₁and R₂values over a period of approximately one month, and with the subject returning to play in vigorous activity at respective numbers of days.

FIG. 10 is a similar series of graphs showing another subject with concussion, this one returning to play early during the period of concussion.

FIG. 11 is a graph representing percentage of subjects of a large group indicating biometric concussion over a month of time.

FIG. 12 is graph showing percentage of concussion subjects who appear to achieve concussion free status based on NSI values, and with a lower line showing percentage achieving biometric indication of concussion free status, over a period of one month.

DESCRIPTION OF PREFERRED EMBODIMENTS

FIG. 1 schematically shows a system 10 according to the invention, indicating a subject's head in profile at 12, and with the subject's brain indicated, also in dashed lines, at 14. The system 10 includes a headset 16 worn on the head in the manner similar to a hair band. On the headset 16 is an accelerometer 18, retained securely to the headset and in a position near the subject's temple, for determining a HeadPulse signal generated by a series of heartbeats indicated schematically at 20, 22 and 24, from contractions of the user's heart, indicated at 26. The pulsing of the heartbeat in the head causes slight physical movements of the head, i.e. of the skull, in X, Y and Z orthogonal directions, those directions being the vertical direction, anterior-posterior direction and the lateral direction. Accelerations are schematically indicated by arrows 28 in the drawing. Accelerations in one, two or all of these directions can be sensed with the accelerometer 18, with a multi-axis accelerometer preferably provided (although a single axis accelerometer can be used). An important aspect is that acceleration data can be taken at a single location on the head.

The headset is preferably wireless and thus includes a battery indicated at 30, although a non-wireless form of the headset could be provided, receiving power through a cable. The wireless device includes a wireless data transmitter or interface 32, sending data wirelessly to an analyzer 34 which could be, for example, contained in an app of a smart phone nearby. Alternatively, the analyzer 34 can be contained in the headset, and processed data can then be sent by cable or wirelessly to a smart phone for indication of analysis and results. In another embodiment data could be stored in memory storage on the headset itself, to be retrieved at an appropriate time. The data could be stored in a removable memory such as a card, cartridge or memory stick.

The drawing also indicates that examination or clinical data 36, obtained contemporaneously through other observations or instrumentation, can be fed to the analyzer 34 for correlation with the HeadPulse signal and any transforms, typically to the frequency domain.

As part of the system 10 of the invention, a high level or gross movement accelerometer is also included on the headset device, as indicated at 38. This gross movement accelerometer has a function of sensing and recording high level movements as from vigorous exercise, typically in the range of 0.5 g to 20 g of acceleration as discussed above. This information, which is fed to a high level analyzer 40 and also communicated to the analyzer 34, correlates a subject's activity and events of that activity with the analyzed headpulse signals, for purposes discussed above, i.e. to determine whether a return to play (RTP) has caused biometric indication of concussion recurrence. The high level accelerometer 38 can be placed anywhere on the headset. As an alternative, as discussed above, the high level accelerometer 38 could be on a separate device such as on a FitBit device worn on the wrist or arm. Further, it could be combined with the sensitive accelerometer 18 provided a wide range accelerometer is used. In any event, the HeadPulse signals and the high-level, gross acceleration signals are separate streams of data from the headset.

FIG. 2 shows the HeadPulse headset device 16 in a preferred embodiment. Two circuit boards are mounted on a headset similar to a plastic hairband, and interconnected by a ribbon cable extending around the band. A protective cap 42 is shown on one side, but the cap is removed at the right side for purposes of clarity. Circuit boards each contain a sensitive three-axis accelerometer in this embodiment, sampled at 250/second and stored on a micro-SD card that slides into the unit. The device has a Bluetooth interface to a custom smart phone app, indicated in FIG. 3, which also includes NSI (Neuro-behavioral System Inventory) queries. In one embodiment the subject is required to respond to all questions posed by the app with each biometric recording with HeadPulse. This system is used with the participating subjects in the study reported below. Instructions were displayed to the subject as on the leftmost panel, FIG. 3A. The NSI question responses are cell-reported using a 5 point Likert scale ranging from 0 to 4, 0 indicating “non” or no systems and 4 indicating very severe symptoms for each question. Overall symptom severity was calculated as the sum of the values provided by the subject with possible score range of 0 to 88.

The head-attachable device can take different forms. For example, it could simply be a patch (e.g. an adhesive patch) that contains both accelerometers (or at least the sensitive accelerometer) and Bluetooth or other wireless technology, along with a small battery. The data can be wirelessly sent to the cloud for processing, or the processing can occur on a smart phone or on the device itself. Note that data could be sent from the head-attached device, or from a smart phone or other device, directly to a medical professional to evaluate a patient's recovery or to provide a record of the recovery with the correlation of the high g accelerometer data and the HeadPulse, low g accelerometer data.

Concussion Study with Australian Athletes

This was a prospective cohort study. Subjects were enrolled in feasibility (A1) and validation (A2) phases. A1 was recruited between Aug. 5, 2021, and Sep. 10, 2021. A2 was recruited between May 5, 2022, and Sep. 3, 2022. The study evaluated a longitudinal cohort of concussed and control athletes of amateur Australian Rules Football (ARF) to investigate the time course of HeadPulse signal changes seen in recently concussed athletes (Australian University of California San Francisco concussion study in athletes, AUSSIE-1, A1). The model was then validated in a second cohort (AUSSIE-2, A2). Athletes in ARF do not wear padding or head protection. Concussion is a frequent complication. Recruitment occurred within the Adelaide Football League, the highest level of amateur Australian Rules Football in the state of South Australia, Australia. Adult male or female athletes participated. Controls were concussion free. All subjects provided consent at season start. HeadPulse measurements were obtained using a lab developed headset utilizing cranial accelerometry to transduce head forces (the headpulse) generated by cardiac contraction. Subjects were followed for one month with repeat recordings and completed a Neurological Symptom Inventory (NSI) with each recording.

HeadPulse waveform analysis involved frequency transformation, then evaluation using a prespecified algorithm. Calculated waveform patterns of concussed athletes were compared to control subjects using Z scores. HeadPulse concussion was characterized as Z scores>2 (2 standard deviations from control HeadPulse Z scores). Sensitivity, time of onset, and duration of HeadPulse concussion were investigated.

100 subjects participated with 58 control and 42 concussed subjects and 44 total concussions. HeadPulse biometric parameters became abnormal in 82% of injured subjects (50% identified by day 2, 90% by day 14). HeadPulse biometrics remained elevated 12 days longer than reported symptoms on average. HeadPulse abnormality exacerbations symptoms occurred in those who returned to play or engaged in unsupervised physical activity.

Results show that serial HeadPulse measurement has diagnostic value for concussion and recovery and may identify potential subconcussive changes during return-to-play within one month. This objective metric may support clinical decisions.

The study was designed to align with local procedures including concussion determination and RTP decisions. Research coordinators attended games and were alerted to players with concussion. An attempt was made to record from concussed athletes early. All but one initial recording were conducted within one hour of concussion. Coordinators traveled to subjects' homes every 1-3 days for a month following injury to obtain additional recordings. The A2 cohort was encouraged to wear a wristband accelerometer after injury to document physical activity (FitBit, Inspire 2. Alphabet, San Francisco, USA).

Recordings were obtained from a battery-powered noninvasive headset device as described above, comprised of highly sensitive accelerometers attached to a headset. The headset was placed on the subject's head in the coronal plane (FIGS. 1 and 2). The HeadPulse is comprised of forces on the head produced by cardiac contraction in the 10-15 mg (g, gravitational force unit, 9.8 m/s²) range. In A1, bilateral triaxial accelerometers were placed anterior to the ear over the temporal bone. In A2, a commercial device (MindRhythm, Inc, Cupertino, CA) with a unilateral triaxial accelerometer was used.

Devices transduced the electrocardiogram (ECG) using a standard 3-lead system to provide heartrate information. Devices connected to an iPhone using Bluetooth and custom software. Recordings lasted 180 seconds in A1 and 90 seconds in A2 and were obtained in the seated position while the subject was asked to hold the head still and not to speak or chew. The A2 headset (MindSafe, MindRhythm, Inc, Cupertino, CA) was designed to provide feedback to participants when motion was detected. Subjects completed an adapted Neurobehavioral Symptom Inventory (NSI) 23,24 on the iPhone with each recording.

Data files were submitted to custom software (MATLAB, MathWorks, Natick, MA) to analyze frequency data as previously described. Time domain accelerometry signals from the right-sided accelerometer for 45 heart beats were converted to frequency domain using Fourier transformation.

The average heartrate during recordings was derived from R-wave analysis of the ECG and provided the fundamental and harmonics of the heartrate. The frequency transform was sampled at the fundamental and successive harmonics 2-9. Concussion biometric R₁was calculated as the ratio of the mean of the 5^thand 6^thharmonics to the max of harmonics 1-3 (R1), and concussion biometric, R₂, was calculated as the mean of the 7th and 8th harmonic divided by the max of harmonics 1-3 R₁and R₂values for controls were calculated.

Concussion biometric, R₁and R₂, values were analyzed as Z scores above or below mean control values. Higher concussion biometric, R₁and R₂, Z scores represent higher frequency shift of HeadPulse following TBI. The analysis defined any recording exceeding two standard deviations (SD) of controls to be significant, and the earliest value to exceed this threshold was defined as concussion biometric onset time of abnormality.

Concussion biometric R₁and R₂, values and corresponding NSI scores were plotted over time. A Kaplan-Meier curve representing freedom from biometric abnormality was calculated to illustrate diagnostic sensitivity compared to controls and to illustrate latency from concussive event to first abnormal device recording. To document recovery from concussion, the duration from biometric concussion onset to a zero NSI score was tabulated; if the subject did not return to a zero NSI score during the recording period the subject was considered not clinically recovered. For subjects who did not have a threshold biometric onset time, the biometric duration was defined as zero. For subjects with threshold biometric onset, the recovery period was defined as duration from clinical concussion to time that all ratios fell under one SD from the mean. The choice of one SD was based on the observation that concussion biometric, R₁and R₂, fell over time in a characteristic fashion, allowing for objective quantization of biometric parameters.

The study found characteristic HeadPulse changes after acute injury and marked asymptomatic HeadPulse abnormalities in those with RTP. Of 762 athlete volunteers, a total of 100 subjects had HeadPulse measurements including control (n=58) and concussed (n=43) subjects; one control subject was later concussed (so counted twice) and one concussed subject in A2 sustained two concussions separated by more than 1 month for a total of 44 concussions.

Subjects had similar years of education, female athletes had fewer self-reported prior concussions than males, and male athletes had historically played more games than female athletes (proxy for concussion exposure) but there were no significant differences of games played between control and concussion subjects within genders. Among concussed subjects, 4 (10%) had loss of consciousness (LOC), 16 (39%) had alteration in consciousness, and 7 (17%) had transient post-traumatic amnesia (PTA) at time of injury. Concussion was diagnosed by a physician in 7 (17%), team staff member in 25 (61%), the subjects themselves in 8 (20%), and 1 was unknown (2.4%). The diagnosis of concussion followed AFL guidelines. No subject sustained a concussion within the month following injury although many returned to play within this timeframe.

In A1, 40 of 184 (22%) of recordings were motion degraded. In A2, an improved device provided feedback regarding excessive body motion (MindSafe device, provided by MindRhythm, Inc, Cupertino, CA) resulting in fewer rejected recordings (24/276=8.7%).

FIG. 3 shows a typical HeadPulse recording, which is raw data in the time domain for each of three axes on which accelerometer readings are taken. The first three panels are HeadPulse forces, in mg (milli-g in the vertical, anterior-posterior and lateral axes, respectively (i.e. the X, Y and Z axes, respectively). The bottom trace in FIG. 3 is the ECG.

FIG. 4 is the Fourier-transformed data from a time domain sample similar to that in FIG. 3, but for a subject immediately following concussion. Harmonics of the heart rate are shown as vertical dotted lines and denoted by number. The height of the frequency transform at each harmonic is used to calculate the concussion biometric, R