id large_stringlengths 7 16 | title large_stringlengths 0 1.14k | author large_stringlengths 0 51.6k | publication_date large_stringclasses 121
values | doi large_stringlengths 0 91 | language large_stringclasses 42
values | license large_stringclasses 7
values | terms large_stringclasses 1
value | discipline large_stringclasses 689
values | source large_stringlengths 0 151 | page int64 1 3.67k | text large_stringlengths 1 2.71M | word_count uint32 1 120k | supra large_stringclasses 7
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 21 | resistance due to higher airflow. The optimal $f_R$ is found between these two extremes [57]. However, in COPD patients, adaptations of $V_T$ may be restricted because of lungs hyperinflation, loss of elasticity or even the reduction of muscles capacities. Similarly, $f_R$, and specially the duration of expiration, is ... | 493 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 22 | Another factor that influences the ventilation is the compliance of the lungs. The lungs tissue has elastic properties that enables them to distend and to recoil, following inspiration and expiration respectively. The compliance is the capacity of the lungs to stretch and is inversely proportional to the elastance (Equ... | 443 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 23 | Static hyperinflation appears when the decreased lung capacity to recoil leads tidal res-
piration to occur at larger lung volumes. Dynamic hyperinflation is related to a temporary
increase in FRC. It happens when the demand in $\dot{V}_E$ is increased, as during exercise. A
raise in $f_R$ leads to shortened expiratory... | 509 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 24 | ### 2.6.1 Spirometry
The spirometric examination is used for measuring the ventilatory function, by quantifying the respiratory volumes and flows at specific conditions. It is the most important examination for diagnosing the COPD because it allows assessing the airway obstructions.
For the test, the patient breaths ... | 374 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 25 | This difference is limited in the flow-volume loop of COPD patients. The flow limitation that characterizes the disease can also be observed by a reduction of the peak expiratory flow. If the patient has an hyperinflation, it will be perceived by a drift in the volume axis, the loop remaining at higher lung volumes (no... | 503 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 26 | the activity of the respiratory muscles can be adapted without exceeding their maximum capacity.
In the other hand, a patient with COPD, even at his baseline, has a precarious load-capacity balance. Its basal level of load is already high because of the resistance in the airways (inflammation of the bronchi, overprodu... | 213 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 27 | # Chapter 3
## Telemedecine
Telehealth and telemedicine are two terms used to encompass different types of interventions that use information and communication technologies to provide or support health-care. Their goals include the reduction of the number or duration of hospitalisations, improvement of quality of lif... | 381 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 28 | of the patient's physiological variables.
Sporadic spirometry examination, for instance, can now be replaced by more frequent measures thanks to modern spirometer models. With these devices, patients can take daily or weekly measures from home, possibly allowing for a richer comprehension of the patient's disease evol... | 550 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 29 | measures were outside chosen thresholds. If considered appropriate, the pneumologist was
then contacted to treat the event, by prescribing a medical treatment by phone, visiting
the patient at home or advising the patient to go to the hospital, according to the gravity
of each case. This study has shown that it is poss... | 505 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 30 | The Anthonisen Criteria is commonly used. It is also called symptom-based criteria, as it defines exacerbation as an increase in three or more symptoms, including at least one major symptom (dyspnea, sputum amount and sputum purulence), during two consecutive days [9]. Variations of this definitions may be used. For ex... | 571 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 31 | symptoms. However, most other authors use fixed duration.
To make opposition to the exacerbation phases, studies also define control or baseline
periods. These are the periods when no acute event was recorded and with sufficient
distance from previous and next exacerbations.
This sufficient distance before and after ... | 564 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 32 | or weekly, some studies reported inconsistent measurement frequency [68, 88, 43]. Also, measuring devices can sometimes be used by other people, such as a patient's relative, creating outliers in the dataset [43].
Less dependent on manual measures, some monitoring devices coupled with non-pharmacological treatments re... | 550 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 33 | <table><caption>Table 3.1 – Summary of publications on exacerbation detection.</caption><thead><tr><th>Features</th><th>Performance</th><th>Advantages</th><th>Drawbacks</th><th>Reference</th></tr></thead><tbody><tr><td>Breathing rate</td><td>Sensitivity: 63%<br>Specificity: 85%</td><td>Online learning of “normality” li... | 135 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 34 | <table><caption>Table 3.1 – Summary of publications on exacerbation detection.</caption><thead><tr><th>Features</th><th>Performance</th><th>Advantages</th><th>Drawbacks</th><th>Reference</th></tr></thead><tbody><tr><td>Wavelets from respiratory sounds</td><td>Sensitivity: 78.1%<br>Specificity: 95.9%</td><td>Good perfor... | 143 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 35 | <table><caption>Table 3.1 – Summary of publications on exacerbation detection.</caption><thead><tr><th>Features</th><th>Performance</th><th>Advantages</th><th>Drawbacks</th><th>Reference</th></tr></thead><tbody><tr><td>FEV1, SpO2 and weight</td><td>Sensitivity: 61.1%<br>Specificity: 80.4%</td><td>Chosen method (CART) c... | 146 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 36 | <table><caption>Table 3.1 – Summary of publications on exacerbation detection.</caption><thead><tr><th>Features</th><th>Performance</th><th>Advantages</th><th>Drawbacks</th><th>Reference</th></tr></thead><tbody><tr><td>Variability of inspiratory reactance from forced oscillation technique</td><td>AUC: 0.72<br>Sensibili... | 50 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 37 | ### 3.3.4 Statistical methods
A variety of statistical methods were applied to the exacerbation prediction problem. Most of them are supervised methods, including logistic regression [49, 88, 44], linear mixed-effects models [112], Classification And Regression Tree (CART) [60], random forest [31], among others.
All ... | 347 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 38 | # Chapter 4
## Artificial intelligence
Remote monitoring devices allow for more detailed monitoring of COPD patients. However, these physiological measurements can quickly amass into large amounts of data, because of both the frequency of measurements and the number of patients concerned. Manual treatment of these da... | 418 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 39 | ### 4.1.1 Fourier transform
According to the theory of Fourier series, any periodic signal can be decomposed in sines and cosines at different frequencies and magnitudes (Equation 4.1.1).
$$g(t) = \frac{1}{2}a_0 + \sum_{n=1}^{\infty} (a_n \sin(2\pi nt) + b_n \cos(2\pi nt)) \quad (4.1.1)$$
The coefficients $a_n$ and ... | 381 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 40 | ## 4.1.2 ARIMA
Autoregressive Integrated Moving Average (ARIMA) is a modelisation approach for time series data relying on the assumption that the signal is autoregressive, that is the value at each time point can be written as resulting from a linear model according to the preceding points and their errors [17].
An ... | 487 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 41 | ## 4.2 Classification in times series
With the extraction of the right features, the monitoring of longer characteristic time can evidence physiological changes in different scales of time, like days, weeks and months. For instance, a hidden Markov model can be used to identify different hidden states with a non super... | 282 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 42 | ## Estimate the model parameters
A hidden Markov model is described by the parameter $\Theta = (A, B, \Pi)$. $\Theta$ can be estimated from the observed data using the Baum-Welch algorithm, which is a special case of the expectation-maximization (EM) algorithm [2].
The objective of the algorithm is to find the local ... | 370 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 43 | $$
\begin{align*}
\pi_i^* &= \frac{\sum_{r=1}^{R} \gamma_0(1)}{R} \\
a_{ij}^* &= \frac{\sum_{r=1}^{R} \sum_{t=1}^{T-1} \xi_{r,t}(i, j)}{\sum_{r=1}^{R} \sum_{t=1}^{T-1} \gamma_{r,t}(i)} \\
b_i^*(v_k) &= \frac{\sum_{r=1}^{R} \sum_{t=1}^{T} 1_{o_{r,t}=v_k} \gamma_{r,t}(i)}{\sum_{r=1}^{R} \sum_{t=1}^{T} \gamma_{r,t}(i)}, \... | 371 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 44 | In some cases, it is necessary to take into account dependency between observations. For example, when repeated measures are originated from each subject. In this case, the random effect related to individuality needs to be combined to the fixed effects of interest by applying mixed effect models.
### 4.3.1 SuperLearn... | 540 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 45 | $$E(Y) = X\beta + Zu \qquad (4.3.2)$$
Typically, the choice of the link function depends on the type of the response data. For example, a log link function is proposed when the response variable is a count of occurrences and a logit link function when the response variable is binary.
## 4.4 Novelty detection
As prev... | 482 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 46 | <figure><img src="image_9.png" /><figcaption>Figure 4.3 – Example of Mahalanobis distances from reference points in the breathing rate-amplitude plan</figcaption></figure>
distribution of mass $\mu$ that describes the pile of sand. We wish to move each grain of sand so the final distribution is $\nu$. The ground cost ... | 325 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 47 | distributions over the same set of events and are commonly used as loss functions in classification learning algorithms.
The cross-entropy $H(P, Q)$ of the estimated probability distribution $Q$ with respect to the reference (true) probability distribution $P$ is given by the Equation 4.4.4, by comparing the probabili... | 427 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 48 | and used to forecast the following point, which is then compared to the observed point.
ARIMA models are also often used with this approach.
In the Hidden Markov Model-based approaches, a HMM is trained with the training
time series. Considering that HMM really captures the “normal” process, the probability of
observi... | 307 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 49 | # Part II
## Technology development | 6 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 50 | # Chapter 1
## TeleOx® monitoring device
Worldwide, COPD is a major cause of mortality and loss of quality of life. The disease evolution is traditionally assessed with respect to changes in the respiratory mechanics and in the capacity to execute daily activities. Spirometry and 6MWT are two examinations that allow ... | 235 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 51 | ## 1.1 Commercial TeleOx®
TeleOx® is a medical device class IIa CE marked. It was developed to be a portable and efficient device for remote monitoring patients under LTOT with flow rates between 0.5 and 5 liters per minute. With a total weight of 35g (including the battery), it has a power autonomy of one year. It wa... | 559 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 52 | version is hereafter called raw data mode.
With this configuration, the device performs continuous acquisition (no sleep phases). Parameters computation follows the same algorithm as previously described. Thus, they are estimated over windows of 45 or 12.8 seconds, according to the presence of the oxygen source.
Ther... | 163 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 53 | # Chapter 2
## Validation of breathing rate measurements
The interest of monitoring the breathing rate has been described in many different health-related situations [66]. For COPD patients in particular, changes in the breathing rate have been linked to the development of exacerbations in several studies. Thus, the ... | 345 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 54 | <figure><img src="image_11.png" /><figcaption>Figure 2.1 – Patient 1 - 7-hour recording from polygraph and TeleOx®</figcaption></figure>
<figure><img src="image_12.png" /><figcaption>Figure 2.2 – Patient 2 - 7-hour recording from polygraph and TeleOx®</figcaption></figure>
## 2.3 Filtering
Filtering is done accordin... | 202 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 55 | <figure><img src="image_13.png" /></figure>
Figure 2.3 – Representation of the filtering steps for a 45-second window. Signals from polygraph and from TeleOx® are presented in black and blue, respectively. Red dashed line indicates the estimated baseline.
<figure><img src="image_14.png" /></figure>
Figure 2.4 – 30 s... | 152 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 56 | subjects breathing. After the estimation of the Rohrer's equation coefficients, the study has pointed to a prevalence of turbulent flow, with $k_2 >> k_1$.
In another study, an alternative method consists of describing the pressure-flow relationship by the power equation $\Delta P = aQ^b$ [107]. With this equation, $b... | 365 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 57 | which may lead to some distortion from truth. Thus, different methods were tested to identify inspiration and expiration cycles.
First, the signal is interpolated using a piece-wise cubic interpolation at twice the original sampling rate (resulting in 20 Hz). This is done to ensure more precision for the time detectio... | 318 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 58 | ### 2.5.2 Signal crossing zero
Times corresponding to a zero crossing in the filtered signal are detected and identified as beginning of inspiration or beginning of expiration according to the slope around the given time. The beginning of an inspiration is characterized by a positive slope around the zero crossing. Th... | 303 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 59 | The Laplacian is a differential operator that is given by the sum of the second partial derivatives with respect to each independent variable. In the case of our 1-dimensional signal, it corresponds to the second derivative of the signal with respect to time. Since derivatives are very sensible to noise, we start by ru... | 452 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 60 | breathing in the period. Whenever the measured durations are too variable, TeleOx® chooses to identify this period as poor quality, not outputting a breathing measure.
### 2.6.1 Validation of breathing rate measurements
From the polygraphy protocol, the recordings from 14 patients resulted in 1099 valid TeleOx® measu... | 350 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 61 | device was Oxycon Mobile® (CareFusion®, San Diego, CA, USA), a portable device capable of spirometry and oxygen saturation measures.
Best agreements were achieved with the chest-band and the accelerometer with bias -1.60 and -2.18 and limits of agreement [-9.99, 6.80] and [-8.63, 4.27], respectively. Compared to these... | 91 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 62 | # Chapter 3
## Estimation of indicators of respiratory mechanics profiles
As we have been able to verify, the treated signal from TeleOx® pressure sensors resembles to the nasal pressure signal measured by the polygraph. Therefore, it contains information about the patient's breathing. At this point, this information... | 328 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 63 | In this trial, oxygen therapy was simulated by replacing the oxygen source by an air source. A raw data mode TeleOx® was placed in the air circuit, between the air source and the nasal cannula. The air flow rate was constant during each recording. Each subject followed instructions for a total of 30 minutes. They were ... | 323 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 64 | Subject A recording was done under 3 L/min air flow rate and subject B under 1 L/min
3.1.2 COPD patients
The second study protocol was named "Analysis of changes in respiratory parameters preceding exacerbation in COPD patients under oxygen therapy". This was a monocentric, observational and retrospective study. This... | 427 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 65 | and is addressed by one of the multidisciplinary healthcare professionals (doctor, nurse, physiotherapist, dietitian, etc). This session last about an hour, where patients are invited to listen to the professional, watch a video, discuss and ask questions.
Most patients were only monitored with TeleOx® in the commerci... | 304 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 66 | <figure><img src="image_25.png" /><figcaption>Figure 3.3 – Nasal pressure during different situations for subject 01. Red triangles indicate detected beginning of inspiration.</figcaption></figure>
<figure><img src="image_26.png" /><figcaption>Figure 3.4 – Nasal pressure during different situations for subject 02. Red... | 170 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 67 | <figure><img src="image_27.png" /><figcaption>Figure 3.5 – Nasal pressure during different situations for subject 03. Red triangles indicate detected beginning of inspiration.</figcaption></figure>
<figure><img src="image_28.png" /><figcaption>Figure 3.6 – Nasal pressure during different situations for subject 04. Red... | 170 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 68 | ### 3.2.4 Rest and exercise
Quiet breathing from different subjects vary in rate, magnitude and shape. Subject 02 presents longer and more ample respiratory cycles than the other subjects presented here. The normal respiration period presented for subject 04 starts with a deeper respiration, followed by quite constant... | 488 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 69 | Besides, the inspiration-to-expiration times ratio is only computed in periods where at least three inspirations and three expiration were detected. This threshold corresponds to three respiratory cycles in the 35.4 seconds valid period, or a breathing rate of 5 breaths/minute.
### 3.3.2 Flow-volume loop
Flow-volume ... | 370 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 70 | Thus, we analyse the changes in variability in our datasets by computing the coefficients of variation of the detected lengths of respiratory cycles, according to the pressure minima and maxima method.
### 3.3.5 ARIMA coefficients
ARIMA was used to model each valid period in healthy and COPD datasets. For future comp... | 445 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 71 | <figure><img src="image_30.png" /><figcaption>Figure 3.8 – Inspiration-to-expiration times ratio estimation over all valid recorded periods in healthy and COPD datasets</figcaption></figure>
In both rest, exercise and post-exercise examples, the zero crossing points do not cor-
respond precisely to the end of expirati... | 396 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 72 | <figure><img src="image_31.png" /><figcaption>Figure 3.9 – Examples of periods with computed inspiration-to-expiration times ratio above 1.5 from healthy dataset. Red and blue triangles indicate detected inspiration and expiration beginnings respectively.</figcaption></figure>
Inside each period, loops present similar... | 166 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 73 | <figure><img src="image_32.png" /><figcaption>Figure 3.10 – Periods with computed inspiration-to-expiration times ratio above 2 from COPD dataset. Red and blue triangles indicate detected inspiration and expiration beginnings respectively.</figcaption></figure>
In this sense, the approach based on TeleOx® recordings i... | 395 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 74 | <figure><img src="image_33.png" /><figcaption>Figure 3.11 – Flow-volume loops at different periods for the same subject. Flow and volume signals were estimated from pressure signal, units are arbitrary.</figcaption></figure>
<figure><img src="image_34.png" /><figcaption>Figure 3.12 – Flow-volume loops for periods pres... | 39 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 75 | <figure><img src="image_35.png" /><figcaption>Figure 3.13 – Comparison between inspiration-expiration amplitude and inspiratory amplitude</figcaption></figure>
<figure><img src="image_36.png" /><figcaption>Figure 3.14 – Periods resulting in outlier amplitudes</figcaption></figure>
<figure><img src="image_37.png" /><f... | 38 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 76 | <figure><img src="image_38.png" /><figcaption>Figure 3.16 – Periods where inspiration-expiration amplitude and inspiratory amplitude could not be estimated in COPD dataset.</figcaption></figure>
a depletion of the oxygen cylinder. Succeeding periods also present a decreasing pulse counter signal until a period where t... | 272 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 77 | <figure><img src="image_39.png" /><figcaption>Figure 3.17 – Within periods variability in healthy and COPD datasets</figcaption></figure>
### 3.4.5 ARIMA coefficients
Figure 3.18 gives examples of ARIMA coefficients estimated for a healthy subject and a COPD patient.
As expected, $\hat{\mu}$ values oscillate very cl... | 309 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 78 | <figure><img src="image_40.png" /><figcaption>(a) Healthy subject</figcaption></figure>
<figure><img src="image_41.png" /><figcaption>(b) COPD patient</figcaption></figure>
Figure 3.18 – Examples of series of estimated ARIMA coefficients from complete raw data recordings.
Consequently, features that are kept as pote... | 66 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 79 | <figure><img src="image_42.png" /><figcaption>Figure 3.19 – Examples of frequency spectrum from complete raw data recordings.</figcaption></figure> | 14 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 80 | # Chapter 4
# Validation of indicators of respiratory mechanics profiles
In this chapter, previously computed indicators are compared according to their capacities of detecting changes in the rest-effort context.
First, the performances of the indicators, alone or combined, are tested using a large number of supervi... | 332 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 81 | other words, differences between rest and effort periods are considered as independent from
individuality.
SuperLearner [105, 72] is used in order to limit the influence of the used method. For the present analysis, supervised classification includes rest and effort periods from all subjects from the same dataset (hea... | 577 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 82 | define the first 5 principal components. Mahalanobis distance and the method described above is then completed using the projected data.
Sensitivity and specificity is given for the cut-off threshold that minimizes the distance from the upper-left corner of the respective ROC curve, that is $\sqrt{FPR^2 + (1 - TPR)^2}... | 358 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 83 | <figure><img src="image_43.png" /><figcaption>Figure 4.1 – Extracted features example from a healthy subject recording. a. the original pressure signal. b. breathing rate, inspiratory amplitude and ARIMA coefficients extracted from 45-second windows of the pressure signal. c. Fourier coefficients from 45-second windows... | 120 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 84 | <figure><img src="image_44.png" /><figcaption>Figure 4.2 – Extracted features example from a COPD patient recording. **a.** the original pressure signal. **b.** breathing rate, inspiratory amplitude and ARIMA coefficients extracted from 45-second windows of the pressure signal. **c.** Fourier coefficients from 45-secon... | 70 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 85 | <figure><img src="image_45.png" /><figcaption>Figure 4.3 – ROC curves for the detection of exercise periods with SuperLearning using combinations of the proposed features</figcaption></figure>
<table><caption>Table 4.2 – Performance of SuperLearner in exercise detection for the COPD patients dataset using different pr... | 88 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 86 | <table><caption>Table 4.3 – Performance of GLMM in exercise detection for the healthy individuals dataset using different predictor variables and performance indices.</caption><thead><tr><th>Predictive variables</th><th>Accuracy</th><th>Sensitivity</th><th>Specificity</th><th>AUC</th></tr></thead><tbody><tr><td>Breathi... | 169 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 87 | <table><caption>Table 4.5 – Performance of one-class classification models in exercise detection for the healthy individuals dataset using different predictor variables and performance indices.</caption><thead><tr><th>Predictive variables</th><th>Accuracy</th><th>Sensitivity</th><th>Specificity</th><th>AUC</th></tr></t... | 330 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 88 | activities in the rest periods. For some of those patients, any movement can become really challenging and be a physical effort, as walking, standing up, showering, etc.
In both cases, this study demonstrates a significant gain in combining breathing rate with amplitude and potentially ARIMA coefficients, or using Fou... | 559 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 89 | # Chapter 5
## Remote monitoring
Following the study of the previous chapter, two updates have been proposed to improve the use of TeleOx® devices.
The first update concerns the firmware of the device which gains a new monitored variable. This update is directly related to the conclusions of the presented study.
Th... | 327 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 90 | sent through Internet to the server and can be assessed by healthcare providers in real time
through an online interface.
This project was developed in collaboration with the *Service de Pneumologie et Réan-
imation médicale* at the Hospital Pitié-Salpêtrière. During more than two months, I fre-
quently visited the un... | 207 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 91 | # Part III
## Analysis and results for exacerbation detection | 10 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 92 | # Chapter 1
## Medium-term monitoring of chosen features
The combination of breathing rate and amplitude allows for the detection of a greater number of respiratory mechanics adaptation in short-term changes. After implementation of these features in the monitoring device, we analyse if they are also relevant for the... | 372 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 93 | 6MWT results from patients medical records. Thus, each recording, composed by TeleOx®
data and medical data, corresponds to a patient's stay in the SSR respiratoire.
1.2 Data treatment
Every day, a TeleOx® records 288 data points. Each point is a vector, containing the time, status (with information about oxygen sour... | 539 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 94 | <table><caption>Table 1.1 – Clinical characteristics of COPD patients during periods of follow-up</caption><thead><tr><th></th><th>Total<br>(n = 70)</th><th>With exacerbation<br>(n = 29)</th><th>Without exacerbation<br>(n = 41)</th></tr></thead><tbody><tr><td>Men</td><td>42 (60.0%)</td><td>22 (75.9%)</td><td>20 (48.8%)... | 289 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 95 | <figure><img src="image_49.png" /><figcaption>Figure 1.2 – Time series from TeleOx® for a week from recording 22. Patient was on oxygen therapy during day and night. Although amplitude measures are high, there is no evidence that the TeleOx® was plugged in a NIV machine.</figcaption></figure>
After filtering NIV relat... | 166 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 96 | <figure><img src="image_51.png" /><figcaption>Figure 1.4 – Daily distributions corresponding to valid data from previously presented weeks of recordings from TeleOx®</figcaption></figure>
1.4 Discussion
It is interesting to notice the periodicity of the series presented above (Figures 1.1, 1.2 and 1.3). Patients typi... | 281 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 97 | Thus, in the next chapter, we seek at describing typical respiratory profiles present in
the data. | 16 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 98 | # Chapter 2
## Automatic detection of respiratory profiles
Previous observation pointed to the presence of daily seasonality in the recordings. Every day, patients need to adapt to different conditions. At the *SSR respiratoire* a daily planning is proposed for the patients. Still, participation to activities vary fr... | 345 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 99 | ## 2.1.2 Barycenters
In both clustering sets, for each selected week, daily distributions of breathing rate and amplitude were used to compute the corresponding Wasserstein barycenter.
Based on the Wasserstein distance between discrete distributions, the Wasserstein barycenter is a weighted average distribution $\nu$... | 471 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 100 | ## 2.2 Results
### 2.2.1 Week mean distribution
Figure 2.1 gives as examples the barycenters of the three weeks presented in the previous chapter.
<figure><img src="image_52.png" /><figcaption>Figure 2.1 – Wasserstein barycenters from weeks presented in Figure 1.4</figcaption></figure>
The barycenters are consisten... | 231 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 101 | The silhouette score for this classification is 0.29. The barycenters of the five obtained clusters are presented in Figure 2.3.
<figure><img src="image_54.png" /><figcaption>Figure 2.3 – Wasserstein barycenters representing the clusters for all weeks.</figcaption></figure>
Clusters 1, 2 and 3 are characterised by th... | 316 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 102 | This artifact can increase the performance according to the silhouette coefficient, while not giving interesting information about the similarity between the elements.
The next best classification seems to be obtained using the complete linkage method with 6 clusters. The silhouette score is 0.31. Its dendrogram is pr... | 175 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 103 | The other clusters 1, 2, 5 and 6 have more representative weeks. Cluster 1 is composed by 5 weeks, where patients have moderate breathing rate and amplitudes that vary along the days. In cluster 2 (8 weeks), there are patients who breathe at higher rates and also use the amplitude to adapt their breathing. Cluster 3, a... | 535 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 104 | The respiratory profiles could not be easily associated with exacerbation or stability, which suggests that population-based algorithms may be not adapted for exacerbation detection. Our hypothesis is that it is the changes from patient's own baseline that may be indicative of exacerbations, thus they need to be identi... | 49 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 105 | # Chapter 3
## Automatic detection of respiratory changes
The use of Wasserstein barycenters to identify types of respiratory profiles within the recordings indicates that most patients present a consistent weekly profile during their stay in the rehabilitation unit and that there is no common respiratory profile for... | 358 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 106 | number of points used is also 288. Those distributions represent what a day or night period is expected to look like for this individual.
The next step consists of comparing the series to these reference distributions. A sliding window is used to extract periods from the recording. The breathing rate-amplitude distrib... | 422 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 107 | <figure><img src="image_59.png" /><figcaption>Figure 3.2 – Performance for each recording with sliding window of 4 hours. Red and black lines represent recordings with and without exacerbations, respectively.</figcaption></figure>
## Example of sleep and wakefulness detection
The last week from recording 42 is presen... | 154 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 108 | <figure><img src="image_61.png" /><figcaption>Figure 3.4 – Barycenters for day (orange) and night (blue) from last week of recording 42</figcaption></figure>
demonstrates that is is not possible to identify a single point as sleep or wakefulness by itself. Instead, it is necessary to analyse each point according to it... | 311 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 109 | <figure><img src="image_63.png" /><figcaption>Figure 3.6 – Performance in function of the distance between reference distributions for day and night. Red and dark points represent recordings with and without exacerbations, respectively.</figcaption></figure>
Recording 30 (Figure 3.7c) is from a patient with less sever... | 357 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 110 | <figure><img src="image_64.png" /><figcaption>Figure 3.7 – Complete night detection for three recordings with exacerbation. First day of exacerbation is represented in red. Reference nocturnal periods are indicated by blue areas.</figcaption></figure>
### 3.2.1 Methods
Recordings from patients who did not exacerbate ... | 239 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 111 | transition score is used to estimate to which week a day is closer.
Days from the first and last weeks are used to confirm the correctness of the method.
## Variations of the method
The proposed method is compared with state of the art used descriptors. To do so, we test if changes could also be detected based on br... | 498 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 112 | <figure><img src="image_65.png" /><figcaption>Figure 3.8 – Comparison of individual accuracies for methods using daily distribution or mean of breathing rate alone or combined with the amplitude.</figcaption></figure>
or the last week. Figure 3.9 shows the individual accuracies with respect to the distances between th... | 173 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 113 | ## Comparison with clinical evolution
In Figure 3.10, the weeks classification performances with the barycenter of breathing rate and amplitude for all patients are compared to the change in the their 6MWT relative results (the percentages relative to expected distances). Points are colored according to the GOLD level... | 456 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 114 | <figure><img src="image_68.png" /><figcaption>(a) Series from first week</figcaption></figure>
<figure><img src="image_69.png" /><figcaption>(b) Barycenter for first week</figcaption></figure>
<figure><img src="image_70.png" /><figcaption>(c) Series of the last week</figcaption></figure>
<figure><img src="image_71.p... | 154 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 115 | <figure><img src="image_74.png" /><figcaption>Figure 3.13 – Initial and final reference barycenters of recording 27. Barycenters are estimated from first and last weeks of recording, respectively.</figcaption></figure>
<figure><img src="image_75.png" /><figcaption>Figure 3.14 – Wasserstein distances between each pair ... | 184 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 116 | Concerning the results obtained with the present dataset, there is a significant improvement when combining the features breathing rate and amplitude and comparing a distribution of points. In fact, when data is limited to a single data point, much of the information contained in the daily respiratory distribution is m... | 580 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 117 | # Chapter 4
## Taking temporality into account
using a Hidden Markov Model
We previously described the modeling of the reference distribution using the Waserstein barycenter. In this method the temporal character of the series is not taken into account.
In this chapter, we describe a different modeling approach, bas... | 395 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 118 | 4.1.1 Modeling example
The last week of recording 52 is used to model its baseline respiratory profile with HMM. For
this example, only breathing rate and amplitude are used, to simplify data visualisation.
The series used to train the HMM correspond to those presented in Figure 3.11c. The
resulting model is described... | 409 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 119 | <figure><img src="image_77.png" /><figcaption>Figure 4.2 – Series from last weeks of recording 52. Each point is colored according to the state that has been attributed to it by the Viterbi algorithm. State 0 is indicated in blue and state 1 in orange.</figcaption></figure>
<figure><img src="image_78.png" /><figcaptio... | 179 | Sciences médicales et de la santé | ||
tel-04015145 | Détection des changements de profil respiratoire des patients sous oxygénothérapie de longue durée | Juliana Alves Pegoraro | 2021 | fr | CC0 | Sciences du Vivant; Mathématiques | TEL | 120 | activity, although the patient can perform physical activities inside their room or rest outside.
Individual models with two hidden states are trained over daily series extracted from the last week of each recording. Hidden states are estimated using the individual models and the Viterbi algorithm. Hidden states 0 and... | 370 | Sciences médicales et de la santé |