Python编程与心血管疾病风险预测：结合Google应用与图像分析

需积分: 7 153 浏览量更新于2024-09-09 收藏 1.29MB PDF 举报

本篇资源主要关注的是Python在医疗健康领域的具体应用，特别是与谷歌研究相关的项目。文章标题“python ebook”虽然直接提及Python，但结合提供的描述和部分内容，我们可以推断这并非一本单纯的Python编程书籍，而是围绕如何利用Python技术处理和分析与心血管疾病风险预测相关的数据。文章讨论了风险分层在心血管疾病管理中的核心作用，强调全球范围内这类疾病仍然是首要死因，尽管已有诸如Pooled Cohort Equations、Framingham模型和SCORE等风险计算器广泛使用。文章深入探讨了如何通过结合表型信息，尤其是血管健康方面的数据，来细化或重新分类个人的风险预测。其中一个关键示例是冠状动脉钙化，这种成像技术提供了额外信号，能显著提升风险评估的准确性。在这一背景下，Python作为数据分析的强大工具，可能被用于开发算法，处理来自Google Research的大型数据集，如冠状动脉钙化的图像数据，以及整合多源数据进行风险模型的构建和优化。作者们，包括来自Google、Verily Life Sciences（旧金山）和斯坦福大学的研究者，如Ryan Poplin、Avinash V. Varadarajan等人，共同贡献了这篇研究，强调了Python在生物医学领域，特别是在心血管疾病风险管理中的实际应用。他们指出，通过利用Python的高效性和灵活性，可以实现风险预测模型的个性化和精确度提升，这对于公共卫生和临床决策具有重要意义。此外，文章还可能提到了Python库，如NumPy、Pandas和SciPy等，这些工具对于处理大规模生物医学数据、执行机器学习算法以及可视化结果至关重要。同时，由于引用了DOI链接，表明这篇研究已经发表在某个科学期刊上，这意味着其内容经过同行评审，具备较高的学术价值和可信度。这本关于Python的电子书或论文可能聚焦于如何通过Python技术推动心血管疾病风险评估的创新方法，展示了Python在医疗数据分析中的实际应用和科研价值。读者可以从中了解到如何将Python与医学成像数据相结合，以提高疾病风险预测的精度，并了解如何在这个领域开展前沿研究。

Articles

https://doi.org/10.1038/s41551-018-0195-0

Google Research, Google, Mountain View, CA, USA.

Verily Life Sciences, South San Francisco, CA, USA.

Division of Cardiovascular Medicine,

Stanford School of Medicine, Stanford, CA, USA.

These authors contributed equally: Ryan Poplin, Avinash V. Varadarajan, Lily Peng and Dale R. Webster.

*e-mail: lhpeng@google.com

isk stratification is central to identifying and managing groups

at risk for cardiovascular disease, which remains the leading

cause of death globally

. Although the availability of cardio-

vascular disease risk calculators, such as the Pooled Cohort equa-

tions

, Framingham

3–5

and Systematic COronary Risk Evaluation

(SCORE)

6,7

, is widespread, there are many efforts to improve

risk predictions. Phenotypic information, particularly of vascu

lar health, may further refine or reclassify risk prediction on an

individual basis. Coronary artery calcium is one such example,

for which it has been shown that additional signals from imaging

improve risk stratification

. The current standard-of-care for the

screening of cardiovascular disease risk requires a variety of vari

ables derived from the patient’s history and blood samples, such

as age, gender, smoking status, blood pressure, body mass index

(BMI), glucose and cholesterol levels

. Most cardiovascular risk

calculators use some combination of these parameters to identify

patients at risk of experiencing either a major cardiovascular event

or cardiac-related mortality within a pre-specified time period, such

as ten years. However, some of these parameters may be unavail

able. For example, in a study from the Practice INNovation And

CLinical Excellence (PINNACLE) electronic-health-record-based

cardiovascular registry, the data required to calculate the 10-year

risk scores were available for less than 30% of the patients

. This

was largely due to missing cholesterol values

, which is not surpris-

ing given that a fasting blood draw is required to obtain these data.

In this situation, BMI can be used in the place of lipids for a prelimi

nary assessment of cardiovascular health

11–13

. We therefore explored

whether additional signals for cardiovascular risk can be extracted

from retinal images, which can be obtained quickly, cheaply and

non-invasively in an outpatient setting.

Markers of cardiovascular disease, such as hypertensive reti

nopathy and cholesterol emboli, can often manifest in the eye.

Furthermore, because blood vessels can be non-invasively visual

ized from retinal fundus images, various features in the retina,

such as vessel calibre

14–20

, bifurcation or tortuosity

, microvascular

changes

22,23

and vascular fractal dimensions

24–26

, may reflect the sys-

temic health of the cardiovascular system as well as future risk. The

clinical utility of such features still requires further study. In this

work, we demonstrate the extraction and quantification of multiple

cardiovascular risk factors from retinal images using deep learning.

Machine learning has been leveraged for many years for a vari

ety of classification tasks, including the automated classification

of eye disease. However, much of the work has focused on ‘feature

engineering’, which involves computing explicit features specified

by experts

27,28

. Deep learning is a family of machine-learning tech-

niques characterized by multiple computation layers that allow an

algorithm to learn the appropriate predictive features on the basis

of examples rather than requiring features to be hand-engineered

Recently, deep convolutional neural networks—a special type of

deep-learning technique that has been optimized for images—have

been applied to produce highly accurate algorithms that diagnose

diseases, such as melanoma

and diabetic retinopathy

31,32

, from

medical images, with comparable accuracy to that of human experts.

Results

We developed deep-learning models using retinal fundus images

from 48,101 patients from the UK Biobank (http://www.ukbio

bank.ac.uk/about-biobank-uk) and 236,234 patients from EyePACS

(http://www.eyepacs.org) and validated these models using images

from 12,026 patients from the UK Biobank and 999 patients from

EyePACS (Table 1). The mean age was 56.9 ± 8.2 years on the UK

Biobank clinical validation dataset and 54.9 ± 10.9 years in the

EyePACS-2K clinical validation dataset. The UK Biobank popula

tion was predominantly Caucasian, while the EyePACS patients

were predominantly Hispanic. Haemoglobin A1c (HbA1c) mea

surements were available only in 60% of the EyePACS population.

Because this population consisted of mostly diabetic patients pre

senting for diabetic retinopathy screening, the mean HbA1c level of

this population was 8.2 ± 2.1%—well above the normal range. UK

Biobank participants were recruited from a UK general population,

Prediction of cardiovascular risk factors from

retinal fundus photographs via deep learning

Ryan Poplin

1,4

, Avinash V. Varadarajan

1,4

, Katy Blumer

, Yun Liu

, Michael V. McConnell

2,3

Greg S. Corrado

, Lily Peng

1,4

* and Dale R. Webster

1,4

Traditionally, medical discoveries are made by observing associations, making hypotheses from them and then designing and

running experiments to test the hypotheses. However, with medical images, observing and quantifying associations can often

be difficult because of the wide variety of features, patterns, colours, values and shapes that are present in real data. Here, we

show that deep learning can extract new knowledge from retinal fundus images. Using deep-learning models trained on data

from 284,335 patients and validated on two independent datasets of 12,026 and 999 patients, we predicted cardiovascular

risk factors not previously thought to be present or quantifiable in retinal images, such as age (mean absolute error within 3.26

years), gender (area under the receiver operating characteristic curve (AUC) = 0.97), smoking status (AUC = 0.71), systolic

blood pressure (mean absolute error within 11.23 mmHg) and major adverse cardiac events (AUC = 0.70). We also show that

the trained deep-learning models used anatomical features, such as the optic disc or blood vessels, to generate each prediction.

NATURE BIOMEDICAL ENGINEERING | VOL 2 | MARCH 2018 | 158–164 | www.nature.com/natbiomedeng

158

下载后可阅读完整内容，剩余8页未读，立即下载

cepacia

粉丝: 0

Python编程与心血管疾病风险预测：结合Google应用与图像分析

分享几本python经典电子书籍

Python库 | wattpad_ebook-1.0.2.tar.gz

nlp自然語言處理python ebook

python-handbook_handbook_python_ebook_

python-ebook-tools:电子书工具的Python端口

python_ebook.tar.gz

wenfujing-python_ebook-master.zip

Learning Penetration Testing with Python [eBook] pdf和代码

python unittest ebook

Python库 | ebook_homebrew-2.1.0.tar.gz

最新资源