Systems and methods for workload balancing and prediction

Abstract

A system for workload balancing and prediction based on a cognitive load of a user has a sensor and a processor configured to acquire personal data for each of a plurality of users, the personal data including one of a user identifying information and a user state determined based on sensing data acquired via the sensor, determine a cognitive load of each of the plurality of users based on the user state and generate an output based on the determined cognitive load and the user identifying information for each of the plurality of users.

Description

SYSTEMS AND METHODS FOR WORKLOAD BALANCING AND PREDICTIONBACKGROUND

[0001] Efficiency has been a primary goal of modem healthcare. Improving efficiency, however, generally increases pressure on radiologists, especially in the context of using a Picture Archiving and Communication System (“PACS”). A PACS is a medical imaging technology that provides economical storage and convenient access to images from multiple modalities (source machine types). More specifically, a PACS is a specialized apparatus for medical imaging that provides support for a radiologist in reading (e.g., accessing, processing, and / or analyzing) images created by different imaging modalities such as, but not limited to, Digital Radiography (DR), Computed Tomography (CT), Magnetic Resonance (MR), and Nuclear Medicine (NM). A PACS can be designed in many different ways.

[0002] An exemplary PACS technology infrastructure may include imaging device interfaces, storage devices, host computers, communication networks, and display systems often integrated by a flexible software package for supporting a radiologist in reading a patient case through a diagnostic workflow. Such a patient case is often referred to as an “image study.” Common specialized hardware components may include for example, patient data servers, data / modality interfaces, PACS controllers with database and archive, and display workstations connected by communication networks for handling and managing efficient data / image flow. A PACS is a synergy of specialized hardware and flexible software. The flexible software also includes various functions such as for example, but not limited to, measurement, segmentation, tumor or lesion identification, landmark detection, visualization, and reporting on findings.

[0003] Medical image studies are received / retrieved and / or transmitted digitally, e.g., over a secure network, and / or portable physical storage medium. Medical images are communicated to and / or from a PACS electronically / digitally via the Digital Imaging and Communications in Medicine (DICOM) protocol, which includes a file format definition and a network communications protocol and uses Transmission Control Protocol (TCP) / Internet Protocol (IP) TCP / IP to communicate between systems. Non-image data, such as a scanned document, may be incorporated, e.g., using formats such as Portable Document Format (PDF).

[0004] Patient data is typically handled as follows. In general, an examination order is transmitted from a Hospital Information System (HIS) to a Radiology Information System (RIS) via HL.7 and an imaging modality via DICOM. The patient is scanned with the imaging modality. The images are transmitted to a PACS Server. The images are assigned to a radiologist(s) and displayed in a worklist for a radiologist of the group who will read the images at a PACS. The radiologist loads and reads the images at the PACS and creates a report, which is saved (e.g., to the RIS). The report is transmitted to the HIS and / or Electronic Medical Record (EMR) via HL.7 and the images are transmitted to the HIS via DICOM.

[0005] Increasing pressure in combination with decreasing resources within the radiology department in hospitals have been shown to cause a significant increase in workload and stress on radiologists and technicians who use PACS technology. Thus, optimal support in workflow, balancing workload, and enhancing staff experience, where possible, has become increasingly important.

[0006] In practice, radiologists and radiological technologists generally have a very hectic day resulting from issues such as, for example, case scheduling switches due to urgency measures and distracting inquiries / interruptions. In some instances, images are read more than once, partially or entirely. For instance, a radiologist may need to temporarily leave the PACS, e.g., to attend to another matter. In such an instance, when the radiologist returns to the PACS to resume their reading of the image study, the radiologist may need to re-read part of the image(s) to recall where they left off. Sometimes alarming conditions and complexities arise in the radiology working context that require quick action. This can result in chaotic and stressful situations in which radiology staff may become overburdened and / or overwhelmed.

[0007] Current technological developments in radiology include improving reading efficiency. However, this may increase pressure and affect the overall reading experience of radiologists. This, in combination with the inconsistency and decrease in active staff within the radiology department in hospitals, can cause a significant increase in workload and stress on the radiologists. This is reflected in the increase in burnout rates of radiologists indicated in the“Statement of Support National Academy of Medicine Collaborative on Clinician Well-Being and Resilience,” American College of Radiology, 5 January 2018.

[0008] In view of at least the above, there is an unresolved need for an improved technological development s) for reading medical image data, and more specifically a need to improve current PACS technology or systems in the technical field of radiology to improve efficiency of the radiology staff without overburdening or overwhelming the radiology staff.SUMMARY

[0009] Some example embodiments are related to a computer-implemented method for providing workload balancing and prediction based on a cognitive load of a user. The method includes acquiring personal data for a plurality of users, the personal data including one of a user identifying information and a user state determined based on sensing data acquired via a sensor, determining a cognitive load of one or more of the users based on the user state and generating an output based on the determined cognitive load and the user identifying information for one or more of the users.

[0010] Other example embodiments are related to a system for workload balancing and prediction based on a cognitive load of a user. The system includes a sensor and a processor, the processor executing an executable program stored on a non-transitory computer readable storage medium to cause the processor to acquire personal data for a plurality of users, the personal data including one of a user identifying information and a user state determined based on sensing data acquired via the sensor, determine a cognitive load of the plurality of users based on the user state and generate an output based on the determined cognitive load and the user identifying information for the plurality of users.BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Fig. 1 shows a schematic drawing of a system according to an exemplary embodiment.

[0012] Fig. 2 shows another schematic drawing of the system according to Fig. 1.

[0013] Fig. 3 shows a table of a classification of speed factors according to an exemplary embodiment.

[0014] Fig. 4 shows a flow diagram of a method according to an exemplary embodiment.

[0015] Fig. 5 shows a schematic drawing of a system according to another exemplary embodiment.

[0016] Fig. 6 shows another schematic drawing of the system according to Fig. 5.

[0017] Fig. 7 shows a schematic drawing of an exemplary cognitive performance distribution of a plurality of users.

[0018] Fig. 8 shows a flow diagram of a method according to another exemplary embodiment.Detailed Description

[0019] The exemplary embodiments may be further understood with reference to the following description and the related appended drawings, wherein like elements are referred to with the same reference numerals. The exemplary embodiments relate to systems and methods for workload balancing and prediction. In particular, exemplary embodiments describe a system and method for gathering a user’s personal data and determining a case data for each of the cases to be reviewed to facilitate, for example, individual and / or department wide workload balance. A user’s personal data may include information from a user profile including, for example, a user’s expertise and experience, along with sensed data indicative of, for example, a user’s cognitive load including factors such as attentiveness, fatigue, and / or stress. Case data may include a complexity of the case to be reviewed. This information may be used in any of a variety of ways. In an exemplary embodiment, the user’s personal data and the case data may be used to calculate a reading time for each of the cases to be reviewed, which may be useful for determining worklist scheduling. In another exemplary embodiment, personal data of individual users may be used to determine imbalances in cognitive performance of users, department wide. Imbalances in cognitive performance may be addressed via, for example, personalized feedback,workload shifting within department, adjustment of individual worklist scheduling, and / or shifting of interruptions. It will be understood by those of skill in the art that although the exemplary embodiments show and describe the users as radiologists, a user may include any persons working within any of a variety of departments in which a user’s cognitive load (e.g., attentiveness, fatigue, stress) may be affected by workload.

[0020] As shown in Figs. 1 and 2, a system 100 according to an exemplary embodiment determines reading times of cases to be reviewed to facilitate, for example, workload balance and advanced worklist scheduling capabilities. The cases to be reviewed may include, for example, image exams (e.g., X-ray, MRI, CT) requiring reading via a user (e.g., radiologist). As shown in Fig. 1, the system 100 comprises a processor 102, a user interface 104, a display 106, and a memory 108. The system 100 may also include sensors 120 for determining a user identity and / or a user state. In an exemplary embodiment, the processor 102 may include a case review data module 110 for analyzing image exams to be reviewed, a personal data module 112 for gathering personal data based on the user’s identity and a sensed state of the user, and a reading time calculation module 114 for calculating a reading time of each image exam to be reviewed based on the derived personal data. In a further exemplary embodiment, the processor 102 may also include an output module 116 determining an output based on the calculated reading times and / or an optimization module 118 for optimizing calculation of the reading times of the image exams. The memory 108 may include a database of cases to be reviewed 126, a personal profile database 128, and a database of cases previously reviewed 130.

[0021] The case review data module 110 is configured to analyze case review data 122 of each case (e.g., image exam) to be reviewed. Each of the cases to be reviewed may be stored in the database 126 and may include case review data 122 associated therewith. The case review data 122 may include, for example, an urgency (e.g., urgency indicia or urgency factors), a type of image (e.g., X-ray, MRI, CT), a type of patient (e.g., patient condition or illness), and a request of the referring physician. The case review data module 110 may analyze the case review data 122 to determine a complexity / difficulty score based on, for example, an anatomical structure of the image exam, results of initial Al findings using diagnostic algorithms, and any corresponding confidence and / or abnormality scores via the diagnostic algorithm. Based on the analysis of thecase review data 122, the case review data module 110 determines a quantified complexity score between, for example, 1-7. According to an exemplary embodiment, a complexity score of 1 may indicate that the case is not very complex while a score of 7 may indicate that the case is very complex.

[0022] The personal data module 112 is configured to gather personal data 124 of an individual user, the personal data including, for example, user identification and user state indicative of cognitive load. In an exemplary embodiment, the user may be identified via a camera-based user identification method including, for example, badge recognition, facial recognition, digital profiles, etc. It will be understood by those of skill in the art, however, that user may be identified using any of a variety of methods including, for example, logging in to a user workstation using personalized credentials (e.g., login information). Each individual user will have an associated personal profile (which includes the personal data 124) that is continuously updated over time to include experience (e.g., seniority) and expertise (e.g., specialization) that is based on case review details of image exams previously read by the user. These personal profiles are stored in the personal profile database 128.

[0023] The user cognitive load may include factors such as, for example, attentiveness, fatigue, and stress, degrees of which may be determined and / or quantified. The user cognitive load may be determined via measurements derived from sensing data from the sensors 120, which may sense a user state including, for example, eye-gaze behavior (including pupil dilation), body and head posture, heart rate variability, skin conductance, facial expression, and distraction sensing. Thus, it will be understood by those of skill in the art that the personal data module 112 may receive sensor data from the sensors 120 to determine the user’s cognitive load.

[0024] The personal data module 112 may utilize existing Al models to determine a quantification of predicted user efficiency based on the user’s personal profile and the user state. In an exemplary embodiment, the user’s determined cognitive load (e.g., user’s state) and / or information included in the user’s personal profile (e.g., expertise, experience) may be used to determine an efficiency score of between, for example, 1-7, corresponding to user efficiency. For example, a score of 1 may indicate that the user is under greater cognitive load and / or haslittle experience reading a type of image and is thus less efficient and a score of 7 may indicate that the user is very experienced and / or under less cognitive load and may thus be more efficient.

[0025] As shown in Fig. 2, the reading time calculation module 114 calculates a predicted reading time for each of the cases to be reviewed based on the case review data 122 and the gathered personal data 124. In an exemplary embodiment, the user’s complexity score derived from the personal data 124 and the complexity score derived from the case review data 122 may be used to determine a review speed factor between, for example, 0.5 to 1 and 1 to 1.5. The review speed factor may be used to calculate a predicted and / or expected reading time for each case. The review speed factor may be based on, for example, the speed factor classification table shown in Fig. 3. It will be understood by those of skill in the art, however, that the speed factor classification table of Fig. 3 is exemplary only and that the speed factor classifications may be evaluated and optimized over time as predicted reading times are compared to actual reading times.

[0026] The review speed factors may be used to calculate the predicted (or expected) reading times using the below formula:Expected Reading Time = ART (Average Reading Time)(ERT) ASF (Analyzed Speed Factor)

[0027] In particular, the predicted or expected reading time is calculated based on an average reading time of similar cases and the determined speed factor. The average reading time may be determined by analyzing cases stored in the previous cases database 130. According to the speed factor classification table shown in Fig. 3, a user with an efficiency score of 5 will read a case having a complexity score of 3 with a review speed factor 1.2. To provide an example, it may be considered that the average reading time of similar cases is 790 seconds (e.g., 13 min and 10 sec), the predicted reading time of the case for the user is 650 seconds, as shown below:790 sec. (ART) / 1.2 (x5;y3 ASF) = 650 sec (ERT)650 sec. = 10 min. and 50 sec. (compared to 13 min and 10 sec. (ART))Thus, in this example, the user is predicted to be able to read the case faster than the average read time of similar cases.

[0028] The calculated reading time for each of the cases to be reviewed by the user in the current user state may be utilized in any of a variety ways. The output module 116 may generate output based on the predicted reading times to facilitate, for example, individual worklist scheduling, department workload prediction and balancing, predictive analytics assessments, decreasing a mental or cognitive load of the user, educational services (e.g., ensuring maintenance of expertise), etc. In an exemplary embodiment, the output module 116 may simply display predicted reading times for all pending cases to be reviewed on the display 106 so that the user may utilize the calculated reading times to manage their own worklist schedule. For example, a user may prefer to read or review some “short” cases before diving into complex cases (e.g., cases that are predicted as taking longer to read). The worklist schedule may be managed via the user interface 104. The user interface 104 may include any of a variety of input devices such as, for example, a mouse, a keyboard and / or a touch screen via the display 106.

[0029] In another embodiment, however, the output module 116 may automatically generate individualized worklist schedules based on the user preferences. User preferences may include, for example, scheduling the cases from shortest predicted reading time to longest predicted reading time. In another embodiment, user preferences may include a desired level of complexity and / or predicted reading time so that users may ensure they are maintaining a desired level of experience and / or expertise. User preferences may also be input via the user interface 104.

[0030] In another embodiment, the output module 116 may determine department workload prediction and balancing. Workload may be assigned to individual users based on predicted reading time and individual user state(s). In another embodiment, the output module 116 maydetermine department-wide scheduling such as, for example, when breaks may be taken and / or whether additional users may be necessary at any given time to review / read cases.

[0031] In yet another embodiment, the output module 116 may provide predictive analytics to, for example, an administrator for the department, so that the administrator may determine staff and / or user scheduling based on the predictive analytics. The analytics may also indicate when individual users perform the best and / or when the individual users may need a break so that administrators may schedule accordingly. Alternatively, the predictive analysis may also aid in determining timing and scheduling of particular types of image scans.

[0032] The optimization module 118 compares the predicted reading time with actual reading times of each of the cases so that a difference therebetween is translated to a coefficient to correct and / or improve the review speed factor. As shown via the feedback loop in Fig. 2, upon review and / or reading of the cases, the cases and their actual reading times may be stored to the database of previous cases 130. The corrected speed factor may be stored to the speed factor classification table of the reading time calculation module 114. It will be understood by those of skill in the art, however, that although the speed factor classification is described as being stored to the reading time calculation module 114, in another embodiment the speed factor classifications may be stored in a separate database stored to the memory 108, which may be accessible via the reading time calculation module and / or the optimization module 118.

[0033] As described above, the personal data 124 (e.g., user state, user cognitive load) gathered via the personal data module 112 may be determined via measurements derived from the sensors 120 and stored in memory 108 (e.g., as part of the personal profile database 128) or sent directly to the processor 102 for further processing. One or more of the sensors 120 may be distributed throughout a workspace (e.g., cameras located overhead of a user) and / or may include a variety of hardware and software that can generate sensor data such as for example, but not limited to, microphone, light sensor, accelerometer, gyroscope, camera, heart rate sensor, and skin conductance sensor. In an exemplary embodiment, one or more of the sensors 120 may be configured as a wearable device. It will be understood by those of skill in the art that while the wearable device may include multiple sensors such as a heart rate sensor and skin conductancesensor, these sensors may also exist or be independently used in different or separate wearables or devices. One or more of the sensors 120 may also sense data including, for example, one or a combination of eye-gaze behavior, posture, heart-rate variability, skin conductance, and distraction sensing of a user. It will also be understood by those of skill in the art that, in some embodiments, the sensors 120 may also be configured for user identification (e.g., badge recognition, facial recognition).

[0034] The processor 102 may be configured to execute computer-executable instructions for operations from applications that provide functionalities to or for the system 100. For example, the case review module 110 may include instructions for analyzing case review data of a case to be reviewed to determine a complexity score thereof, the personal data module 112 may include instructions for acquiring personal data of the user from memory 108 including personal information and user state to determine an efficiency score, and the reading time calculation module 114 may include instructions for calculating a predicted / expected reading time for each of the cases to be reviewed based on the complexity of the case and the user’s predicted efficiency. Similarly, the output module 116 may include instructions for generating a desired output based on the calculated reading times and the system optimization module 118 may include instructions for optimizing, for example, the speed factor classifications. It should be noted, however, that functionalities described with respect to the case review module 110, the personal data module 112, the reading time prediction module 114, the output module 116 and the optimization module 118 may also be represented as a separately incorporated component of the system 100, a modular component connected to the processor 102 or as a functionalities achievable via more than one processor 102. For example, the system 100 may be comprised of a network of computing systems, each of which includes one or more of the components described above.

[0035] System 100 may be for example a PACS. System 100 may also be a teleradiology system or workstation (including for example a PACS) operating remotely from a hospital and using the cloud for communication with hospital systems such as imaging modalities. In other embodiments, system 100 may also be a specialized console workstation at the imaging modality. An example of a suitable PACS includes for example, but is not limited to a systemsuch as Philips® Intellispace® PACS, Philips VuePACS®, successor PACS products, or similar. The PACS may be an advanced visualization system. The PACS may also be a system like Philips® Intellispace® Portal. System 100 may also be a group of hardware and software modules, some of which are incorporated into a PACS while others exist separate, but in communication with the PACS-integrated modules in any number of permutations of possible combinations of integrated and external modules. System 100 may also include an Application Specific Integrated Circuit (“ASIC”) or specialized semiconductor or other hardware, specially configured or adapted (e.g., based on specialized instructions) for carrying out any number of functions of the modules described herein. System 100 (and related methods and functionalities described herein) embodied as a PACS represents a significant improvement to PACS technology (e.g., one or more of PACS workflow, PACS workload assignment, PACS workload balancing, PACS predictive analytics, etc), including for example improvements to Philips® Intellispace® PACS, Philips VuePACS®, successor PACS products, or similar. Indeed, both Philips® Intellispace® PACS and Philips VuePACS have some sort of PACS workflow, PACS workload assignment, or PACS workload balancing features that would be improved by the present PACS embodiment and related methods or functionalities. System 100 (and related methods and functionalities described herein) embodied as a PACS also represents a significant improvement to the technical field of radiology and radiology arts.

[0036] In addition, although the exemplary embodiments show and describe the databases 126- 130 as being stored to the memory 108, it will be understood by those of skill in the art that the data described with respect to the databases 126-130 may be stored to a plurality of devices connected to and accessible via the system 100 via, for example, a network connection for receiving and transmitting information. As is well known and previously described in the case of PACS, communication (e.g., acquisition of information) can be electronic / digital via for example the Digital Imaging and Communications in Medicine (DICOM) protocol, which includes a file format definition and a network communications protocol and uses Transmission Control Protocol (TCP) / Internet Protocol (IP) TCP / IP to communicate between devices or systems. In one exemplary embodiment, the current cases to be reviewed, the user profiles, and / or the previous case studies may be acquired from one or more remote and / or network memories and stored to a central memory 108. Alternatively, the current cases to be reviewed, the user profiles,and / or the previous case studies may be collected and stored to any remote and / or networked memory.

[0037] Fig. 4 shows an exemplary method 200 for calculating a predicted reading time for cases (e.g., image exams) to be reviewed based on a complexity of the case and user data 124. In 210, case review data 122 of each image exam to be reviewed is analyzed to determine a complexity / difficulty score based on, for example, an urgency, a type of image (e.g., X-ray, MRI, CT) and / or an anatomical structure of the image exam, a type of patient (e.g., patient condition or illness), a request of the referring physician, any preliminary Al findings using diagnostic algorithms and any corresponding confidence and / or abnormality scores associated with the Al findings. As described above, in an exemplary embodiment, the complexity score may be a quantified score between the values of 1 and 7.

[0038] In 220, personal data 124 of the user is gathered from both the personal user profile including information such as, for example, expertise and experience, and a user state measured via sensors 120. The user state may be indicative of cognitive load including factors such as attentiveness, fatigue, and stress. The personal data 124 is used to determine the user efficiency score which, in an exemplary embodiment, is a quantified score between the values of 1 and 7. A score of 1 may indicate that the user is under greater cognitive load and is thus less efficient and a score of 7 may indicate that the user is under less cognitive load and may thus be more efficient.

[0039] In 230, the predicted reading time of each of the current cases to be reviewed is calculated based on the case review data 122 and the personal data 124. In an exemplary embodiment, the complexity score and the efficiency score derived in steps 210, 220, along with an analysis of reading times of similar cases, may be used to calculate the predicted reading time of each case. In an exemplary embodiment, the predicted reading time may be calculated using the following formula:Expected Reading Time = ART (Average Reading Time) (ERT) ASF (Analyzed Speed Factor)

[0040] In 240, the predicted reading times of all pending cases to be reviewed may be displayed on the display 106 so that the user may manage their own worklist schedule, as desired. Alternatively, or in addition, outputs related to individual worklist scheduling, department workload prediction and balancing, predictive analytics assessments, decreasing a mental or cognitive load of the user, educational services, etc. may be displayed and / or transmitted to administrators and / or individual users.

[0041] In 250, upon reading one or more of the cases to be reviewed, the case(s) are stored to the database of previously reviewed cases 130. The predicted reading time may be compared with the actual reading times the cases so that a difference therebetween is translated to a coefficient to correct and / or improve the review speed factor, thereby optimizing calculation of the predicted reading time.

[0042] As shown in Figs. 5 and 6, a system 300 may be substantially similar to the system 100 described above and, according to an exemplary embodiment, automatically or semi- automatically (e.g., with minimal interaction with a user or administrator) aligns and predicts cognitive load levels of a plurality of users (e.g., radiologists) to automatically or semi- automatically (e.g., with minimal interaction with a user or administrator or by for example providing suggestions to a user or administrator) distribute workloads and / or provide personalized suggestions based on the cognitive loads of individual users, in real time. In other words, the system 300 may be utilized to provide department- wide workload and / or cognitive load (e.g., performance) balance automatically or semi-automatically (e.g., minimal interaction with a user or administrator or by for example providing suggestions to a user or administrator). This process can automatically and continuously update balancing of workload among users. As shown in Fig. 5, the system 300 may be substantially similar to the system 100, as described above, comprising a processor 302, a user interface 304, a display 306, and a memory 308. The system 300 may also include sensors 320 for determining a user identity and / or a user state. Similar to the system 100 described above, in an exemplary embodiment, the processor 302 may include a personal data module 312 for gathering personal data for each of the plurality of users,the personal data including, for example, information included in the user’s personal profile and the user’s state. The processor 302, however, further comprises a cognitive performance module 332 for calculating cognitive performance scores for each of the users and determining whether there are cognitive imbalances between the users, in real time. Similar to the system 100, the memory 306 may also include a database of cases to be reviewed 326, a user profile database 328, and / or a database of previously reviewed cases 330.

[0043] As shown in Fig. 7, imbalances between users can be recorded in parallel so that when significant performance imbalances occur, a suggested follow-up action for balancing the cognitive performance may be provided via a workload balancing module 334. According to a further exemplary embodiment, the processor 302 may additionally comprise a long-term effects module 336 and / or a communication module 338, as will be described in further detail below.

[0044] The personal data module 312 may be substantially similar to the personal data module 112, the personal data module 312 being configured to gather personal data of each of the plurality of users. As described above with respect to system 100, the personal data may include, for example, user identification and user cognitive load. User identification information may include data associated with the user profile including, for example, a user’s experience and expertise. The user state may include measurements acquired using the sensors 320, which may be indicative of cognitive load including, for example, attentiveness, fatigue, and stress. As described above with respect to the system 100, the sensors 320 may include, for example, cameras and / or a variety of hardware and software that can generate sensor data such as for example, but not limited to, a microphone, light sensor, accelerometer, gyroscope, camera, heart rate sensor, and skin conductance sensor. In an exemplary embodiment, one or more of the sensors 320 may be included in a wearable device.

[0045] The user cognitive load may be determined via measurements derived from sensing data from the sensors 320, which may sense a user state including, for example, eye-gaze behavior (including pupil dilation), body and head posture, heart rate variability, skin conductance, facial expression, and distraction sensing. In an exemplary embodiment, Galvanic Skin Response (GSR) may be utilized to sense hormone composition in sweat particles on the skin, which maybe indicative of stress levels. Thus, stress induced cognitive impairment may be estimated approximately 1 hour ahead of time.

[0046] The cognitive performance module 332 may process the user state data including, for example, the stress induced cognitive impairment data from the GSR. The cognitive performance module 332 may determine a cognitive performance score for each of the users based on their user state data and / or user profile information and compare the cognitive performance scores of the plurality of users, in real time. Thus, the cognitive performance module 332 is able to identify imbalances in cognitive performances between users within, for example, a department (e.g., radiology department). In an exemplary embodiment, the stress induced cognitive impairment may be an inverse variable compared to cognitive performance. As shown in Fig. 7, cognitive load imbalances of each of the users may be recorded in parallel so that a significant depletion of one of multiple cognitive performance balances happens, a suggested follow-up action is suggested on the system that is running stationary on all computers in the radiology department. This also facilitates accessible and easy communication to directed users.

[0047] The workload balancing module 334 processing a signal for one or more follow-up actions based on a cognitive load imbalance determined via the cognitive performance module 332. Providing the follow-up actions may be an automated process and may include, for example, suggesting breaks and giving personal feedback to individual uses, assigning breaks, shifting workload to other users with a specific assignment of a case(s) or worklist to a different user, adjusting complexity of worklist, and forwarding interruptions to other users to prevent some users from being overloaded if their cognitive performance is low, as will be described in detail below.

[0048] While the follow-up actions are intended to balance cognitive performance across all users of the department, suggesting breaks and giving personal feedback is a personally directed action. In this embodiment, the workload balancing module 334 provides feedback to a user with low cognitive performance to take a break. Over time, the processor 302 learns each user’s preferences with respect to regaining cognitive performance and is able to provide more specificbreak suggestions. These preferences may be manually input by the individual user via the user interface 304, or, in another embodiment, may be inferred via the processor based on the measured impact via the sensors 320.

[0049] In another embodiment, the workload balancing module 334 may shift workload to other users (e.g., radiologists, technicians) across the department. Users with “excessive” cognitive performance may be capable of handling more cases based on their current cognitive state (e.g., higher mental capacity), thereby allowing users with lower current cognitive capacity to restore their cognitive capacity to at least a baseline level.

[0050] In another embodiment, the workload balancing module 334 may shift a composition and / or order of the worklist of an individual user with lower cognitive load. For example, the user may desire to review some simpler cases to regain some cognitive performance before diving into a more complex case. The worklist may be ordered based on, for example, a reading time for each of the cases, as discussed above with respect to system 100.

[0051] In another embodiment, when a user’s cognitive performance is low, the workload balancing module 334 may block interruptions (e.g., phone calls) to the user, forwarding or redirecting them to users with higher cognitive performance.

[0052] It will be understood by those of skill in the art that although each of the follow-up actions described above are described as separate embodiments, the workload balancing module 334 may provide one or more follow up actions. It may be iterative as well. Follow-up actions may be simultaneously provided or, in another embodiment, provided in combination so that when a first follow-up action does not result in a balancing of cognitive performance across users, a second follow-up action is provided. For example, the workload balancing module 334 may first adjust a complexity of an individual user’s worklist. However, if this worklist adjustment fails to increase the user’s cognitive performance, the workload balancing module 334 may provide a second follow-up action - e.g., shifting workload to other users.

[0053] According to a further exemplary embodiment, in addition to reducing workload and / or cognitive performance imbalances, the system 300 may store user data to, for example, the user profile database 328, so the data may be analyzed via the long-term effects module 336 for patterns which may be useful in determining long term measures that may improve the overall resilience of the department. The communication module 338 may facilitate communication between the system 300 and the head of the department.

[0054] The long-term effects module 336 gathers performance data, records patterns and trends in the data, and predicts long term effects. Individual performance data may be saved to, for example, to the user profile database 328. The long-term effects may be predicted via data analysis of the growing data, which may become increasingly valuable and insightful as the amount of data becomes larger. One example of a pattems / trends that may be identified includes specific cognitive spikes in days, weeks, or months per year which may correlate to external factors that may influence cognitive performance such as, for example, increased workload due to holidays. Another example of pattems / trends may include an average cognitive performance on specific types of cases.

[0055] The system can also be trained through machine learning, a neural network, or other artificial intelligence to optimize assignment and workload balancing based collection of data in an operational system database including instances of system improvement from system actions (e.g., reassignments, blocking or forwarding of interruptions, or other actions taken) and instances of system failure based on system actions (e.g., reassignments, blocking or forwarding of intermptions, or other actions taken). The system can also be trained through machine learning, a neural network, or other artificial intelligence to detect patterns, and particularly factors (external or otherwise) that affect workloads, assignments, and workloads balancing, and automatically take actions to optimize assignment of work and / or workload balance. Acquired data of patterns and specifically factors (e.g., environmental or industry factors) can be collected and stored in a workload interruption database. Over time, this database can also be modified based on changes in the environment, changes in factors, improvements based on system actions (e.g., reassignments, blocking or forwarding of interruptions, or other actions taken) and / orfailures based on system actions (e.g., reassignments, blocking or forwarding of interruptions, or other actions taken).

[0056] The operational system database and / or interruption database and / or modified data set(s) (or any combination thereof) may be used to train a model, neural network, or other Al in one or more stages to improve workload assignment or workload balancing, outcomes in general, and / or automated detection of patterns or factors that affect workloads, assignments, and workload balancing. The model can be used to guide the system in changing or provide additional automated actions through an iterative or optimized artificial intelligence or machine learning approach.

[0057] For example, an artificial intelligence module (embodied for example as an ASIC, semiconductor chip with instructions, instructions executable by a processor, etc.) trains a model for example with the interruption database and modified data set(s) as training data through one or more stages. The trained model can be applied to processing of new data (e.g., current realtime data) for taking actions. New data can include for example, user state, efficiency score, predicted read times, detected patterns, detected external factors, etc.). Examples of suitable Al algorithms for training include, but are not limited to, different types of neural networks, deep neural networks or “deep learning” models, or for example, sequence to sequence models. Throughout this description, the term “neural network” is used to describe a plurality of processing nodes that are densely interconnected. Sometimes, the neural network is organized into layers of nodes, but it is not a requirement. In a layer model, for example, a node may be connected to one or more nodes in a lower layer, from which it receives data, and one or more nodes in a higher layer, to which it sends data. It should also be understood that a neural network is a subset of machine learning, which is a method of data analysis that automates analytical model building.

[0058] Thus, throughout this description, it should be understood that use of a neural network is only exemplary, and any functions or operations described as being performed by a neural network may be performed by any type of machine learning and should not be interpreted as being limited to only a neural network. Some examples of neural networks include feed-forwardneural network, Radial Basis Function (RBF) Neural Network, Multilayer Perceptron, Convolutional Neural Network (i.e., densely connected neural networks, residual neural networks, networks resulting from architecture search algorithms, capsule networks etc.), Recurrent Neural Network(RNN), Modular Neural Network, and any similar types of algorithms which are known in the art or are suitable for the purpose. Such an Al module can exist as part of for example a PACS or separate from a PACS and may be used for training in a location remote and separate from the PACS, where a trained Al model may be implemented in the PACS through various types of connections or communication known in the art.

[0059] The communication module 338 provides feedback back to the head of radiology about the developments and trends found in the analysis of the user cognitive performance / load data, as well as current developments in real time. This way, the head of radiology is informed with the status (long and short term) of the entire department. Additionally, since the longer-term effects are also predicted and shown, the department head may be able to plan for appropriate actions in response to problematic developments. For example, after implementation of the system 300, the head of radiology may desire to approve any follow-up actions suggested via the system 300 in order to minimize the imbalance in cognitive performance. Once accuracy of the system 300 has sufficiently increased, the system 300 may make its own “decisions”. Alternatively, other users or staff members may be assigned to take control on the approval / rej ection of certain suggestions that are made.

[0060] Similar to the processor 102, the processor 302 may be configured to execute computerexecutable instructions for operations from applications that provide functionalities to or for the system 300 - e.g., to the personal data module 312, the cognitive performance module 332, the workload balancing module 334, the long-term effects module 336, and the communication module 338. It should be noted, however, that functionalities described with respect to each of the modules 312, 332, 334, 336 may also be represented as a separately incorporated component of the system 300, a modular component connected to the processor 302 or as a functionalities achievable via more than one processor 302. It will also be understood by those of skill in the art that although not shown or described with respect to the system 300, the modules described withrespect to the system 100 may be similarly incorporated into the system 300 to provide functionalities described with respect thereto.

[0061] Fig. 4 shows an exemplary method 400 utilizing the system 300 for balancing workload and / or cognitive performance balance based on a determined imbalance in cognitive performance between users within, for example, a single department (e.g., radiology). In 410, personal data may be gathered from memory 108 for each of the users of a plurality of users within the department via the personal data module 312. The personal data may include user identifying information such as, for example, the user’s expertise and experience, along with a user state. The user state may be determined based on data acquired from the sensors 320, which may be indicative of cognitive load / performance such as, for example, attentiveness, fatigue, and stress.

[0062] In 420, the personal user data acquired in 410 may be processed to determine a cognitive load / performance for each of the users based on their user state data and / or user profile information. In an exemplary embodiment, the processor 302 may determine a cognitive performance score for each of the users and compare the cognitive performance scores of the plurality of users, in real time. The cognitive scores for each of the users are measured in parallel so that the cognitive performance module 332 is able to identify imbalances in cognitive performances between users within, for example, a department (e.g., radiology department), in real time.

[0063] In 430, based on the determined cognitive imbalance, the workload balancing module 334 may provide a follow-up action to one or more of the users. The follow-up action may be directed to individual users having low cognitive performance and / or may be directed to a plurality of users to balance the cognitive load between users. Follow-up actions may include, for example, suggesting breaks and giving personal feedback to individual users, shifting workload to other users, adjusting complexity of worklist, and forwarding interruptions to other users. As described above with respect to the system 300, more than one follow-up action may be provided, if desired.

[0064] According to a further exemplary embodiment, in 440, cognitive scores of individual users and / or department wide imbalances may be stored to a database so that the data may be further analyzed for long term effects. In an exemplary embodiment, individual cognitive scores may be stored to the user profile database 328. It will be understood by those of skill in the art, however, that cognitive data of users may be stored to a separate database which records data for each of the plurality of users, department-wide, and any determine imbalances between users. Changes in cognitive performance / load after implementation of a follow-up action may also be recorded.

[0065] In 450, the long-term effects of individual user cognitive data and / or imbalances between users may be predicted via big data analysis of the growing data. In an exemplary embodiment, the long-term effects module 336 may, for example, identify cognitive changes / patterns correlating to, for example, external factors and / or types of cases, to determine long-term effects.

[0066] In 460, the long-term effects and / or patterns identified in 450 and / or currently determined cognitive performance parameters of individual users and / or department wide imbalances may be communicated to, for example, the head of the department. The head of the department may take appropriate action based on identified patterns.

[0067] Those skilled in the art will understand that the above-described exemplary embodiments may be implemented in any suitable software or hardware configuration or combination thereof. An exemplary hardware platform for implementing the exemplary embodiments may include, for example, an Intel x86 based platform with compatible operating system, a Windows OS, a Mac platform and MAC OS, a mobile device having an operating system such as iOS, Android, etc. In a further example, the exemplary embodiments of the above-described method(s) may be embodied as a program containing lines of code stored on a non-transitory computer readable storage medium that, when compiled, may be executed on a processor or microprocessor.

[0068] Although this application described various aspects each having different features in various combinations, those skilled in the art will understand that any of the features of one aspect may be combined with the features of the other aspects in any manner not specificallydisclaimed or which is not functionally or logically inconsistent with the operation of the device or the stated functions of the disclosed aspects.

[0069] It is well understood that the use of personally identifiable information should follow privacy policies and practices that are generally recognized as meeting or exceeding industry or governmental requirements for maintaining the privacy of users. In particular, personally identifiable information data should be managed and handled so as to minimize risks of unintentional or unauthorized access or use, and the nature of authorized use should be clearly indicated to users.

[0070] It will be apparent to those skilled in the art that various modifications may be made in the present disclosure, without departing from the spirit or the scope of the disclosure. Thus, it is intended that the present disclosure cover modifications and variations of this disclosure provided they come within the scope of the appended claims and their equivalent.

Claims

What is claimed is:

1. A computer-implemented method for providing workload balancing and prediction based on a cognitive load of a user, comprising the steps of: acquiring personal data for a plurality of users, the personal data including one of a user identifying information and a user state determined based on sensing data acquired via a sensor; determining a cognitive load of one or more of the users based on the user state; and generating an output based on the determined cognitive load and the user identifying information for one or more of the users.

2. The method of claim 1, further comprises: analyzing case data for one or more cases to be reviewed to determine a complexity of the one or more cases; and calculating a predicted reading time of the one or more cases based on a speed factor determined via the complexity of the case and the cognitive load of one or more of the of users.

3. The method of claim 2, wherein the speed factor used to calculate the predicted reading time is optimized over time by comparing an actual reading time of the cases with the calculated predicted reading time thereof.

4. The method of claim 2, wherein the complexity of the one more cases is determined via one or more of an anatomical structure thereof, an initial diagnostic finding, an urgency, and a type of image.

5. The method of claim 1, wherein the user identifying information includes an expertise and experience level of the one or more users.

6. The method of claim 1, wherein the user state includes one of attentiveness, fatigue, and stress, degrees of which may be determined via measurements derived from the sensing data acquired via the sensor.

7. The method of claim 1, wherein the sensing data includes one of eye-gaze behavior, body and head posture, heart rate variability, skin conductance, facial expression, and distraction sensing.

8. The method of claim 1, further comprising comparing a cognitive load of a plurality of users to one another to identify a cognitive performance imbalance between users so that the output addresses the cognitive performance imbalance.

9. The method of claim 1, wherein the output includes one of suggesting breaks and providing personal feedback, shifting workload between users, adjusting the complexity of individual worklists, and forwarding interruptions to other users.

10. The method of claim 1, further comprising identifying patterns in the cognitive load of the one or more users, over time.

11. A system for workload balancing and prediction based on a cognitive load of a user, comprising: a sensor; and a processor, the processor executing an executable program stored on a non- transitory computer readable storage medium to cause the processor to: acquire personal data for a plurality of users, the personal data including one of a user identifying information and a user state determined based on sensing data acquired via the sensor; determine a cognitive load of the plurality of users based on the user state; and generate an output based on the determined cognitive load and the user identifying information for the plurality of users.

12. The system of claim 11, wherein the processor executes the executable program to cause the processor to: analyze case data for a plurality of cases to be reviewed to determine a complexity of the plurality of cases; and calculate a predicted reading time of the plurality of cases based on a speed factor determined via the complexity of the case and the cognitive load of the plurality of users.

13. The system of claim 12, wherein the processor executes the executable program to cause the processor to: optimize the speed factor over time by comparing an actual reading time of the cases with the calculated predicted reading time thereof.

14. The system of claim 12, wherein the processor executes the executable program to cause the processor to determine the complexity of the cases based on one or more of an anatomical structure thereof, an initial diagnostic finding, an urgency, and a type of image.

15. The system of claim 11, wherein the user identifying information includes an expertise and experience level of the users.

16. The system of claim 11, wherein the user state includes one of attentiveness, fatigue, and stress, degrees of which may be determined via measurements derived from the sensing data acquired via the sensor.

17. The system of claim 11, wherein the sensor is configured to sense one of eye-gaze behavior, body and head posture, heart rate variability, skin conductance, facial expression, and distraction sensing.

18. The system of claim 11, wherein the processor executes the executable program to cause the processor to compare a cognitive load of the plurality of users to one another toidentify a cognitive performance imbalance between users so that the output addresses the cognitive performance imbalance.

19. The system of claim 11, wherein the output includes one of suggesting breaks and providing personal feedback, shifting workload between users, adjusting the complexity of individual worklists, and forwarding interruptions to other users.

20. The system of claim 11, wherein the processor executes the executable program to cause the processor to identify patterns in the cognitive load of the plurality of users, over time.

Images