagnosis
    • A61B5/444—Evaluating skin marks, e.g. mole, nevi, tumour, scar
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H30/00—ICT specially adapted for the handling or processing of medical images
    • G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0059—Measuring for diagnostic purposes; Identification of persons using light, e.g. diagnosis by transillumination, diascopy, fluorescence
    • A61B5/0077—Devices for viewing the surface of the body, e.g. camera, magnifying lens
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
    • G16H10/20—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for electronic clinical trials or questionnaires
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
    • G16H40/60—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the operation of medical equipment or devices
    • G16H40/63—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the operation of medical equipment or devices for local operation

Definitions

  • PVD peripheral vascular diseases
  • the device may comprise means for obtaining visual information associated with a skin of the patient and means for detecting an anomaly on the skin of the patient based at least in part on the obtained visual information. Moreover, the device may comprise means for assigning a scoring value to the detected anomaly, wherein the scoring value is associated with the health risk for the patient.
  • the device may be implemented as a remote entity such as, e.g., a server, but also as a local entity, such as a smartphone, for example.
  • a remote entity such as, e.g., a server
  • a local entity such as a smartphone
  • the visual information may comprise information of a locally restricted area of the skin of the patient (e.g., a locally restricted area of the skin at the neck of the patient and/or a locally restricted area of the skin at an upper leg of the patient and/or any other (suitable) locally restricted area of the skin of the patient).
  • the locally restricted area may, e.g., be between 1 to 30 cm 2 , preferably between 5 to 25 cm 2 , more preferably between 10 and 20 cm 2 or any other suitable area.
  • An anomaly may be understood as a feature derivable from the visual information indicative for a health risk for the patient. Additionally or alternatively, an anomaly may be understood as a feature in the obtained visual information associated with a diseased patient which is not present in respective visual information associated with a patient considered as healthy.
  • the scoring value may be provided as a number (e.g., an integer number) within a predefined numerical interval which may indicate a likelihood that the patient suffers from a certain heath risk. As an example, the scoring value may be any number between 0 and 100, wherein the number may indicate a likelihood (in percent) that the patient suffers from a certain health risk.
  • the scoring value spans the interval of 0 to 10, wherein 0 may be an indication for the likelihood that the patient does (most likely) not suffer from a certain health risk whereas 10 may be understood as an indication for the likelihood that the patient (most likely) suffers from a certain health risk.
  • the means for obtaining visual information are means for continuously obtaining visual information and/or the means for detecting an anomaly are means for continuously detecting an anomaly and/or the means for assigning a scoring value are means for continuously assigning a scoring value. Accordingly, a tighter monitoring of the health risk of the patient can be achieved.
  • This device may allow a noninvasive diagnostic and/or monitoring tool for a potential health risk for a patient.
  • the device may be used by a patient (e.g., due to its simplicity and/or its preferably compact form factor) to track and recognize existing symptoms of the patient or as a tracking tool for physicians to gain access to the progressive symptoms of a potentially prevalent disease.
  • This may allow an early diagnosis of a disease and may thus support a better and safe treatment of the patient with respective decreased costs for a health system (e.g., as a potential treatment may be initialized in an early stage of a progressing disease).
  • a gathering of visual information may be done by the patient alone, the burden on doctors (e.g., in terms of workload) and other members of a medical staff which may commonly be part of the diagnosis procedure, may be reduced as a main part of the diagnoses may be substituted by the interplay between the device, for monitoring a health risk for the patient, and the patient without the need for a doctor. Moreover, the risk that a patient may not realize a potential hazard arising from a potentially present health risk may be minimized and the potential health risk may be determined before it may become an emergency.
  • Another aspect of the present invention relates to a further device for monitoring a health risk for a patient.
  • the device may comprise means for obtaining visual information associated with a skin of the patient and optionally means for detecting an anomaly on the skin of the patient based at least in part on the obtained visual information. Moreover, the device may comprise means for receiving a scoring value to be assigned to the detected anomaly, wherein the scoring value is associated with the health risk for the patient.
  • the means for obtaining visual information may comprise a sensor configured to acquire visual information associated with the skin of the patient.
  • the sensor may be provided as a camera, e.g. including a CCD chip.
  • the means for receiving the scoring value may comprise means for communicating with a remote entity such that the scoring value may be received from the remote entity.
  • the means for communication may also comprise means for transmitting the visual information to another device, e.g. a server.
  • the another device may then detect the anomaly and/or assign the scoring value to the anomaly.
  • the device may comprise means for determining the scoring value at least partially locally, e.g., by the device itself.
  • the another device respectively, the means for determining the scoring value may be configured to execute an algorithm which determines or assists in determining the scoring value.
  • a portable device may be provided for monitoring a health risk for a patient.
  • the device may not necessarily be provided with means for performing the calculation of the scoring value and/or detecting the anomaly which may be accompanied by a respective high energy demand, which may disadvantageous ⁇ affect a battery lifetime of the device. Therefore, by outsourcing the calculation of a scoring value, the battery lifetime of the device may be extended, and the device may be provided with a more compact formfactor. Additionally, the manufacturing costs of the device may be decreased.
  • the device for monitoring a health risk for a patient may be adapted to monitor a prevalence, presence and/or progression of a peripheral arterial disease, PAD, and/or a peripheral vascular disease, PVD, and/or a diabetes.
  • the device for monitoring a health risk for a patient may be adapted to monitor a likelihood, a degree of progression, a probability of having, not having, developing and/or not developing PAD, PVD and/or diabetes and/or a likelihood of complications from PAD, PVD and/or diabetes.
  • the monitoring of a prevalence may comprise a monitoring of a likelihood whether a patient suffers from one or more of the aforementioned diseases and/or whether an existing disease has deteriorated over time.
  • an early diagnosis and/or treatment of a potential prevalence of at least one of said diseases may be supported. This may in particular be advantageous since the prevalence of said diseases may oftentimes progress without any symptoms or with only a negligible deterioration of potentially existing symptoms such that the prevalence and/or a deterioration of any of said diseases may not be recognized during an early stage.
  • the device may thus advantageously contribute to a decrease of the morbidity of diseased patients.
  • the visual information may be a video sequence.
  • a video sequence may relate to a video sequence with a duration of 1-10 s, preferably of 2- 8 s, more preferably of 5-6 s.
  • the video sequence may be a color video.
  • the video may be a black and white video.
  • the visual information may comprise a series of photographs.
  • the video sequence may in some cases additionally comprise one or more freeze images.
  • a larger skin area (as compared to a single freeze image) of the skin may be provided. This may allow a monitoring of respective larger area of the skin of the patient and may thus more likely support an early diagnosis (and preferably a respective early treatment of a prevailing disease) as a potential small anomaly may less likely be overlooked.
  • the means for detecting the anomaly may be adapted to detect a lesion at at least one location on the skin of the patient and preferably an appearance of the lesion.
  • a lesion may be understood as any damage or abnormal change of the skin of the patient. This may comprise a wound, a hematoma, an ulcer, a discoloration, a pigmentation, combinations therefrom or any other lesion which may be associated with a health risk for a patient.
  • An appearance may be understood as a dimension of the lesion (e.g., a maximum (e.g., a top to bottom) extension in length units and/or a surface area of the lesion), a color and/or a color change of the lesion (as compared to a previously detected anomaly) and/or any other visually acquirable parameters which may be associated with a health risk for the patient.
  • a dimension of the lesion e.g., a maximum (e.g., a top to bottom) extension in length units and/or a surface area of the lesion
  • a color and/or a color change of the lesion as compared to a previously detected anomaly
  • any other visually acquirable parameters which may be associated with a health risk for the patient.
  • This may advantageously contribute to a non-invasive monitoring of a health risk for a patient as only visual information, derivable from a skin of the patient, may be required for the monitoring.
  • This may simplify the monitoring as such, may decrease the overall monitoring costs (e.g., as no complex medical equipment may be required and/or since no physically present doctors may be required), may allow a gapless monitoring (e.g., since a monitoring of a health state of the patient may not be limited to distinct screenings by, e.g., a doctor) of a health state of a patient and may thus advantageously contribute to an improved health monitoring of the patient.
  • the means for obtaining may comprise means for receiving the visual information from a local device.
  • the device for monitoring a health risk for a patient may be physically separated from the patient and/or the local device, i.e., the device for monitoring may not be in close vicinity to the patient and/or the local device and may instead be located in another room as the patient and/or another building and/or may be accessible by means of a network and/or internet connection.
  • the means for receiving may comprise means for communicating with the local device by means of a wired connection and/or preferably by means of wireless connection (e.g., Bluetooth, WiFi, LTE, 5G, etc.).
  • the means for obtaining may further comprise means for decompressing a received visual information in case the visual information is transmitted in a compressed manner.
  • the device for monitoring may support the implementation of the device as a remote entity such as, e.g., a server.
  • a remote entity such as, e.g., a server.
  • This may facilitate an at least partially centralized data warehouse system at which received visual information may be collected and/or analyzed. Therefore, a centralized data analysis may be facilitated.
  • This may additionally allow that the main computing power for detecting an anomaly and/or for assigning a scoring value may be provided by the device for monitoring such that the local device may not be required to be provided with dedicated data storage capabilities and/or may not be adapted with a dedicated computation power for detecting the anomaly and/or for assigning the scoring value.
  • the local device may be adapted with a compact form factor, decreased manufacturing costs and an extended battery lifetime.
  • the means for obtaining visual information are configured to obtain a first visual information associated with a first part of the skin and a second visual information associated with a second part of the skin
  • the means for detecting an anomaly are configured to detect a first anomaly on the first part of the skin at least in part on the obtained first visual information, and a second anomaly on the second part of the skin at least in part on the obtained second visual information
  • the means for assigning a scoring value are configured to assign a first scoring value to the first anomaly and a second scoring value to the second anomaly.
  • the device further comprises means for making a first comparison between the first scoring value and the second scoring value, and means for assigning a first comparison scoring value to the first anomaly based on the first comparison.
  • the first part of the skin is distinct from the second part of the skin. For instance, the first part of the skin may be part of the left leg of the patient, whereas the second part of the skin is part of the right leg.
  • the device provides a more accurate and robust determination of a health risk to a patient by comparing healthy skin with skin potentially comprising an anomaly.
  • the first comparison scoring value is associated with the health risk of the patient and allows a more accurate and robust determination of a health risk to a patient.
  • the means for obtaining visual information are further configured to obtain a third visual information associated with the first part of the skin and a fourth visual information associated with the second part of the skin, wherein the third visual information and the fourth visual information are obtained later in time than the first visual information, respectively, the second visual information.
  • the means for detecting an anomaly are further configured to detect a third anomaly on the first part of the skin at least in part on the obtained third visual information, and a fourth anomaly on the second part of the skin at least in part on the obtained fourth visual information
  • the means for assigning a scoring value are configured to assign a third scoring value to the third anomaly and a fourth scoring value to the fourth anomaly
  • the means for making the first comparison are further configured to make a second comparison between the first scoring value and the third scoring value, and a third comparison between the second scoring value and the fourth scoring value
  • the means for assigning the first comparison scoring value are further configured to assign a second comparison scoring value to the third anomaly based on the second comparison and the third comparison and/or the first comparison and/or the fourth comparison.
  • the third and the fourth visual information are obtained at some time later than the first and second visual information in order to capture a potential progression of the potential anomaly.
  • a reasonable time between obtaining the first, respectively, second visual information and the third, respectively, the fourth visual information may be hours, days, weeks or even months.
  • the means for obtaining visual information are continuously obtaining visual information of the first and /or second part of the skin at more than two points in time. According to this embodiment, the progression of an anomaly in time may be determined and compared to the progression of healthy skin over time.
  • the device provides an even more accurate and robust determination of a health risk to a patient.
  • the means for obtaining visual information are configured to obtain a first visual information and a second visual information, wherein the first visual information and the second visual information are associated with a part of the skin of the patient and the second visual information is obtained later in time than the first visual information.
  • the means for detecting an anomaly are configured to detect a first anomaly on the part of the skin at least in part on the obtained first visual information, and a second anomaly on the part of the skin at least in part on the obtained second visual information, and the means for assigning a scoring value are configured to assign a first scoring value to the first anomaly and a second scoring value to the second anomaly.
  • the device further comprises means for making a comparison between the first scoring value and the second scoring value, and means for assigning a comparison scoring value to the second anomaly based on the comparison.
  • the second visual information is obtained at some time later than the first visual information in order to capture a potential progression of the potential anomaly.
  • a preferrable duration of the time period between obtaining the first visual information and the second visual information may be hours, days, weeks or even months.
  • the means for obtaining visual information are configured to continuously obtain visual information of the part of the skin. According to this embodiment, the temporal evolution or the progression of an anomaly may be determined.
  • the comparison scoring value is associated with the health risk of the patient and allows a more accurate and robust determination of the health risk to the patient as it takes the temporal evolution of the anomaly into account.
  • the first and second comparison scoring value may also be provided as a number (e.g., an integer number) within a predefined numerical interval which may indicate a likelihood that the patient suffers from a certain health risk.
  • the first and second comparison scoring value may be any number between 0 and 100, wherein the number may indicate a likelihood (in percent) that the patient suffers from a certain health risk.
  • the first and second comparison scoring value spans the interval of 0 to 10, wherein 0 may be an indication for the likelihood that the patient does (most likely) not suffer from a certain health risk whereas 10 may be understood as an indication for the likelihood that the patient (most likely) suffers from a certain health risk.
  • the means for detecting the anomaly on the skin of the patient comprise an artificial intelligence, Al, engine or model.
  • the Al engine may comprise an autoencoder (AE), a variational autoencoder (VAE), a generative adversarial network (GAN) or a convolutional neural network (CNN) for an accurate and robust determination of a health risk to a patient.
  • AE autoencoder
  • VAE variational autoencoder
  • GAN generative adversarial network
  • CNN convolutional neural network
  • the device for monitoring a health risk for a patient may further comprise means for providing the patient with at least one question associated with the health state of the patient, wherein the at least one question is preferably associated with an extent of a pain experienced by the patient and/or a location of the pain experienced by the patient.
  • the device may additionally or alternatively comprise means for receiving at least one answer from the patient in response to the at least one question (wherein the question may optionally be provided to the patient by a different device).
  • An extent of pain may be understood as a ranking of the currently experienced pain by means of a number. As an example, the currently experienced pain may be indicated by an integer number between 0 and 10, wherein 0 may indicate that currently no pain is experienced wherein 10 may indicate that a very strong pain is experienced. It is emphasized that any other suitable interval (e.g., from 0 to 5) may equally be implemented to rank the pain experienced by the patient.
  • a location of the pain may refer to an area of the body of the patient at which pain is experienced, e.g., an area comprising an upper leg of the patient, a neck of the patient, etc.
  • the means for providing the patient with at least one question may comprise means for displaying the question (such as, e.g., a graphical display and preferably a graphical user interface (GUI)) and/or may comprise means for acoustically providing the patient with the at least one question (such as, e.g., at least one loudspeaker which may provide the question to the patient as a spoken sequence).
  • GUI graphical user interface
  • the reply may be received by displaying possible replies to the patient (e.g., on a touch-sensitive display).
  • a scale of 0 to 10 may be provided to the patient wherein the patient may select one of the numbers as a measure for the extent of pain the patient is currently experiencing.
  • the means for receiving may further comprise a microphone.
  • the microphone may be configured to receive a spoken reply to the at least one question from the patient.
  • the patient may answer “5” as an indication of the extent of pain the patient is currently experiencing in his/her right upper leg. It may be possible that the received reply is further processed/evaluated at the local device. However, additionally or alternatively, it may also be possible that the received answer is transmitted to a remote entity.
  • the means for providing may comprise a database wherein the database may be provided with the at least one question to be provided to the patient.
  • the means for providing may further comprise means for communicating with a local device, wherein the device may preferably comprise means for transmitting the question to a local device.
  • the local device may then provide the at least one question to the patient as outlined above.
  • the means for receiving may comprise means for receiving a reply to the at least one question from the local device.
  • the local device may be implemented as outlined above such that it is capable of receiving a respective response to the at least one question.
  • the monitoring of a health risk for the patient may further be improved as the monitoring (and a potential diagnosis) does not only rely on a single parameter (e.g., visual information) but may further be refined by the at least one question. Therefore, an early and more reliable diagnosis of potentially prevailing health risk for the patient may advantageously be supported.
  • a single parameter e.g., visual information
  • the device for monitoring a health risk for a patient may further comprise means for weighting the received at least one answer and/or the scoring value, respectively, the first scoring value and/or the second scoring value and/or the third scoring value and/or the fourth scoring value and/or the first comparison scoring value and/or the second comparison scoring value to provide an estimation of the prevalence of PAD, PVD and/or diabetes.
  • the means for weighting may comprise means for analyzing the received reply and extracting information from the reply.
  • the at least one question may comprise the question “Is the estimated total distance you were walking today more than 200 m or less?”.
  • the patient may reply “more than 200 m” (“more than 200 m” may be considered as information).
  • the means for weighting may assign a weighting value (e.g., an integer number between 1 to 4) to the reply, indicating to which extent the received answer and/or the scoring value, respectively, the first scoring value and/or the second scoring value and/or the third scoring value and/or the fourth scoring value and/or the first comparison scoring value and/or the second scoring value may be associated with the prevalence of PAD, PVD and/or diabetes.
  • a weighting value e.g., an integer number between 1 to 4
  • the first scoring value and/or the second scoring value and/or the third scoring value and/or the fourth scoring value and/or the first comparison scoring value and/or the second scoring value may be associated with the prevalence of PAD, PVD and/or diabetes.
  • a weighting value of 4 may be assigned to the reply, if the patient replies that pain occurred after walking less than 200 m, a weighting value of 3 may be assigned, if the patient replies that pain occurred after walking more than 200 mm, a weighting value of 2 may be assigned and if the patient replies that pain occurred after walking 1 km, a weighting value of 1 may be assigned.
  • the means for weighting may be implemented at the local device (i.e., the local device may at least partially locally assign the weighting factor to the provided response) and/or may be implemented at a remote device (i.e., the remote device may at least partially assign the weighting factor to the provided response).
  • At least one question concerning the health state of the patient may be taken into account (besides the visual information) when monitoring the health state of the patient.
  • the weighting of the at least one question may facilitate providing a probability value (encoded in a total sum of the assigned weighting values associated with the at least one question) indicating a likelihood with which a patient may suffer from PVD, PAD and/or diabetes.
  • the device for monitoring a health risk for a patient may further comprise means for requesting, preferably periodically, visual information of the skin of the patient.
  • the means for requesting may comprise means for outputting a push notification to the patient indicating that new visual information of the skin of the patient shall be supplied.
  • the means for outputting may comprise a display (if the device is implemented as a local device) such that the push notification may be supplied visually as a message on the display and/or may comprise an LED, wherein the LED may be configured to flash if new visual information is to be supplied.
  • the means for outputting may comprise a loudspeaker such that the indication that new visual information is required may be output as a beeping sound and/or a vocal prompt (e.g., a spoken sequence).
  • the (local) device may internally be configured (e.g., by means of a hardcoded configuration) to request the visual information.
  • the (local) device may be provided with means to receive a request message from a remote device and may comprise means for outputting the received message.
  • a periodic request may relate to a request which is output once a day, once a week, once a month, once a year and/or at any other suitable interval.
  • the means for requesting may additionally or alternatively be configured to output a request for visual information upon an externally supplied request such as a request provided remotely by a doctor of the patient.
  • the device By providing the device with means for (e.g., periodically) requesting visual information, it may be ensured that a gapless tracking and monitoring of the health state of a patient occurs. This may decrease the risk that a deterioration of the health state of the patient may remain unrecognized.
  • Another aspect of the present invention relates to a method for monitoring a health risk for a patient.
  • the method may comprise obtaining visual information associated with a skin of the patient and detecting an anomaly on the skin of the patient based at least in part on the obtained visual information.
  • the method may comprise assigning a scoring value to the detected anomaly, wherein the scoring value is associated with the health risk for the patient.
  • the visual information, the assigning of a scoring value, the health risk and/or the anomaly may be implemented as outlined above.
  • Another aspect of the present invention relates to a method for monitoring a health risk for a patient.
  • the method may comprise obtaining visual information associated with a skin of the patient and, optionally, detecting an anomaly on the skin of the patient based at least in part on the obtained visual information, wherein detecting an anomaly may comprise comparing the obtained visual information with previous obtained visual information.
  • the method may comprise receiving a scoring value to be assigned to the detected anomaly, wherein the scoring value is associated with the health risk for the patient.
  • the visual information, the assigning of a scoring value, the health risk and/or the anomaly may be implemented as outlined above.
  • the method for monitoring a health risk for a patient may further comprise receiving the visual information at a remote entity, and preferably performing the detecting and/or the assigning at least partially at the remote entity.
  • the visual information may initially be obtained by a local device (e.g., a camera, a medical diagnostics tool and/or a portable device (e.g., a smartphone), etc.). Subsequent to the obtaining, the obtained information may be provided to a remote device.
  • a local device e.g., a camera, a medical diagnostics tool and/or a portable device (e.g., a smartphone), etc.
  • the obtained information may be provided to a remote device.
  • computation and/or storage intensive tasks may be outsourced from the local device to the remote entity. This may allow to extend the battery lifetime of the local device and may generally allow a more compact form factor for the portable device. Moreover, by outsourcing computation intensive tasks to the remote entity (which may preferably be optimized for computation intensive tasks), a faster detecting and/or assigning may be supported. Moreover, if the detecting/assigning is performed at the remote entity, the detecting/assigning may be based at least in part on a previously acquired (historic) data set which may further improve the reliability of the detecting and/or the assigning.
  • the method for monitoring a health risk for a patient may further comprise obtaining a second visual information associated with a second part of the skin of the patient, detecting a second anomaly on the skin of the patient based at least in part on the obtained second visual information, assigning a second scoring value to the detected second anomaly, wherein the second scoring value is associated with the health risk for the patient, making a first comparison between the first scoring value and the second scoring value, assigning a first comparison scoring value to the first anomaly based on the first comparison.
  • the method provides a more accurate and robust determination of a health risk to a patient by comparing healthy skin with skin potentially comprising an anomaly.
  • the method provides an improved determination of the extent of the health risk to a patient. Accordingly, the first comparison scoring value is associated with the health risk for the patient.
  • the method for monitoring a health risk for a patient may comprise obtaining a third visual information associated with the first part of the skin of the patient after obtaining the first visual information, detecting a third anomaly on the skin of the patient based at least in part on the obtained third visual information, assigning a third scoring value to the detected third anomaly, wherein the third scoring value is associated with the health risk for the patient, making a second comparison between the first scoring value and the third scoring value, assigning a second comparison scoring value to the third anomaly based on the second comparison.
  • the second comparison scoring value can be associated to the health risk for the patient.
  • a first period of time of hours, days, weeks or months may lie in between obtaining the third visual information associated with the first part of the skin of the patient and obtaining the first visual information.
  • the frequency of obtaining visual information may be adjusted to the disease (PAD, PVD and/or diabetes), to the extent and/or progression of the disease.
  • the method may comprise obtaining visual information and/or detecting an anomaly and/or assigning a scoring value in a continuous manner. Accordingly, a tighter monitoring of the health risk of the patient can be achieved.
  • the method may comprise obtaining a fourth visual information associated with the second part of the skin of the patient after obtaining the second visual information; detecting a fourth anomaly on the skin of the patient based at least in part on the obtained fourth visual information; assigning a fourth scoring value to the detected fourth anomaly, wherein the fourth scoring value is associated with the health risk for the patient; making a third comparison between the second scoring value and the fourth scoring value, and/or a fourth comparison between the third scoring value and the fourth scoring value, and assigning a third comparison scoring value to the third anomaly based on the second comparison and the third comparison and/or the first comparison and/or the fourth comparison.
  • the embodiment combines the advantages of the embodiment comprising a comparison between scoring values associated to different parts of the skin and the embodiment comprising a comparison between scoring values associated with the same part of the skin, but with different points in time.
  • a second period of time of hours, days, weeks or months may lie in between obtaining the second visual information associated with the second part of the skin of the patient and obtaining the fourth visual information.
  • the first period of time and the second period of time are the same.
  • the method for monitoring a health risk for a patient may further comprise providing the patient with at least one question associated with the health state of the patient, wherein the at least one question is preferably associated with an extent of a pain experienced by the patient and a location of the pain experienced by the patient and receiving at least one answer from the patient in response to the at least one question.
  • the method may further comprise preferably weighting the received at least one answer to provide an estimation of a prevalence of PAD.
  • the patient may only be provided with at least one question when the method is performed for the first time. Additionally or alternatively, the patient may be provided with the at least one question periodically (e.g., once a day, once a week, once a month, once a year, etc.) and/or on an irregular basis (e.g., if a deterioration of an anomaly in the visual information is determined).
  • the detection of an anomaly may comprise providing the obtained first visual information, second visual information, third visual information and/or fourth visual information to a trained artificial intelligence, Al, engine, wherein the trained Al engine, preferably including an autoencoder, has been trained on previously obtained visual data considered as healthy and preferably based on a skin type of the patient.
  • a trained artificial intelligence, Al preferably including an autoencoder
  • the detection may further comprise deriving a first anomaly in the obtained first visual information by means of the trained Al engine, preferably by reconstructing the obtained first visual information by the autoencoder and comparing the reconstructed information to the obtained first visual information, and/or deriving a second anomaly in the obtained second visual information by means of the trained Al engine, preferably by reconstructing the obtained second visual information by the autoencoder and comparing the reconstructed information to the obtained second visual information, and/or deriving a third anomaly in the obtained third visual information by means of the trained Al engine, preferably by reconstructing the obtained third visual information by the autoencoder and comparing the reconstructed information to the obtained third visual information, and/or deriving a fourth anomaly in the obtained fourth visual information by means of the trained Al engine, preferably by reconstructing the obtained fourth visual information by the autoencoder and comparing the reconstructed information to the obtained fourth visual information.
  • the training on previously obtained visual data may be understood as training the Al engine such that it may be capable of reproducing one or more data sets comprised in the obtained visual data used for training.
  • the Al engine may comprise a neural network, comprising an input layer, an output layer and one or more hidden layers in between the input layer and the output layer. Each node of a certain layer may be interconnected to each node of a preceding and/or a subsequent layer. Each interconnection may be provided with a weighting factor. The weighting factors may be determined by training the neural network.
  • the detection of the anomaly may be based at least in part on determining a difference between the reconstructed obtained visual information and the obtained visual information which potentially comprises an anomaly.
  • the training of the Al engine to a skin type of the patient may allow an adaption of the anomaly detection to different appearances of an anomaly on the skin of the patient relative to the skin of the patient.
  • an anomaly may look different on a colored skin as compared to a Caucasian skin.
  • a training of the Al engine to different skins of the patient may thus support a more precise and reliable determination of an anomaly. It may be a separate aspect of the present invention.
  • the reliability of the monitoring may further be improved. This may be the case since the Al engine may be trained to a large data set of images and may thus potentially also be trained to rare and/or small anomalies which may, e.g., be overlooked by a doctor.
  • Another aspect of the present invention relates to a system for monitoring a health risk for a patient.
  • the system may comprise one or more of the devices as outlined above.
  • the system may comprise the device configured to act as a remote entity. Furthermore, the system may preferably comprise the device configured to act as a local entity.
  • the system may allow an advantageous synergetic interplay of the local device (as described above) and the remote entity (as described above). Such a synergetic interplay may be facilitated by combining one or more advantageous properties of the local device (e.g., portability, easy handling, easy operability, optimized user experienced, low-cost, etc.) with one or more advantageous properties of the remote entity (e.g., computation power, large storage capability, centralized data storage, centralized maintenance, etc.). This may advantageously support a reliable monitoring of a health risk for the patient.
  • one or more advantageous properties of the local device e.g., portability, easy handling, easy operability, optimized user experienced, low-cost, etc.
  • the remote entity e.g., computation power, large storage capability, centralized data storage, centralized maintenance, etc.
  • Another aspect of the present invention relates to a computer program comprising code, which when executed on a computer, performs the method as outlined above.
  • Fig. 1 Illustration of an exemplary procedure to monitor a health risk for a patient
  • Fig. 2 Illustration of exemplary claudication locations
  • Fig. 3 Illustration of exemplary locations for determining a pulse strength of a patient
  • Fig. 4 Exemplary system overview according to aspects of the present invention
  • Fig. 5 Exemplary flow chart of a method for anomaly detection in visual information
  • Fig. 6 Illustration of an exemplary neural network configured to act as an autoencoder
  • Figs. 7A and 7B Illustration of an exemplary method for anomaly detection in visual information according to aspects of the present invention
  • Fig. 8 Exemplary flowchart for assigning a scoring value to visual information obtained from two sides of a body of a patient. Attempts to diagnose PAD in a non-invasive manner exist for a long time. Typical approaches to diagnose PAD, as used in the art, are shown and compared in the table below.
  • Table 1 overview of different approaches to diagnose a PAD and/or a PVD.
  • Fig. 1 shows an exemplary flow chart for monitoring 100 a health risk for a patient, in particular a PAD vs. a PVD venous disease.
  • a commonly used monitoring flow is called “VESSEL” which is represented by the exemplary flow chart depicted in Fig. 1.
  • each of the steps 110-160 corresponds to one letter of the term “VESSEL” and each of the steps of method 100 intends to distance a potential PAD from a PVD, i.e., during each step of 100 individual characterization steps may be carried out to distance a potential PAD from a PVD.
  • arteries carry oxygen-rich blood away from the heart whereas veins return oxygen-poor blood back to the heart.
  • the individual steps to be performed during each of the method steps 110-160 is outlined below.
  • the monitoring method may start with step 110 corresponding to the letter “V”.
  • Step 110 intends to characterize various positions that may help alleviate discomfort/pain for a patient.
  • an elevation of at least one leg may decrease swelling of the at least one leg and whether the elevation of the at least one leg contributes to an improvement of a blood flow through the at least one leg. Further in this regard, it may be characterized whether a dangling of the at least one leg or standing/sitting for long periods of time (e.g., for several minutes, for several hours, etc.) leads to an increase of the pain and a potential edema.
  • step 120 corresponding to the letter “E” an explanation for the pain may be investigated.
  • This may comprise a characterization under which circumstances the pain may be experienced and how the pain may be experienced (e.g., sharp, etc.).
  • the pain may be characterized whether the pain is experienced as sharp (e.g., worst at night), whether the patient experiences “rest pain” (e.g., the patient wakes up from sleep with pain (when the legs are in the horizontal position it may impede blood flow) and whether the patient dangles the extremity off the bed to alleviate the pain.
  • “rest pain” e.g., the patient wakes up from sleep with pain (when the legs are in the horizontal position it may impede blood flow
  • an intermittent claudication may comprise a characterization whether an activity (e.g., running, walking etc.) causes severe pain in the calf muscles, thighs, buttocks etc. whereas, and whether, when the activity is stopped, the pain eases up. This may be due to fact that the muscle is being deprived of blood flow due to a prevailing PAD which may then cause the experienced pain.
  • a skin of a lower extremity e.g., at least one leg of the patient
  • a skin of a lower extremity may be characterized, preferably with respect to its color and/or temperature.
  • the skin is thin, dry/scaly, hairless (or at least loss of hair at at least one extremity (such as, e.g., at least one leg)) and whether thick toenails may be observable.
  • it may further be characterized whether a dangling of at least one leg leads to a rubor of the at least one leg and whether an elevation of at least one leg leads to a paling (or bluish skin color) of the at least one leg.
  • At least one leg shows a thick, tough skin and/or whether the at least one leg shows a brownish color.
  • step 140 corresponding to the letter “S”, a strength of pulse in at least one lower extremity may be characterized.
  • a pulse In case of investigating whether a patient suffers from a venous disease, it may be elicited whether a pulse is present (and preferably whether the pulse is normal). This may be explained due to the aspect that in case of a prevailing venous disease, there may generally not be a blockage in the blood flow from the heart to the at least on extremity. Instead, a blockage may be present in the return path of the blood flow, e.g., from the at least one extremity to the heart. Subsequently, in step 150, corresponding to the letter “E”, the presence of an edema may be investigated.
  • an edema may be present, and it oftentimes tends to deteriorate towards an end of a day.
  • step 160 corresponding to the letter “L”, the presence (and preferably the appearance) of a lesion may be characterized.
  • a lesion e.g., an ulcer
  • its location may be characterized.
  • the lesion may most likely be located at the end of at least one toe (e.g., close to a respective toenail), at the top of at least one foot (e.g., dorsum) and/or at a lateral ankle region (e.g., malleolus) of at least one foot.
  • any kind of drainage is present (e.g., very little drainage), whether some (e.g., little) tissue granulation (e.g., pale/very light pink) is present or whether the tissue granulation is necrotic/black. Further in this context, it may be elicited whether a potentially present ulcer is deeply “punched out” with noticeable margins/edges that give it a round-shaped appearance.
  • a lesion e.g., an ulcer
  • the appearance of the ulcer may be characterized. In this regard, it may be characterized whether the ulcer appears as swollen with a drainage, if a granulation is present (e.g., colored between deep pink to red) and whether the edges of the ulcer are irregular with a shallow depth.
  • FIG. 2 exemplarily illustrates the conjunction 200 between a potential claudication (e.g., a stenosis or occlusion) of at least one artery in a leg of a patient and the associated location of pain experienced (corresponding to the letter “E” of the term “VESSEL” as outlined with reference to Fig. 1, above) by the patient as a result of a stenosis or occlusion of a particular artery.
  • a potential claudication e.g., a stenosis or occlusion
  • a claudication may typically be experienced by a patient as an aching or burning in at least one leg muscle. Moreover, the pain may reliably be reproduced after walking for a certain distance and may be relieved within minutes of rest. Moreover, in case of a present claudication, pain may never be present at rest and the pain may not be exacerbated by certain positions/orientations of the leg.
  • the location of the pain may provide an indication of the site of the disease, e.g., which artery may experience a stenosis and/or occlusion.
  • a stenosis or an occlusion of an aorta 210a of a leg of a patient may cause a claudication at a bilateral buttock, a thigh and a calf 210b.
  • a stenosis or an occlusion of common iliac artery 220a may lead to pain experienced in a buttock area 220b of a leg of a patient.
  • a stenosis or an occlusion of a common femoral artery 230a may lead to pain experienced in a thigh area 230b of a leg of a patient.
  • a stenosis and or an occlusion of a superficial femoral artery 240a may cause pain in a calf area 240f of a leg of a patient.
  • Fig. 3 exemplarily shows different locations 300 at which the strength of a pulse of a patient may be determined. It is noted that a strength of a pulse may thus be associated with a location at which the pulse was determined. This location- specific determination of the pulse of a patient may provide additional information on the potential prevalence of a PAD, venous disease and/or a diabetes. It may thus be possible determine a strength of the pulse of a patient in the thigh of a leg of a patient such as in the femoral artery 310 of a patient. Additionally or alternatively, it may be possible to determine a strength of a pulse in a femoral vein 320 of a patient.
  • a strength of a pulse of a patient in a popliteal cavity of a leg of a patient such as in a popliteal artery 330 of a patient.
  • a strength of a pulse of a patient at a feet of a patient e.g., in a dorsalis pedis artery 340 of a patient.
  • Fig. 4 exemplarily shows a system 400 overview of communicating devices according to an aspect of the present invention.
  • the system 400 may comprise a local device 410 which may be adapted with means for obtaining visual information 420 of a skin of a patient.
  • the visual information 420 may comprise visual information associated with a foot and/or a leg of a patient and/or a hand and/or an arm of the patient.
  • the visual information 420 may be or may be converted into a video sequence 430.
  • the video sequence 430 may be forwarded to a remote entity 440.
  • the forwarding of the video sequence 430 may preferably occur in an encrypted manner.
  • the remote entity 440 may comprise a cloud-based system.
  • the remote entity 440 may comprise means for storing at least the video sequence 430 received from the local device 410.
  • the remote entity 440 may comprise an artificial intelligence (Al) engine for at least partially processing the received video sequence 430.
  • the processing of the video sequence 430 may comprise detecting an anomaly in the video sequence 430 and/or assigning a scoring value to a potentially detected anomaly.
  • an indication 445 may be provided to the local device 410.
  • the indication may comprise, e.g., a scoring value assigned to a detected anomaly and/or answers to questions associated with the health state of the patient. Additionally or alternatively, the indication may warn a patient that a potential existing health risk has deteriorated, that a health risk was detected for the first time and/or that medical supported may be indicated.
  • the remote entity 440 may further comprise means for communicating with a clinical system 450.
  • the clinical system 450 may comprise a hospital information system, one or more clinicians (e.g., a doctor, a nurse, scientists, etc.). By means of the clinical system 450, it may be facilitated that a doctor and/or a nurse of the patient may monitor a potential health risk for a patient based at least in part on one or more video sequences 430 received from the remote entity 440.
  • clinicians e.g., a doctor, a nurse, scientists, etc.
  • the data received at the clinical system 450 may be provided to at least one scientist who may thus gain access to the at least one video sequence 430 for the sake of studying the properties/signatures of PAD, PVD and/or diabetes.
  • the data provided to the clinical system 450 may be limited to at least one video sequence 430. In some cases, the data provided to the clinical system 450 may further comprise a scoring value associated with a prevalence of a health risk for the patient.
  • a doctor, a nurse and/or a scientist may interpret the data received from the remote entity 440 and may determine, e.g., that an error occur in the assigning of a scoring value and/or that an anomaly was detected erroneously.
  • feedback 455 may be provided to the remote entity 440 such that the Al engine of remote entity 440 may be updated based on the feedback 455. This may facilitate a more accurate determination of an anomaly and/or a more accurate assignment of a scoring value for subsequent video sequences 430.
  • the feedback 455 may comprise a direct feedback, sent by a doctor, a nurse and/or a scientist, to a patient.
  • the feedback 455 may be received by the local device 410 and may be indicated to the patient.
  • the feedback may comprise statement by, e.g., a doctor that no health risk is present, that a potentially pre-known health risk has deteriorated or may comprise a warning that a patient should seek for medical assistance.
  • the feedback 455 may comprise a request for an additional video sequence 430.
  • Fig. 5 shows an exemplary flow diagram of an anomaly detection method 500 according to an aspect of the present invention and using the Al engine as outlined above.
  • normal data sets 510 are supplied to the aforementioned Al engine for training the Al engine (e.g., Al engine 440).
  • the Al engine may learn to encode and/or reconstruct the normal data sets 510, wherein the training algorithm may, e.g., at least in part be based on a variational autoencoder (VAE), a generative adversarial network (GAN) and/or a convolutional neural network (CNN or ConvNet).
  • VAE variational autoencoder
  • GAN generative adversarial network
  • CNN or ConvNet convolutional neural network
  • VQVAE-2 vector quantized variational autoencoder-2
  • the underlying Al model may be considered as a constructed anomaly detection model 530.
  • test data set 540 may be provided to the Al engine.
  • the test data set 540 may comprise at least one representative of the training data set 510 (and/or an additional data set considered as normal which was not part of the initial training data set) and one or more data sets 545 which may comprise an anomaly.
  • test data set 540 may then be provided 550 to the trained Al engine.
  • the trained Al engine may decide 560 whether the training data set 540 is similar to the normal data sets 510 used for training the Al engine. If the training data set is similar 520 to the normal data set 510, the training data set may also be considered as normal, i.e., as not comprising an anomaly.
  • an anomaly 580 may be determined.
  • Fig. 6 shows an exemplary neural network 600 as it may be used as an autoencoder (AE).
  • AE autoencoder
  • the AE may comprise an encoder 610, exemplarily comprising an input layer 620, a first hidden layer 630 and a second hidden layer 640. Since the number of nodes in subsequent layers decreases from the input layer 620 to the second hidden layer 640, this transition is accompanied by a loss of information as will further be described below.
  • the encoder 610 may be followed by a decoder 650, exemplarily comprising, the second hidden layer 640, a third hidden layer 660 and an output layer 670. Since the number of nodes per layer increases from the second hidden layer 640 to the output layer 670, the amount of information which can be stored in each of the layers increases as it will further be described below.
  • the input layer 620 may exemplarily comprise four nodes.
  • Input layer 620 may be configured to represent an input image to be processed by the exemplary neural network 600. It is noted that the four nodes of input layer 620 are only shown for the sake of simplicity and that in real applications, the number of nodes in the input layer 620 may typically higher.
  • the neural network 600 is not limited to the processing of greyscale images only and may also be configured to process color images, e.g., in the RGB space, the YMCK space and/or any other suitable color space.
  • the factor of three accounts for the representation of the respective image in the RGB colors space comprising three basis colors.
  • the exact value of each entry in the vector accounts for the brightness of a pixel in, e.g., the red colors space.
  • a first set of 786,432 (consecutive) rows may represent the red colors space (and each row may represent the brightness of a particular pixel in the red color space)
  • a second set of 786,432 (consecutive) rows may represent the green color space (and each row may represent the brightness of a particular pixel in the green color space)
  • a third set of 786,432 (consecutive) rows may represent the blue colors space (and each row may represent the brightness of a particular pixel in the blue color space).
  • the architecture of the input layer 620 is not limited to these examples but may also be configured with any other suitable architecture.
  • Each of the nodes of the input layer 620 may be interconnected to each of three exemplary nodes assigned to the first hidden layer 630. Each interconnection may be accompanied by a weighting factor multiplied to the data propagated between two (subsequent) nodes (i.e. from a node of the input layer 620 and a node of the first hidden layer 630). Since the first hidden layer 630 may only comprise three nodes as compared to the four nodes assigned to the input layer 620, the amount of information which may be represented by the first hidden layer 630 may be lower as compared to the amount of information which may be represented by the input layer 620.
  • the transition from the input layer 620 to the first hidden layer 630 may be understood as a dimension reduction of the image input into the neural network 600.
  • the associated loss of information may, e.g., lead to one or more of a loss of color (e.g., a transition from a colored image to a black and white image, a loss of contrast, a loss of certain boundaries/margins in the image, etc.).
  • Each of the three nodes of the first hidden layer 630 may be interconnected with each of two nodes (only shown exemplarily) assigned to the second hidden layer 640. Each interconnection may be accompanied by a weighting factor multiplied to the data propagated between two (subsequent) nodes.
  • the transition from the first hidden layer 630 to the second hidden layer 640 may again be understood as a dimension reduction. This may again be understood as a loss of information of the image processed by the neural network 600.
  • the output of the second hidden layer 640 may be referred to as a reduced image.
  • a processing of the originally presented image by the decoder 650 may be performed.
  • the decoder 650 Since the number of nodes increases with each layer comprised by the decoder 650, the amount of information storable/processible in each of the layers may increase. This concept may be used by the neural network 600, and in particular by the layers/nodes assigned to the decoder 650, to reconstruct the originally presented image (as fed into the input layer 610) from the reduced image generated at the second hidden layer 640.
  • This may, inter alia, be done adding additional features to the reduced image in each of the third hidden layer 660 and the output layer 670, e.g., by adding color information to the image, contrast information, boundaries/margins, etc.). This may be encoded in respective weighting factors associated with an interconnection of two nodes in respective subsequent layers.
  • the image obtained at the output layer 670 may then be compared to the image originally provided to the input layer 620.
  • the comparing may comprise the calculation of an error function wherein the absolute value of the error function may be large if the deviation between the image provided at the output layer 670 and the image originally fed into the input layer 620 is large (i.e., the image provided at the output layer 670 is significantly different from the image originally provided to the input layer 620, wherein certain features of the original image are not present in the reconstructed image and/or wherein features appear in the reconstructed image which were not present in the original image).
  • the applied training algorithm may then try to determine the associated weighting factors such that the error function may be minimized, i.e., the weighting factors are considered to be at their respective optimum if the image provided at the output layer 670 is nearly identical (comprising at most negligible differences, e.g., within the average noise level) to the image originally provided to the input layer 620.
  • the training may be considered successful and/or complete, if the value of the error function is below a predefined threshold, preferably close to 0.
  • Figs. 7A and 7B exemplarily illustrate the training process of a GAN used for determining an anomaly in obtained visual information according to an aspect of the present invention.
  • Fig. 7A illustrates a high-level overview of an exemplary training of a GAN and an exemplary determination of a potential anomaly in the obtained visual information.
  • Raw data i.e., obtained visual information such as, e.g., the visual information 420 as described with reference to Fig. 4, above
  • preprocessed e.g., color adjustments, contrast adjustments, sharpening filters, cropping, etc.
  • the training data set 710 may be provided to the GAN training 720.
  • the GAN training 720 may be based on a set of preprocessed images 720A which may be considered as healthy, i.e., which do not comprise an anomaly, and a model 720B (e.g., comprised by the Al engine as described with reference to Fig. 4 and/or 6, above) to be trained.
  • a model 720B e.g., comprised by the Al engine as described with reference to Fig. 4 and/or 6, above
  • the GAN training 720 may exemplarily be based on the training of a generator.
  • a generator may be based on a neural network, e.g., the generator may be based on layers 640-670 of exemplary neural network 600 as described with reference to Fig. 6, above (but it may also be implemented differently, e.g. including another generative neural network).
  • the neural network of GAN training 720 may be trained to reconstruct (or more generally: generate) an artificial image which may be identified (e.g., by a discriminator as further outlined below) as a representative of the preprocessed images 720A.
  • an anomaly detection 730 may subsequently be initiated based at least in part on the trained model 720B.
  • the anomaly detection 730 may comprise providing a set of yet unseen data 730A (i.e., data which has not been used for training the model 720B) to the trained model 73 OB, identifying potential candidates 730C for an anomaly and determining 730D that the potential candidates 730C represent an anomaly.
  • Trained model 730B may additionally (as compared to the trained model 720B) comprise a trained encoder, which e.g., may be based on a neural network architecture similar to the architecture of layers 620-640 of the neural network 600 as described with reference to Fig. 6, above).
  • the unseen data may first undergo an encoding (as described with reference to Fig. 6, above, and a subsequent reconstruction by means of the trained generator comprised by the trained model 720B.
  • potential anomalies present in the unseen data 730 A may not be properly reconstructed by the trained model 730B as the underlying neural network has not been trained for reconstructing anomalies present in provided data sets. More specifically, the neural network has not been trained to reconstructed anomalies in the unseen data 730 A.
  • Identifying potential candidates 730C for an anomaly may comprise comparing the reconstructed image of the unseen data 730 A to an image of the training data 710 and/or to the unseen data 730 A originally provided to the trained model 730B. Any deviation identified from the training data thereby may be understood as indicators for the presence of a potential anomaly in the unseen data 730 A. Only deviations may be considered which exceed a predefined noise level. Noise in this case may be understood as disturbances in provided data (e.g., in the training data 720A and/or the unseen data 730A) which are not related to the remaining content of the respective data.
  • a potential anomaly may only be seen as an anomaly if it may be separated from noise (e.g., a potential anomaly may only be seen as a potential anomaly if it may be represented by more than 10 adjacent pixels and/or by a certain area in an image (comprising the potential anomaly) which appears darker than surrounding areas in the image).
  • the thus effectively arising difference an anomaly has to have as compared to image data not representing an anomaly may depend on the progression of the potential disease.
  • a potential disease, monitored in an early stage may be represented by a potential anomaly which only lies slightly above the noise level whereas a progressed disease may be represented by a potential anomaly which more clearly differs from the noise level as an anomaly associated with an early stage of a disease.
  • the deviation may, e.g., be expressed by the value of an error function. If the value of the error function exceeds a predefined threshold, a potential anomaly may be referred to as a determined anomaly 730D.
  • Fig.7B shows the GAN training process as described with reference to Fig. 7A in more detail.
  • GAN training 740 may comprise generative adversarial training sub-step 740A and may comprise encoder training 740B.
  • Sub-step 740A may comprise a generator 740C and a discriminator 740D (considered as the adversarial).
  • Generator 740C may be implemented as the generator described with reference to Fig. 7A, above.
  • Generator 740C may generate an artificial image (based on input data (e.g., random values, noise, encoded visual information and/or any other suitable input parameter) provided to the respective one or more input layers of generator 740C) which may represent visual information according to an aspect of the present invention.
  • input data e.g., random values, noise, encoded visual information and/or any other suitable input parameter
  • Said generated image is passed to the discriminator 740D which may determine if there is a difference between the artificially generated image and a real image 740E (considered as healthy, i.e., not comprising an anomaly, which may be comprised by the training data set 720A), e.g., based on calculating a value of an error function (and/or a rewards function) as outlined above.
  • the value of the error function may be fed back into the generator 740C and the generator 740C may generate, based at least in part on the value of error function, a new artificial image which may then again be passed to the discriminator 740D. This process may iteratively be repeated until the value of the error function is below a predefined threshold. In this case, the generative adversarial training sub-step 740A is considered to have successfully been terminated.
  • encoder training 740B may be carried out.
  • Encoder training 740F may take advantage of the generative adversarial training sub-step 740 A and an encoder 740F.
  • Encoder training 740B may comprise training an underlying neural network such that a dimension reduction of the visual information provided to the encoder may occur and such that the input data provided to generator 740C may be obtained as an output of the encoder 740F.
  • the encoder 740F may be implemented as the encoder described with reference to Fig. 6, above.
  • the training of the encoder 740F may be based at least in part on the training of a neural network as described elsewhere herein (e.g., based on a minimization of an error function).
  • Anomaly detection 750 may comprise providing unseen data 750B (i.e., data which has not been part of the training data 740A), which may comprise an anomaly to be detected, to the trained model 750A.
  • the trained model 750A may be based on the trained generator 740C and the trained encoder 740F.
  • unseen data 750B (or at least one representative of unseen data 750B, such as, e.g., an unseen image) may be provided to the trained model 750A.
  • the provided unseen data 750B may undergo a dimension reduction by means of the trained encoder 740F.
  • the encoded unseen data 750B may be provided to the trained generator 740C as input data based on which the trained generator 740C may generate an artificial image.
  • the generated artificial image may be compared to the unseen data 750B originally provided to the encoder 740F.
  • any difference (wherein a difference may be defined as described elsewhere herein) derivable from said comparison may be understood as an indication for a potential anomaly in the unseen data 750B as both the encoder 740F and the generator 740C have only been trained to encode and reconstruct (generate) real images 740E considered as healthy (i.e., not comprising an anomaly). Additionally or alternatively, it may also be possible that the generated artificial image is compared to one or more images of real images 740E to derive an indication for a potential anomaly in the unseen data 750B. A potential anomaly in the unseen data 750B may be seen as an anomaly if the aforementioned difference exceeds a predefined value (which preferably exceeds the noise level in the generated artificial image).
  • a key aspect of training a GAN may be seen in obtaining a trained generator 740C which is optimized (due to the training) to provide an image which will be accepted by a discriminator 740D.
  • a trained generator 740C of a GAN may be optimized to generate an artificial image which may be that close to a real image 740 that it may not be rejected by discriminator 740D (e.g., such that it may not be distinguishable from a real image 740 for the discriminator 740D).
  • a trained auto-encoder may be optimized to memorize how to reconstruct an originally presented input data set in greatest detail and efficiency.
  • the GAN-based approach for anomaly detection and the autoencoder-based concept for anomaly detection may be implemented as alternative solutions.
  • it may also be possible to combine the GAN-based approach with the autoencoder-based approach e.g., the layers 640-670 of the neural network 600 as described with reference to Fig. 6 may be replaced by generator 640C and may be trained according to an interplay of a respective generator and a respective discriminator).
  • Fig. 8 shows an exemplary flowchart of a method 800 for detecting that at least one side of a body of a patient is developing a PAD (and/or a PVD and/or a diabetes).
  • the method 800 may be based on a first step 810A obtaining visual information associated with a skin of a patient and determining whether the visual information comprises an anomaly according to aspects as presented herein.
  • the step 810A may be carried out for a first side (e.g., a right side) of the body of the patient (such as, e.g., a right leg of the patient).
  • the method 800 may further comprise obtaining visual information associated with a skin of the patient and determining whether the visual information comprises an anomaly according to aspects as presented herein.
  • the step 81 OB may be carried out for a second side (e.g., a left side) of the body of the patient (such as, e.g., a left leg of the patient).
  • a scoring value may be assigned to a respective right side of the body of the patient (scoring value 820A) and to a respective left side of the body of the patient (scoring value 820B).
  • the individually assigned scoring values (820A, 820B) may then be forwarded to a scoring comparison algorithm 830.
  • the scoring comparison algorithm 830 may compare each of the scoring values 820 A and 820B to respective previously determined scoring values (e.g., historic data) associated with previously obtained visual information. This may allow to determine a temporal evolution of the previously assigned scoring values to the presently assigned scoring values and may thus allow to determine whether a deterioration of an anomaly has occurred over time.
  • An increase between a currently assigned scoring value 820A and 820B, respectively, may be an indication for a deterioration of a health risk for a patient, e.g., of a PAD and/or a PVD and/or a diabetes.
  • scoring comparison algorithm 830 may further be configured to compare the currently assigned scoring values 820A, 820B to each other to determine a numerical difference between the currently assigned scoring values 820 A, 820B. The determined numerical difference may be compared to previously determined numerical differences to determine how the difference between assigned scoring values 820A and 820B may have changed over time.
  • the method may only comprise step 810A or 810B, and the scoring comparison algorithm 830 may only compare the scoring value 820 A or 820B to respective previously determined scoring values (e.g., historic data) associated with previously obtained visual information. This may allow to determine a temporal evolution of the previously assigned scoring value(s) to the presently assigned scoring value and may thus allow to determine whether a deterioration of an anomaly has occurred over time.
  • previously determined scoring values e.g., historic data
  • the calculation results of scoring comparing algorithm 830 may be provided as feedback 840.
  • Feedback 840 may comprise the currently assigned scoring values 820A and/or 820B.
  • the feedback may additionally or alternatively comprise the difference value between the currently assigned scoring values 820A and/or 820B.
  • the feedback 840 may comprise one or more difference values between respective currently assigned scoring values 820A and/or 820B and previously assigned scoring values.
  • a comparison scoring value may be determined to describe the change over time of the scoring value 820A and/or the scoring value 820B and/or the change of the scoring value 820A, respectively, the scoring value 820B in comparison to each other.
  • the determining whether the obtained visual information comprises an anomaly may be followed by providing the patient with at least one question concerning the health state of the patient.
  • the potentially received replies to the at least one question may then be assigned a weighting factor to estimate the association of the received replies with a certain disease (e.g., a PAD, PVD or a diabetes).
  • a certain disease e.g., a PAD, PVD or a diabetes.
  • IC intermittent claudication
  • the walking distance or speed at which symptoms occur may depend on multiple factors including the severity and site of the arterial disease. Therefore, higher scores may be assigned in the “Explanation of the pain section” of questionaries associated with the “VESSEL” approach as described above.
  • Providing the patient with at least one question concerning the health state of the patient may be done by a weighting algorithm which may, e.g., be implemented in an app, e.g., operating at the local device and/or at the remote device as described above.
  • Table 3 as recited below states several exemplary questions/topics addressed by the at least one question and respective weighting coefficients (to be assigned based at least in part on the contentual information comprised in a response to a provided question) to support the monitoring of a health risk for a patient (preferably a prevalence of PAD).
  • Table 3 Exemplary questions/topics provide to the patient and respective weighting coefficients to monitor a health risk for a patient.
  • the weighting algorithm may be configured to first characterize the “E” of “VESSEL” and may thus provide at least one question to a patient which may be associated with an “Explanation of the pain” and may characterize a pain as the first clinical symptoms of PAD.
  • the limit for the system may be reached when the weighting coefficient in “Explanation of the pain” is 6. In this case, the patient may automatically be flagged for further tests without further checking of other symptoms. If the score for this section may be below 6 (and optionally at least 2 or at least 3), then the system may look at weighting coefficients associated with Lesions, as outlined herein. Next, it may move to other sections and their scores and may compare them with previously assigned weighting factors.
  • At least one weighting coefficient for other sections gets higher over time, then the patient may be seen as more likely developing a PAD. If for example after 6 months of monitoring a health risk for a patient, at least one assigned weighting coefficient, associated with symptoms, increases in at least one section other than “Explanation of the pain”, then the patient may be flagged for further tests by a doctor. The overall history of symptoms progress along with type of pain, location of pain could be provided to the patient or their doctors.
  • the aforementioned weighting algorithm may best be suitable to be used by a patient who has a history of a heart and/or a diabetes disease and may not be implemented as a consumer app (e.g., the app may be used for (professional) monitoring purposes only and/or may be exclusively used by clinicians such as, e.g., doctors).
  • the symptoms weighting reports may be shared with doctor for review and decision making if a patient needs to be called in.
  • the weighting coefficient could be used for detecting a venous disease or a diabetes as both these diseases may have skin manifestation as well.
  • Tables 4 and 5 show two further exemplary question sets to be provided to a patient and associated weighting coefficients to be assigned to the received replies for a monitoring of a health risk for the patient according to the “VESSEL” scheme as outlined above. More specifically, Table 4 states several exemplary questions to be provided to a patient for monitoring whether a patient develops a PAD and/or whether an already existing PAD has deteriorated. In analogy to Table 4, Table 5 states several exemplary questions to be provided 5 to a patient for monitoring whether a patient develops a PVD and/or whether an already existing PVD has deteriorated. Moreover, the Tables 4 and 5 further indicates whether a question is directly related to PAD/PVD and potential weighting factors assigned to a response to a question provided to a patient.
  • Table 4 exemplary questions provided to a patient for monitoring a health risk for the patient
  • the following table shows a further exemplary question set provided to a patient for monitoring a health risk for the patient to monitor a potential venous disease (e.g, PVD).
  • a potential venous disease e.g, PVD
  • Table 5 exemplary questions provided to a patient for monitoring a health risk for the patient
  • the questions as outlined above may be provided to both (potential) PAD and (potential) PVD patients.
  • the weighting algorithm may be capable of assigning a diseased patient to a PAD and/or a PVD. It is noted that some symptoms are in common whereas other symptoms are completely different than others.
  • PAD weighting algorithm
  • PVD weighting assessment may be that the weighting algorithm may start with questions associated with a pain and may then move to Edema. Afterwards it may focus on other symptoms in comparison with PAD.

Landscapes

  • Health & Medical Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Pathology (AREA)
  • Primary Health Care (AREA)
  • Epidemiology (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Biophysics (AREA)
  • Veterinary Medicine (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • Dermatology (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

L'invention concerne, entre autres, un dispositif de surveillance d'un risque de santé pour un patient. Le dispositif (400) comprend un moyen (410), par exemple une caméra de téléphone intelligent, servant à obtenir des informations visuelles (430), par exemple une séquence vidéo, associées à la peau (420) du patient et un moyen (430), par exemple un moteur d'IA en nuage entraîné, servant à détecter une anomalie sur la peau du patient sur la base, au moins en partie, des informations visuelles obtenues, ainsi qu'un moyen (440) servant à attribuer une valeur de notation (445) à l'anomalie détectée, la valeur de notation étant associée au risque pour la santé du patient. La valeur de notation peut aider à diagnostiquer ou à prédire des maladies vasculaires périphériques.
PCT/EP2023/079161 2022-11-15 2023-10-19 Système fondé sur l'ia pour la surveillance de patients souffrant d'une maladie vasculaire périphérique WO2024104704A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
EP22207531.9 2022-11-15
EP22207531 2022-11-15

Publications (1)

Publication Number Publication Date
WO2024104704A1 true WO2024104704A1 (fr) 2024-05-23

Family

ID=84358206

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2023/079161 WO2024104704A1 (fr) 2022-11-15 2023-10-19 Système fondé sur l'ia pour la surveillance de patients souffrant d'une maladie vasculaire périphérique

Country Status (1)

Country Link
WO (1) WO2024104704A1 (fr)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2011087807A2 (fr) * 2009-12-22 2011-07-21 Health Discovery Corporation Système et procédé de dépistage de mélanome à distance
US20120065484A1 (en) * 2002-04-04 2012-03-15 Hull Edward L Determination of a Measure of a Glycation End-Product or Disease State Using Tissue Fluorescence
US20150230712A1 (en) * 2014-02-20 2015-08-20 Parham Aarabi System, method and application for skin health visualization and quantification
CN104887183A (zh) * 2015-05-22 2015-09-09 杭州雪肌科技有限公司 基于光学的肌肤健康监测和预诊断智能方法
US20200107732A1 (en) * 2017-03-22 2020-04-09 Modulated Imaging, Inc. Systems and methods for assessing diabetic circulatory complications
US20210275026A1 (en) * 2020-03-05 2021-09-09 International Business Machines Corporation Automatic measurement using structured lights

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120065484A1 (en) * 2002-04-04 2012-03-15 Hull Edward L Determination of a Measure of a Glycation End-Product or Disease State Using Tissue Fluorescence
WO2011087807A2 (fr) * 2009-12-22 2011-07-21 Health Discovery Corporation Système et procédé de dépistage de mélanome à distance
US20150230712A1 (en) * 2014-02-20 2015-08-20 Parham Aarabi System, method and application for skin health visualization and quantification
CN104887183A (zh) * 2015-05-22 2015-09-09 杭州雪肌科技有限公司 基于光学的肌肤健康监测和预诊断智能方法
US20200107732A1 (en) * 2017-03-22 2020-04-09 Modulated Imaging, Inc. Systems and methods for assessing diabetic circulatory complications
US20210275026A1 (en) * 2020-03-05 2021-09-09 International Business Machines Corporation Automatic measurement using structured lights

Similar Documents

Publication Publication Date Title
Peng et al. Quantifying fractal dynamics of human respiration: age and gender effects
EP3261524A1 (fr) Systèmes et procédés pour une mesure non effractive de la pression artérielle
US20150065826A1 (en) Noninvasive Predictive and/or Estimative Blood Pressure Monitoring
EP3386376A1 (fr) Systèmes et procédés de mesure non invasive de la tension artérielle
US10431343B2 (en) System and method for interpreting patient risk score using the risk scores and medical events from existing and matching patients
JP2017530849A (ja) 生理データに基づく患者の継続的な心拍出量（ｃｃｏ）を予測する方法およびシステム
JP2018534697A (ja) 個人化された予測モデルに基づいて健康モニタリングを容易化するシステム及び方法
CN111603136B (zh) 一种用于足部诊疗的远程会诊系统
WO2024104704A1 (fr) Système fondé sur l'ia pour la surveillance de patients souffrant d'une maladie vasculaire périphérique
CN116548926A (zh) 一种基于冷刺激的糖尿病足筛查系统及方法
US20220254502A1 (en) System, method and apparatus for non-invasive & non-contact monitoring of health racterstics using artificial intelligence (ai)
Fattah et al. Wrist-card: PPG sensor based wrist wearable unit for low cost personalized cardio healthcare system
Melillo et al. Cloud-based remote processing and data-mining platform for automatic risk assessment in hypertensive patients
US9084585B1 (en) System and method to classify left ventricular and mitral valve function
Cornelis et al. Near infrared spectroscopy to evaluate the effect of a hybrid exercise programme on peripheral muscle metabolism in patients with intermittent claudication: an exploratory PROSECO-IC sub study
US9980656B2 (en) Hybrid model and diagnostic process for atrial fibrillation
Paula-Ribeiro et al. Immediate and 24-h blood pressure-lowering effects of arm crank exercise in patients with traumatic lower-limb amputation: a randomized cross-over study
CN118315064A (zh) 乳腺癌术后心理健康检测系统、方法及存储介质
Buthelezi et al. IoT based health monitor for remote patient
US20230346234A1 (en) Wearable photoplethysmography device for detecting clinical decompensation based on heart rate variability
Chazaro et al. Myocardial infarction diagnosis by a neural network
Timokhov et al. Processing and Analysis of the Results of Long-Term Continuous Monitoring of the Health Status of Patients with Arterial Hypertension and Cardiac Arrhythmias
Lima et al. BLOOD PRESSURE CONTROL RATES PREVIOUS AND DURING THE COVID-19 PANDEMIC IN BRAZIL
Rey et al. AMBULATORY BLOOD PRESSURE MEASURMENT INTEREST IN GENERAL PRACTICE
Neuhauser et al. PLEASE TRY THIS AT HOME: STANDARDIZED BLOOD PRESSURE SELF-MEASUREMENT-WILLINGNESS TO PARTICIPATE AND FEASIBILITY STUDY

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23793782

Country of ref document: EP

Kind code of ref document: A1