可解释的机器学习模型用于对未确定来源的栓塞性中风中状动脉斑块进行分类
medRxiv : the preprint server for health sciences
|November 22, 2024
概括
一种可解释的机器学习模型准确地识别了不确定源 (ESUS) 患者的栓塞性中风中的罪祸首 - - 状动脉斑块. 该模型使用斑块和化特征,提供临床解释性和识别高风险斑块指标.
科学领域:
- 心血管成像 - 心血管成像
- 神经学 神经学
- 人工智能在医学中的应用
背景情况:
- 源不明的栓塞性中风 (ESUS) 可能与动脉斑块有关,其狭窄率低于50%.
- 斑块的脆弱性受到诸如斑块内出血 (IPH),富含脂质的死核 (LRNC),周围血管脂肪组织 (PVAT) 和化特征等因素的影响.
- 可解释的机器学习 (ML) 提供了一条克服斑块分类中的"黑子"模型临床解释性限制的途径.
研究的目的:
- 应用一种可解释的ML方法,使用夏普利添加式解释 (SHAP) 框架.
- 在ESUS患者中将化动脉斑块分类为罪祸首或非罪祸首.
- 为了提高ML模型的临床解释性,用于动脉斑块分析.
主要方法:
- 对患有单边前部循环ESUS的患者进行部CT血管造影的回顾性分析.
- 从手动细分和半自动软件中提取斑块级和化级特征.
- 培训和比较三个CatBoost ML模型 (板块级,化级,组合) 结合SHAP进行决策解释.
主要成果:
- 组合的ML模型实现了0.77的曲线下测试面积 (AUC) 和0.82.8的准确性.
- 组合模型的表现优于仅在斑块水平或化水平特征上训练的模型.
- 关键的预测特征包括斑块厚度,IPH/LRNC体积比,PVAT体积,化最小密度和化体积与平均密度比,斑块厚度>2.6毫米被确定为潜在的高风险值.
结论:
- 结合非化斑块和化特征的ML模型可以有效地分类罪祸首化动脉斑块.
- SHAP框架为ML模型决策提供了至关重要的临床解释性.
- 这种方法有助于识别高风险斑块特征和潜在的值,与中风风险分层相关.
相关概念视频
Classifying Matter by Composition
90.1K
Matter: Pure Substances and Mixtures
According to its composition, the matter can be classified into two broad categories — pure substances and mixtures.
A pure substance is a form of matter that has a constant composition throughout with uniform properties. For example, any sample of sucrose has the same composition and same physical properties, such as melting point, color, and sweetness, regardless of the source from which it is isolated.
A mixture is composed of two or...
According to its composition, the matter can be classified into two broad categories — pure substances and mixtures.
A pure substance is a form of matter that has a constant composition throughout with uniform properties. For example, any sample of sucrose has the same composition and same physical properties, such as melting point, color, and sweetness, regardless of the source from which it is isolated.
A mixture is composed of two or...
90.1K
Simplified Synchronous Machine Model
769
The Synchronous Machine Model is a fundamental tool in analyzing and ensuring the transient stability of power systems. This model simplifies the representation of a synchronous machine under balanced three-phase positive-sequence conditions, assuming constant excitation and ignoring losses and saturation. The model is pivotal for understanding the behavior of synchronous generators connected to a power grid, particularly during transient events.
In this model, each generator is connected to a...
In this model, each generator is connected to a...
769
Wind Turbine Machine Models
593
In the growing field of wind energy, incorporating wind turbine models into transient stability analysis is essential. Induction and synchronous machines are the primary models used, with induction machines being prevalent due to their simplicity and reliability.
Induction machines interact through the rotating magnetic field generated by the stator and the rotor. The key parameter is slip, which is the difference between synchronous speed and rotor speed relative to synchronous speed. Slip is...
Induction machines interact through the rotating magnetic field generated by the stator and the rotor. The key parameter is slip, which is the difference between synchronous speed and rotor speed relative to synchronous speed. Slip is...
593
Classifying Matter by State
102.9K
Chemistry is the study of matter and the changes it undergoes. Matter is anything that has mass and occupies space. Matter is all around us; the air, water, soil, mountains, even our bodies are all examples of matter. Matter is divided into three states — solid, liquid, and gas — that are commonly found on earth. The fourth state of matter, plasma, occurs naturally in the interiors of stars.
102.9K
How Data are Classified: Numerical Data
37.7K
Data that are countable or measurable in specific units are called numerical or quantitative data. Quantitative data are always numbers. Quantitative data are the result of counting or measuring the attributes of a population. Amount of money, pulse rate, weight, number of people living in a town, and number of students who opt for statistics are examples of quantitative data.
Quantitative data may be either discrete or continuous. All quantitative data that take on only specific numerical...
Quantitative data may be either discrete or continuous. All quantitative data that take on only specific numerical...
37.7K
Pulmonary Embolism I: Introduction
667
Pulmonary embolism (PE) occurs when a thrombus, fat or air embolus, amniotic fluid, or tumor tissue blocks one or more pulmonary arteries. These blockages originate in the venous system or the right side of the heart.EtiologyPE primarily arises from deep vein thrombosis (DVT) and other hypercoagulable states, such as inherited thrombophilias. Additional etiological factors include venous stasis, commonly seen in obesity, and endothelial injury from surgery and trauma. Less common causes include...
667


