Decision support for the intensive care

重症监护病房的危重病人面临危及生命的健康状况，因此，他们受到持续和密切的监测，以便在病情突然恶化时及时采取干预措施。监测涉及在较长的时间间隔内获取和归档大量信号和数据。监测数据包括生命体征、血气浓度、通气参数和各种实验室结果，以及医生的记录。监测还包括由医生对病人健康的几个方面进行的主观目视评估。当收集到视觉评估数据时，通常没有较少主观的替代方案可用于记录患者状态。视觉评估包括：1-患者的全身运动能力；2-意识水平，通过评价患者的眼球运动来确定；3-任何神经缺陷，通过评估身体某一部分（如腿、手臂或面部）的运动能力下降来判断；4-自发非机械通气呼吸参数，如频率和最大吸气量；5-由呼吸引起的运动，如胸部、腹部和其他次级肌肉的运动，以检测和评估收缩；6-以及所有皮肤方面和颜色，如紫绀、紫癜、红斑、皮肤坏死和烧伤。视觉监控虽然被广泛应用，但本质上仍然是主观和定性的。医生也可以通过触摸和按压病人的四肢来评估周围的低血管化，以评估他们的相对温度。事实上，病人体表的热量分布提供了体内血管化变化的信息。更具体地说，寒冷的四肢（例如手）表明存在血管化集中，这与深层器官（心脏、大脑和肾脏）的高度血管化相对应，损害了表层器官。外周低血管化发生在严重血流动力学应激的情况下，如心力衰竭、严重感染和大出血。这种评估是主观的和定性的，只有当医生在病人的床边时才有可能。我们的目标是支持为危重病人提供护理的医生所做的决定。我们的目标是使用计算机视觉和人工智能来帮助评估重症监护病房患者的痛苦。更具体地说，我们打算利用视频采集和机器学习的新发展，专注于评估患者的呼吸窘迫、躁动、意识水平和皮肤温度分布。我们的研究项目将通过持续、定量地监测危重病人的健康状况，改善对危重病人的护理，包括那些生活在偏远地区的病人。

Critically ill patients in the intensive care unit face life-threatening health conditions and thus, are continuously and closely monitored to enable prompt intervention in case of any sudden worsening of their condition. Monitoring involves acquiring and archiving a large number of signals and data over extended time intervals. Monitoring data includes vital signs, blood gas concentrations, ventilation parameters and various laboratory results, and physicians' notes. Monitoring also includes subjective visual assessments, performed by physicians, of several aspects of a patient's health. When visual assessment data is gathered there typically are no, less subjective, alternatives available to record patient state. Visual assessment includes: 1- the general motility of the patient; 2- the level of consciousness, determined through the evaluation of the patient's eyes movements; 3- any neurological deficit, done by evaluating the decrease in motility of one part of the body such as the leg, arm, or face; 4- the parameters of spontaneous non-mechanically ventilated respiration such as the frequency and the maximum inspired volume; 5- the movement induced by respiration such as the movement of the chest, abdomen, and other secondary muscles to detect and evaluate retractions; 6- and all skin aspects and color such as cyanosis, purpura, erythema, cutaneous necrosis, and burns. Visual monitoring, although widely employed, is nonetheless subjective and qualitative in nature. Physicians also evaluate the peripheral hypo-vascularization by touching and pressing a patient's extremities in order to assess their relative temperature. In fact, the heat distribution across a patient's body surface gives information about changes to the vascularization inside the body. More specifically, cold extremities (hands for instance) indicate the existence of a vascularization centralization which corresponds to a hyper-vascularization of the deep organs - heart, brain and kidneys - to the detriment of the superficial ones. Peripheral hypo-vascularization happens in situations of severe hemodynamic stress such as heart failures, severe infections, and massive bleedings. This evaluation is both subjective and qualitative, and is only possible when the physician is present at the patient's bedside. Our objective is to support the decisions made by physicians providing care to critically ill patients. We aim to use computer vision and artificial intelligence to help in the assessment of patients' distress in the intensive care unit. More specifically, we intend to focus on assessing patient's breathing distress, agitation, consciousness level, and skin temperature distribution using new developments in video acquisition and machine learning. Our research program will improve the care provided to critically ill patients, including those who are in remote regions, by providing continuous, and quantitative monitoring of their health distress situations.