Mathieu Feuilloy

Prediction of blood transfusion donation

The goal of the present study was to develop and evaluate machine learning algorithms for the prediction of blood transfusion donation. The machine learning algorithms studied included multilayer perceptrons (MLPs) and support vector machines (SVMs). The methods were evaluated retrospectively in a group of 600 patients and validated prospectively in a group of 148 patients. We reach a sensitivity of 65.8% and a specificity of 78.2% in the prospective group. This discrimination is very interesting because it could allow to propose to the patients, classified as non-donators, to give their blood in the future. Furthermore, the blood transfusion donation UCI corpus used, has been processed in a different manner than the initial marketing one. Therefore, this recent corpus could give a new training set for testing and improving machine learning methods in the future.

Prevalence and Causes of Normal Exercise Oximetry in the Calf in Patients with Peripheral Artery Disease and Limiting Calf Claudication.

OBJECTIVE In patients with claudication, an ankle brachial index (ABI) under 0.90 is considered to be abnormal and a sufficient argument for the arterial origin of exercise induced pain. Exercise transcutaneous oxygen pressure (Ex-tcpO2) can provide evidence of exercise induced regional blood flow impairment (RBFI) and confirm the arterial origin of walking induced pain. The frequency with which calf Ex-tcpO2 remains apparently normal in patients with claudication and abnormal ABI is unknown. Causes of these discrepant results have yet to be analysed. METHODS A retrospective analysis of 4575 Ex-tcpO2 tests performed on 3,281 patients was conducted. The focus was on patients with a history of calf claudication and ABI under 0.90. Duplicate or non-standard tests were excluded, as were patients with no pain or those able to walk more than 15 minutes (on a treadmill). Searches were conducted for possible explanations of normal calf Ex-tcpO2 in the selected patients. RESULTS Cardiorespiratory limitation was identified in 50 patients and isolated non-calf ischemia in 36 of the 106 patients selected. There was no obvious explanation during Ex-tcpO2, but clinical improvement after non-vascular treatment or total absence of improvement after a technically successful revascularisation was noted in 12 patients. Four patients were lost on follow up. Four patients improved after revascularisation, which suggests that the Ex-tcpO2 result was false negative. CONCLUSIONS Ex-tcpO2 is negative in more than 20% of tests performed in patients with an ABI under 0.90 and a history of calf claudication. In most cases, when excluding re-tests and non-limiting or non-calf claudication on the treadmill, non-calf ischemia or a non-vascular limitation occurring during the test were observed. This observation supports both the value of treadmill testing in patients with calf claudication assumed to be of arterial origin (ABI<0.90) and the use of Ex-tcpO2 to detect non-calf ischemia.

Creep test material rupture prediction by neural networks

This work focuses on acoustic emission analysis of mechanisms damage in fiber composite materials, subjected to heavy loads during a creep test. The goal of the present study was to develop and evaluate machine learning algorithms for the prediction of material rupture with creep test by traction method. This study aimed to predict if a tensile specimen will break in 30 seconds or not. Multilayer Perceptrons were trained retrospectively in a group of 80 samples moments and tested prospectively in a group of 16 tensile specimens. During the 5-cross validations we reached a sensitivity of 88% and a specificity of 88% in the prospective group. The mean area under the ROC (Receiver Operating Curves) was equal to 0.86. Those results are very interesting because they are a first important step in the lifetime prediction of material rupture before significant damages can occur.

Dimension reduction methods for the early syncope prediction by artificial neural networks

The aim of this study is to develop a method to predict unexplained syncope or presyncope occurrences. Diagnosis of syncope is currently based on the reproduction of symptoms in combination with hypotension and bradycardia induced by a 45-min of 60-80 head-upright tilt test (HUTT). The main drawback of this test concerns its duration which can reach 55 minutes if the patient does not faint. Thus we propose a new predicting tool which is only based on the 10 first minutes of the supine position of the HUTT. First, we describe how to acquire the variables of the disease and to achieve the dimension reduction methods. Then, at the end of the variables processing a neuronal method evaluates the prediction quality for a retrospective and a prospective group of patients. The best model compares very favorably with previous published results for other syncope detectors.

Exercise transcutaneous oximetry significantly modifies the diagnostic hypotheses and impacts scheduled investigations or treatments of patients with exertional limb pain

INTRODUCTION In lower extremity peripheral artery disease (PAD), transcutaneous oximetry at exercise (Ex-TcpO2) has been largely validated in research practice, but evidence of routine practice in various vascular laboratories is missing. We hypothesized that Ex-TcPO2 would change the diagnosis hypotheses, investigations and treatments for patients referred for exertional limb pain. MATERIAL & METHODS A multicenter prospective trial was conducted in nine different referral centers. Investigators performed Ex-TcpO2 and recorded investigations and treatments already scheduled for the patient. We encoded referral physicians diagnostic hypothesis. Before Ex-TcpO2, vascular physicians were asked to give their diagnosis hypotheses. A minimal decrease from rest of oxygen pressure (DROP)<minus 15mmHg defined the presence of exercise-induced ischemia on the area of interest. After Ex-TcPO2, we recorded post-test diagnostic hypothesis and investigations and treatments to be cancelled or performed. We compared the diagnosis hypotheses, scheduled investigations and treatments, before and after the Ex-TcpO2. RESULTS We included 603 patients (485 males: 80.4%), aged 64.7±9.8years. The post-test diagnosis hypothesis differed in 266 patients (44.1%; p<0.0001) and in 96 patients (15.9%) from the pre-test hypothesis of referring and vascular physician, respectively. This led to the recommendation to cancel 27 scheduled investigations or treatments of a total cost of ~130,000 euros. DISCUSSION Ex-TcPO2 in patients with exertional limb pain is applicable in various vascular institutions, and significantly modifies the diagnostic hypotheses and impacts scheduled investigations or treatments of patients with exertional limb pain.

Simultaneous pain intensity rating and quantification of ischemia throughout exercise and recovery in proximal versus distal arterial claudication

Data on simultaneous hemodynamic changes and pain rating estimation in arterial claudication while walking are lacking. This study was conducted to determine if a difference in transcutaneous oxygen pressure (tcpO2) exists between proximal and distal localization at pain appearance (PAINapp), maximal pain (PAINmax) and pain relief (PAINrel) in proximal or distal claudication and if a relationship exists between tcpO2 changes and pain intensity. We analyzed the pain rating (Visual Analog Scale (VAS)) to lower limb ischemia, measured with the decrease from rest of oxygen pressure (DROP) tcpO2 index during constant-load treadmill tests in patients with calf (n = 41) or buttock (n = 19) claudication. Calves versus buttocks results were analyzed with ANOVA tests. The R2 correlation coefficient between individual VAS versus DROP was calculated. Ischemia intensity versus pain rating changes were correlated. Significant ischemia was required for pain appearance, but pain disappeared despite the persistence of ischemia. We observed no statistical difference for DROP at PAINapp, PAINmax or PAINrel between proximal or distal claudication. A significant correlation between pain rating versus DROP was found: from PAINapp to PAINmax, R2 = 0.750 (calves) and 0.829 (buttocks), and from PAINmax to PAINrel, R2 = 0.608 (calves) and 0.560 (buttocks); p<0.05. Pain appeared after a significant decrease of hemodynamic parameters but disappeared while parameters were not normalized. No difference in pain rating was found in proximal versus distal claudication.

Comparison of Feature Selection Methods for Syncope Prediction

The aim of this study is to develop a method to predict unexplained syncope. Its diagnosis is currently based on the reproduction of symptoms induced by a 45-min of 60-80deg head-upright tilt test (HUTT). The main drawback of this test concerns its duration which can reach 45 minutes, therefore our study proposes an analysis which is only based on the 10 first minutes of the test. An important number of variables is obtained during the HUTT. To reduce and to select the most relevant variables, many feature selection methods are used and compared to obtain groups of pertinent variables. We used classification tools to achieve significant syncope outcome prediction.