With a few years of experience under my belt now, I’m a little better equipped to think about the question I’ve already spent countless hours trying to answer: What does training load monitoring and injury prevention mean in the context of professional team sports?
I still don’t have a definitive answer, and I realized that the more I knew, the less certain I was. However, there are some interesting avenues for using some of the tools and techniques that are popular today—and there are plenty of reasons to question others.
Why do we need such a flawed system for monitoring training load?
Early in my career, I caught the diagnostic virus. Busy, hurrying through the corridors of Allianz Stadium, a black briefcase filled with GPS in one hand and a computer in the other, I could already see myself making the front pages of popular scientific journals: “Pierre Austruy, the Sherlock Holmes of Physical Preparation.” In my head, as a young graduate not yet dented by the reality on the ground, I imagined an ideal and perfect monitoring program. It was all about clues to find, details to highlight, probabilities to calculate, and—voila!—the game would be played and injuries and poor performance would be avoided.Tired of accumulating data that, in the end, is often not that insightful, I dream of a monitoring routine that is both innovative and refined. Click To Tweet
My ambition echoed the common practice in most sports teams (and practices still widespread), consisting of prescribing players a wide range of weekly tests in order “to not miss anything.” Today, I feel like I have been cured. Tired of accumulating data that, in the end, is often not that insightful, I dream of a monitoring routine that is both innovative and refined.
Functional Movements Test vs. Individualized “Risk Zone” Monitoring
This type of test is a bit like an IQ test: having a great score makes you superior in theory, but rarely extraordinary in practice. (And, by the way, are there any movements that are not, in one way or another, functional?) Very high IQs often make their commute to work standing in a crowded bus, sweating in their tight three-piece suits. They are very good employees but remain anonymous. The geniuses we admire, on the other hand, may not have the most flattering score, but their remarkable creativity or special intelligence reflects their extraordinary mental abilities.
When we consider very high-level athletes, we are dealing with “monsters”: individuals with dimensions and physical and mental potential outside the norm. And it is possible that certain imbalances between one limb and another, certain limitations in range of motion, or even certain unconventional movement strategies are what ultimately make the player remarkable. Why aim for balance and normality as the criterion for satisfactory monitoring, when the desired performance requires the exceptional and the extreme?
I don’t, however, think musculoskeletal system testing should be eliminated from a monitoring program altogether. Having tangible confirmation that, on a biomechanical level, an athlete is in a compromising situation—and therefore, being able to intervene with the individual (specific exercises, massages, etc.) or with training (load, volume, exercise selection, etc.)—is an essential part of injury prevention. I would, though, advocate a minimalist approach, far from the classic spider-web strategy where quantity trumps quality, by adopting as a mantra the sentence, “Better is less, but better!” Couldn’t weekly or daily prophylaxis only focus on previously injured areas?
I have, for example, 100 data points on ankle flexibility for players who have never missed a single practice session due to ankle problems. Does this mean that I am very good at my job and that the monitoring program in place prevents ankle injuries? Or does it mean that, for these players, ankle flexibility is not a good indicator of their readiness?
What is the first predictor of an injury? A previous injury and usually in the same place. As soon as we suffer structural damage, we are forever compromised. Sure, things get better, and we can “fix” a problem for a while (years in some cases), but we never go back to the state we were in before the injury. Perhaps a previous injury should dictate which monitoring tests are truly representative of individual-level readiness and health?
Additionally, each player faces different biomechanical challenges, whether due to anatomical constraints, compromised tissues, or poor movement habits. Just as the annual blood test your doctor prescribes doesn’t include testing for tropical diseases if you haven’t traveled abroad, why not just limit monitoring tests to specific potential problems instead of carrying out a more general screening?
Checking a bit of everything all the time “just in case” so as to “not miss” a threat of injury, in my opinion, only serves to make us feel as we are very professional and conscientious. We proudly distribute our endless spreadsheets and multiple reports, but their actual usefulness is limited. As the season progresses, players lose interest in the procedure, we have more and more data and less and less time to analyze it properly before making decisions, and injuries still aren’t eradicated.Perhaps focusing monitoring routines on the individual profile based on injury history and biomechanics would give us better play participation and more profitable data. Click To Tweet
Perhaps focusing monitoring routines on the individual profile based on injury history and biomechanics would give us better player participation and more profitable data. A player with a history of hamstring injuries, but who has never experienced upper body issues, may skip a shoulder mobility test to focus on a more specific exercise such as a Nordic hamstring curl before continuing on with their day. Most likely, if they were to sustain an injury in training, previously compromised tissue would be among the determining factors.
The Need for More Internal Data
Since a well-designed monitoring program includes much more than simple movement tests, let’s continue our rethinking of current practices with the same desire for renewal. Physiology may well be the mother of sports sciences, but it is poorly represented when it comes to monitoring (which is regrettable). Many practitioners are content to estimate the physiological state of an athlete through markers that are external to them, such as GPS data or RPE and wellness questionnaires (which, for a long time, was my case as well). And when taking the time to study the internal load, these practitioners are satisfied with the minimum.
Heart rate is a great indicator, revealing valuable information through a multitude of clues. Young “sport scientists” new to the world of professional sport know this well. And that’s why they eagerly pounced on studying cardiac data early in their careers. Very often, however, faced with the reluctance of players (a heart rate monitor is uncomfortable, especially in contact sports) and the poor reliability of the data (poorly placed heart rate monitor, loss of signal, etc.), the interest in the cardiac response to exertion tapers off and remains under-interpreted.
Resting heart rate then seems the easy solution, but because it’s extremely subject to fluctuations, it does not provide much critical information. Nowadays, more and more teams use a standard running test (type “box run”), always carried out on the same day in a week of preparation (in general, 48 hours post-match), during which the players are equipped with cardiac monitors. Different data, such as average or peak heart rate, is collected and compared to previous datasets.
An increase in average heart rate during this test compared to the previous week, for example, then indicates less recovery. While this procedure is a step forward, it is subject to variability, which makes the conclusions difficult to interpret.
- If the test is performed on-field, the properties of the surface at the time of the survey partly explain the fluctuations in the results, with heart rate being related to energy demand—the latter itself being related to mechanical demand. The humidity level, the height of the grass, and other environmental concerns have a direct impact on heart rate.
- To obtain quality data, it is necessary that the climatic conditions are standardized. The temperature outside, the dryness or humidity of the air, and even the wind partly determine the heart’s response.
- The individual training load that precedes the test should be similar from week to week. As the goal of the procedure is to produce a physiological index capable of attesting to the player’s state of fitness, it is important to consider the context. For example, if before the previous test a player participated in 30 minutes of the match, and this time they are tested after a match in which they played 60 minutes, should we expect the same results?
Producing a better strategy to recover valid physiological data requires eliminating these fluctuations as much as possible. A stationary bike, Watt bike, or rower has the advantage of having stable physical characteristics, and an indoor test eliminates weather considerations. Instead of a collective test at a standard time, it may be better to perform an individual test when the training loads over the previous seven days are comparable: a monthly test is sufficient. It is also necessary to ensure that no psychological factor too strongly impacts the cardiac response. Conducting this test immediately after a meeting with the coaches, or a review of a match, is not a good idea.
On the other end, checking wellness questionnaires before deciding to go on with the testing procedure is a necessity. If we were to see in them unexpected levels of stress or soreness reported by a player, we would be better off reprogramming the test to a later date for that individual.
Finally, while observing average heart rate during exercise is relevant, looking at recovery heart rate is probably more informative. Recovering heart rate after exercise involves a coordinated interaction of parasympathetic reactivation and sympathetic withdrawal. Since the autonomic nervous system is linked to many other physiological systems, its ability to maintain homeostasis can provide information on the level of muscle inflammation, the amount of plasma creatine kinase due to multiple collisions, and the state of the central nervous system (parasympathetic or sympathetic inhibitions).
The two-minute recovery heart rate test is simple and effective. The athlete provides a maximum effort of two minutes on an ergometer (static bike, rower, etc.), at the end of which the heart rate is recorded (HR1). Two minutes after exercise, the heart rate is recorded again (HR2). Subtracting the HR2 from the HR1 determines a basis for improvement. For example, if HR1 is 150 beats per minute and HR2 is 95, then the recovery heart rate is 55.
Note that for athletes in high-intensity, intermittent sports, the ability to reduce heart rate between strenuous efforts is one of the most critical qualities required to perform, make a good decision under fatigue and pressure, and not get injured. This index therefore allows both monitoring the state of form and predicting the level of performance of the athlete.
It is not mandatory to turn to an active procedure that requires the player to make an effort to collect a picture of their physiological state. Some biological markers do very well and are detectable with a simple blood or saliva test. Admittedly, going down this road requires collaboration with an analysis laboratory and comes with significant costs, but many clubs fund without flinching the frequent trips their players take to and from the clinic for DEXA scans and other anthropometric tests as costly as they are futile. I can observe with the naked eye a worrying fat gain (unless I am its victim, in which case I prefer to doubt such visual evidence…). on the other hand, I am quite unable to estimate a level of blood heat shock proteins (HSPs).
HSPs participate in a cell protection system induced by the presence of reactive oxygen species (ROs), cytokines, or hyperthermia. HSPs increase stress tolerance and contribute to cell repair processes. In addition, HSPs are involved in the remodeling associated with exercise training, where they facilitate mitochondrial biogenesis, regulate apoptotic pathways, and induce improvements in insulin sensitivity. Muscle damage and stress resulting from exercise are considered two of the many stimuli that induce the synthesis of HSPs. Sustained high synthesis of these proteins may indicate a state of inadequate regeneration, even after several weeks of recovery from exhaustive exercise.
HSPs are a critical physiological marker for monitoring the athlete’s state of recovery and adaptation to training. As the understanding of the role of these proteins grows, it wouldn’t be surprising to see more and more teams take an interest in them. Along with HSPs, other reliable indicators of muscle damage can be:
- A measurement of the level of circulating creatine kinase (CK).
- An evaluation of the levels of interleukins (IL) and TNF alpha as representatives of inflammatory markers.
- A measurement of immunoglobulin A (igA), providing information on the state of the immune system.
A monthly or even quarterly profiling of the levels of HSPs, CK, IL-1, TNFa, and igA would make it possible to assess the readiness of a player in a much more relevant way than any rough estimate drawn from subjective questionnaires or derived from specious algorithms.
Glucose Continuous Monitoring
While nutrition is indeed considered one of the cornerstones of recovery and adaptation to training, it does not appear to be a priority in monitoring practices. When claiming to assess an athlete’s level of recovery, not studying what is considered a major factor in recovery is quite paradoxical. If the daily weigh-in gives relevant information to the farmer about his turkeys as Christmas or Thanksgiving approaches, such a monitoring tool is much less relevant for professional athletes.
Body weight fluctuates considerably and constantly, for obvious reasons (which I will be careful not to expose here in case you are finishing your Niçoise salad). Substantial and sudden weight loss is likely to indicate a state of overtraining or a health problem, and players generally have an optimal body weight in competition that is best kept under control. But when it comes to problematic weight loss, it usually occurs with other detectable symptoms—and when it comes to a healthy weight, players themselves pay special attention.New glucose monitoring technologies open up new perspectives on athlete nutrition management never before considered. Click To Tweet
New glucose monitoring technologies open up new perspectives on athlete nutrition management never before considered. The use of the continuous glucose monitoring (CGM) test makes it possible, for example, to study the evolution of the blood glucose level 24 hours a day, without resorting to an invasive method such as blood sampling. This is because the CGM machine is inserted into the back of the arm and measures glucose levels every 1-5 minutes, discreetly measuring the contents of the subcutaneous tissue.
Initially developed to support the treatment of diabetes, this technology is now used by a growing number of athletes, nutritionists, and sports scientists to assess individual responses to different nutritional strategies. As the blood glucose level is also affected by stress, this type of test makes it possible to evaluate the effects of a variety of constraints (emotional stress, jet lag, overtraining, lack of sleep, inflammation, etc.) on the metabolism and the physiological state of the athlete.
Observation of the behavior of a physiological variable for 24 hours in the real life of the athlete—and not only in a controlled environment—is a considerable asset in understanding the possible reasons for poor performance and injuries. A fasting blood glucose level of 6.1 mmol/L is considered to be normal, and within two hours of ingestion of 75 grams of glucose as part of an oral glucose tolerance test (OGTT), the glucose blood level should not exceed 7.8 mol/L. Although sports are often cited as excellent for glycemic control, several studies1,2,3 report more mixed conclusions. Four in 10 athletes have blood glucose levels above 6 mmol/L 70% of the time under control, according to a recent study4, and some researchers have advanced the theory that genetic makeup associated with exceptional performance in power or endurance in elite athletes could also reflect their metabolic characteristics.
Elite power athletes appear to be more resistant to insulin than elite endurance athletes. The training loads act as a stress factor influencing insulin sensitivity—in particular, the repetition of intense efforts that generate a significant release of catecholamines and result in post-exercise hyperglycemia and hyperinsulinemia, as well as nutritional strategies based on the high consumption of carbohydrates. The use of a glucose test is important to explore the different nutritional strategies to optimize the performance of the athlete and to prevent a decrease in insulin sensitivity, which would have significant consequences on their recovery, health, and energy.
But the applications in high-performance sport related to these new technologies do not stop there:
- What strategy should coaches and athletes adopt when traveling to limit fatigue and stress?
- Does the training schedule negatively affect players “sleep quality”?
- Do we train too hard the day before the game?
All these questions can be examined through continuous glucose monitoring. Stress, through the production of glucocorticoids as well as inflammatory markers, acts on the levels of glucose and insulin, which makes the observation of glucose a remarkable physiological test for the sport scientist.
Designing Better Subjective Questionnaires
Finally, there can be no conversation about monitoring without discussing the use of subjective questionnaires. Ah!!! Running after athletes who did not complete their questionnaires on time is the flagship activity of today’s sport scientists. The idea that, coupled with objective measurements, subjective cues make it possible to pinpoint revealing discrepancies (for example, the difference between an athlete’s physical output and their feeling of fatigue) is attractive, and not devoid of practical interest.
However, are the questionnaires used ideally composed?
Let’s stop asking questions we can’t answer ourselves! For the sake of data accumulation, I felt like I was guilty of nonsense. Wellness questionnaires, RPE, play rating scales, emotional profiling… I asked a lot of questions. I thought these numbers would inform decisions about training load, player tracking, and individual management. But what kind of valuable information can you get from a crappy question (forgive my vocabulary!)? As Yogi Berra said, “If you ask me something that I don’t know, I won’t answer.”
Seriously, a lot of us have a question on our wellness questionnaire that sounds like this: “How would you rate your energy level today?” Have you tried to answer this question yourself? Personally, I don’t know if my energy this morning was a 6 or a 7 out of 10. Maybe it was even a 5. And are you talking about my energy before the first coffee or after the fourth? Before making the trip home to the gym while listening to an exciting podcast or after collecting a tax notice in my mailbox? How do we define energy itself in relation to this question? Is there a clearly defined equivalence between each number on the scale and a given series of events? What distinguishes a 7 from a 6?
Another pearl: “Rate the quality of your sleep on a scale of 1 to 10.” What sort of question is that?! We know that it is very difficult to estimate your own sleep quality. It is not something we can clearly state. Collecting RPEs after each session does not provide a complete, individual, and subtle understanding of each player’s response to training.
To me, it has become obvious that most players pick a number at the start of the season and just stick to it as their default value. And when they come up with a different number, the surprise effect isn’t there—obviously, the session was designed to be easier than usual or much more difficult. The default value strategy makes sense. Can you distinguish a “moderate” session from a “somewhat difficult” session? I’m not sure I’ve ever had a “somewhat difficult” session myself…in the Borg scale, 7 and 8 have the same definition. What, then, justifies choosing one over the other?
Players already face enough daily demands that we should not prompt them to think about this type of senseless choice. I debated which statistical methods to use and spent time on “actionable strategies,” gently and indulgently prepared Excel tables and made pretty, colorful reports filled with data that had no real value. Asking ill-conceived questions can quickly make you mistake a noise for a song.If the primary function of a wellness questionnaire is to generate a discussion with the athlete, what prevents us from getting straight to the point with our questions? Click To Tweet
If the primary function of this type of questionnaire is to generate a discussion with the athlete—and this exchange should shed light on the measures to be taken—what is preventing us from getting straight to the point? A coach can verbally ask his players black and white questions, the kind that demand a clear answer: Is there any reason for me to worry about your readiness to train today? These are the types of questions that identify players who are in a physical or psychological state that does not meet expectations and that is not directly the result of training loads or a clearly identifiable reason.
As for the RPE, it is difficult to question it, as it is central today in the management of training loads. This approach, however, can still be improved. The purpose of the RPE is to assess an athlete’s fatigue state by comparing the session performed to the athlete’s feelings. Then, by multiplying the number chosen by the duration of the session, an estimate of the overall workload (at least from a subjective point of view) is proposed. But is the question “Rate your perceived feeling of exertion on a scale of 1 to 10” the only or the best way to evaluate fatigue? I don’t think so.
Besides the limitations inherent in the Borg scale discussed above, this question does not really give athletes the opportunity to communicate on the reality of the effort they put in. Maybe, objectively, I spent 10 minutes above 80% of my maximum heart rate in a 45-minute session containing three 30-second blocks of runs with 30-second recovery at 100% of my maximum aerobic speed (VMA). Let’s call this workout W1.
Consider workout W2, where I do the same work, at the same intensity, and hit the same objective markers. The only difference is that the session is performed on a rower. In this case, very few athletes will score these two sessions identically, where except for the exercise mode, all other things are equal. Those who like running will find W1 easier; others will prefer session W2. From this point of view, the RPE therefore leaves a difference in taste to dictate, at least partially, the estimate of a training load.
In addition, the environment, climate, time of session, or cumulative effect of previous sessions weigh on the choice of RPE. For two similar sessions, the athlete may give two different scores—not because they feel more or less intense physical or psychological fatigue, but simply because their interpretation of elements beyond their control varies. In addition, the view that the physical trainer has of the physical demands of a session does not necessarily match the feelings of the player who goes through it. Sometimes, this discrepancy reflects something in the realm of fitness, but quite often it highlights a difference in appreciation. The RPE also contrasts what the strength and conditioning coach expects from a session with what the player feels. Two different individuals, two different understandings—a clue very open to interpretation indeed.A more complementary RPE strategy would be to compare the expectation that the player places on themselves before the session with their feelings at the end of it. Click To Tweet
A more complementary strategy would be to compare the expectation that the player places on themselves before the session with their feelings at the end of it. Questions could include:
- “How satisfied were you with your performance during the session?”
- “Rate your performance during the session on a simple rating scale from 1 to 10.” (Or even better, from 1 to 5 to avoid the always convenient choice of average.)
Here, the players face themselves. By entering the field and knowing the environmental factors, the type of session, and the expectations of the strength and conditioning and technical staffs, they form an idea of what they are capable of delivering in terms of performance. While making their way out of the training field or court, they are in a position to take a step back from their production, estimate the differential between before and after and between expected and actual, and verbalize their scores accordingly.
Why put the feeling of the fatigue aspect into perspective in order to refocus on the appreciation of the performance aspect? Well, personally, if I have a very tired team that wins every weekend, that’s fine with me. I dare say that even if you are a sensitive soul and have a remarkable capacity for empathy, you would be comfortable in this scenario as well. On the other hand, if my team is a cohort of guys in great shape that do not win a match, I would be very worried.
Complementary to the observation of the relationship between intensity of a workout and subjective fatigue, studying the relationship between intensity of a workout and athletic performance would provide useful and relevant information for directing a physical preparation program. Understanding the effect of training load not only on a subjective fatigue level but on a player’s perception of their own performance opens up new perspectives, and when asking the athlete about their performance, two important indications for a monitoring program appear.
First, a simple number reflecting the performance level of each athlete for each session is created. This score can be analyzed in relation to other training load indices. Unlike the RPE, it does not refer to the concept of “load” itself and breaks away from the purely physical preparation aspect, which makes the comparison with other markers, such as those obtained with GPS, more balanced.
Second, this approach allows, in a mediated way, the detection of certain traits associated with abnormal psychological or physical fatigue, thus complementing the RPE data. Ask a genuinely tired player if they found a session tiring, and the chances are high they’ll say “no.” True competitors, often proud and sensitive, sometimes prefer to curl up when they are not feeling their best. But if this fatigue is indeed present, it probably leads to a decrease in performance, and the athlete knows this. To admit, on top of that, that they are suffering from the difficulty of training is akin to an admission of weakness and sounds like a message directly addressed to the coach: “Don’t select me for Saturday’s game!”
Only the star players and those indisputable spot-holders can afford to be tired. The rest—those who are fighting for their place in the selection—will likely line up with the ratings given by their competition instead. On the other hand, this same competitive spirit makes the players very critical of their performances. Everyone hates being the weak link, but they know that overestimating their performance is frowned upon in the coaching office. Caught between disappointment at a poor score and fear of an overrated score, the athlete is forced to be honest about their performance.
So when by analyzing the data collected we see that for the same typical session a player rates their performance differently, we can make assumptions. If the score is lower, as the player certainly was not intentionally less precise in their tasks, this loss is probably the result of physical or mental discomfort. If there is no clearly identifiable injury or psychological stress, then we are probably finally in the presence of this harmful fatigue that we dread.
To end with this review of a classic monitoring program, let us insist on the fact that the absence of objective measurement of the cognitive state of the athlete is justifiable only because of the absence of technology. This does not prevent us from dreaming a little, and with one foot in the future, from considering a few options.
The Future of Monitoring?
When looking at the next generation of monitoring, I begin with blink rate as a measure of cognitive fatigue. Spontaneous eye blinking occurs much more often than is necessary to maintain the tear film over the eyes. Various factors, such as cognitive demand, level of concentration in performing a task, and fatigue, influence the spontaneous blink rate. When it comes to competition day, decision-making, reaction time, and vision are as important (possibly much more important) than muscle glycogen content or strength level.When we look at the future of athlete monitoring, blink rate as a measure of cognitive fatigue has real possibilities. Click To Tweet
We are good at monitoring physical readiness and making sure we don’t compromise match day performance by creating too much fatigue beforehand in training. Physical fatigue is relatively easy to detect with field tests or in the weight room. In contrast, we tend to forget about cognitive load monitoring. Did we do too much video analysis this week? Is this the best time to add individual technique sessions? What about the weight of meetings?
So far, we’ve relied heavily on subjective measures such as wellness scores to try to identify this cognitive fatigue. Measuring eye blink rate can be done like any other monitoring procedure, through a simple protocol where the athlete performs a cognitive task and eye tracking technology is used. As research on this topic develops, we can imagine that an easy-to-implement system will eventually become available in the market.
If language is preferred over the study of the gaze, a system of discourse analysis could be revealing. A recent study questioned the possibility of detecting fatigue and drowsiness in airline pilots by observing their communication with the control tower.5 Considering that fatigue and drowsiness are the main factors in human-caused air crashes, the researchers undertook a retrospective analysis of pilot communication just before an accident.
The pilots’ speech was analyzed 35 hours before the fatal flight in normal conditions (control condition) and compared to a recording approximately one hour before the fatal flight and during the accident (drowsiness condition). The analysis focused on the temporal organization of speech: hesitation, silent pauses, prolongation of the last syllables, and speed of articulation. The results show that the speech on the day of the accident was characterized by slower speech and significantly slower articulation speed than on the previous days.
Other research, particularly in the military field6, has also successfully explored the relationship between speech characteristics and fatigue, demonstrating that inter-individual variability in speech, choice of vocabulary, and rhythm of syllables is an effective means of measuring cognitive fatigue. What comes up over and over in interviews with strength and conditioning coaches is that nothing beats direct contact with the player and the exchange of a few words to determine their state of form.
Perhaps. But then, if this is so productive that the exchange can also be objectified, isn’t that ideal? Create three standard questions, ask each athlete on a standardized day of the training week, record the answers, then pass these samples to a revealing system for analyzing speech and syllable flow rate…. Here, in part, perhaps lies the future of monitoring in terms of cognitive fatigue.
Since you’re here…
…we have a small favor to ask. More people are reading SimpliFaster than ever, and each week we bring you compelling content from coaches, sport scientists, and physiotherapists who are devoted to building better athletes. Please take a moment to share the articles on social media, engage the authors with questions and comments below, and link to articles when appropriate if you have a blog or participate on forums of related topics. — SF
1. Lime-Ma F, Cotter J, and Schick E. “The Effect of Acute Hyperglycemia on Muscular Strength, Power and Endurance.” International Journal of Exercise Science. 2017;10:390-396.
2. Steffes GD, Megura AE, Adams J, et al. “Prevalence of metabolic syndrome risk factors in high school and NCAA division I football players.” Journal of Strength and Conditioning Research. 2013;27(7):1749-1757.
3. Buell JL, Calland D, Hanks F, et al. “Presence of metabolic syndrome in football linemen.” Journal of Athletic Training. 2008;43(6):608-816.
4. Thomas F, Pretty CG, Desaive T, and Chase JG. “Blood Glucose Levels of Subelite Athletes During 6 Days of Free Living.” Journal of Diabetes Science and Technology. 2016;10(6): 1335–1343.
5. Vasconcelos C, Vieira M, Kecklund G, and Yehia HC. “Speech Analysis for Fatigue and Sleepiness Detection of a Pilot.” Aerospace Medicine and Human Performance. 2019;90:415-418. 10.3357/AMHP.5134.2019.
6. Greeley HP, Friets E, Wilson JP, Raghavan S, Picone J, and Berg J. “Detecting Fatigue From Voice Using Speech Recognition.” International Symposium on Signal Processing and Information Technology. 2006.