Fully-Automated-Detection-and-Quantification-of-Macular-Fluid_2018

需积分: 9 177 浏览量 2020-07-28 17:53:05 上传评论收藏 1.4MB PDF 举报

这篇名为“Fully-Automated-Detection-and-Quantification-of-Macular-Fluid_2018_Ophthalmo.pdf”的文章是关于眼科医学图像处理领域的深度学习应用，特别关注了利用深度学习技术自动检测和量化视网膜黄斑部积液（Macular Fluid）在光学相干断层扫描（Optical coherence tomography, OCT）图像中的方法。该研究由多个医学专家和科学家共同完成，旨在开发并验证一种全自动检测和量化黄斑部积液的算法。文章设计了一个诊断工具，参与者包括患有不同类型眼部疾病的患者，如新生血管性年龄相关性黄斑变性（n = 400），糖尿病黄斑水肿（n = 400）以及视网膜静脉阻塞（n = 400）。这些患者的OCT图像来自两种不同的设备：Zeiss Cirrus和Heidelberg Spectralis。研究方法基于深度学习，目的是自动识别并量化视网膜内囊状积液（Intraretinal Cystoid Fluid, IRC）和视网膜下积液（Subretinal Fluid, SRF）。算法的性能通过与两位独立阅片中心分级员的手动共识阅读结果进行比较来评估，主要衡量指标包括检测的准确性（通过受试者工作特征曲线下的面积，Area Under the Receiver Operating Characteristics curves, AUC）、精确度（Precision）和召回率（Recall）。研究结果显示，新设计的深度学习驱动的全自动诊断方法在所有三种黄斑病变中检测和量化IRC时表现出极高的准确性，平均AUC为0.94（范围：0.91-0.97），平均精确度为0.91，平均召回率为0.84。对于SRF的检测和测量，该方法也表现出高精度，AUC为0.92（范围：0.86-0.98），平均精确度为0.61，平均召回率为0.81。在新生血管性AMD和RVO病例中，该方法的性能优于DME，因为DME在数据集中较为罕见。这项工作展示了深度学习在医疗图像分析中的潜力，特别是在眼科疾病诊断和监测方面。通过自动化和精确的液体检测，可以极大地提高医生的工作效率，减少人为误差，并有助于早期发现和治疗，从而改善患者的预后。这种方法的应用可能为临床实践带来革命性的变化，特别是在处理大量OCT图像的大型医疗中心，可以节省大量时间和资源。同时，这种技术也为未来在其他医学成像领域中开发类似的自动分析工具提供了参考。

资源推荐

资源详情

资源评论

Fully Automated Detection and

Quantiﬁcation of Macular Fluid in OCT Using

Deep Learning

Thomas Schlegl, MSc,

1,4

Sebastian M. Waldstein, MD, PhD,

Hrvoje Bogunovic, PhD,

Franz Endstraßer, MD,

Amir Sadeghipour, PhD,

Ana-Maria Philip, MD,

Dominika Podkowinski, MD,

Bianca S. Gerendas, MD, MSc,

Georg Langs, PhD,

Ursula Schmidt-Erfurth, MD

1,2

Purpose: Development and validation of a fully automated method to detect and quantify macular ﬂuid in

conventional OCT images.

Design: Development of a diagnostic modality.

Participants: The clinical dataset for ﬂuid detection consisted of 1200 OCT volumes of patients with neo-

vascular age-related macular degeneration (AMD, n ¼ 400), diabetic macular edema (DME, n ¼ 400), or retinal

vein occlusion (RVO, n ¼ 400) acquired with Zeiss Cirrus (Carl Zeiss Meditec, Dublin, CA) (n ¼ 600) or Heidelberg

Spectralis (Heidelberg Engineering, Heidelberg, Germany) (n ¼ 600) OCT devices.

Methods: A method based on deep learning to automatically detect and quantify intraretinal cystoid ﬂuid

(IRC) and subretinal ﬂuid (SRF) was developed. The performance of the algorithm in accurately identifying ﬂuid

localization and extent was evaluated against a manual consensus reading of 2 masked reading center graders.

Main Outcome Measures: Performance of a fully automated method to accurately detect, differentiate, and

quantify intraretinal and SRF using area under the receiver operating characteristics curves, precision, and recall.

Results: The newly designed, fully automated diagnostic method based on deep learning achieved optimal

accuracy for the detection and quantiﬁcation of IRC for all 3 macular pathologies with a mean accuracy (AUC) of 0.94

(range, 0.91e0.97), a mean precision of 0.91, and a mean recall of 0.84. The detection and measurement of SRF were

also highly accurate with an AUC of 0.92 (range, 0.86e0.98), a mean precision of 0.61, and a mean recall of 0.81, with

superior performance in neovascular AMD and RVO compared with DME, which was represented rarely in the

population studied. High linear correlation was conﬁrmed between automated and manual ﬂuid localization and

quantiﬁcation, yielding an average Pearson’s correlation coefﬁcient of 0.90 for IRC and of 0.96 for SRF.

Conclusions: Deep learning in retinal image analysis achieves excellent accuracy for the differential detec-

tion of retinal ﬂuid types across the most prevalent exudative macular diseases and OCT devices. Furthermore,

quantiﬁcation of ﬂ uid achieves a high level of concordance with manual expert assessment. Fully automated

analysis of retinal OCT images from clinical routine provides a promisi ng horizon in improving accuracy and

reliability of retinal diagnosis for research and clinical practice in ophthalmology. Ophthalmology 2018;125:549-

558 ª 2017 by the American Academy of Ophthalmology. This is an open access article under the CC BY-NC-ND

license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Supplemental material available at www.aaojournal.org.

OCT has profoundly disrupted conventional dia gnostic and

therapeutic strategie s in clinical management and has led to

paradigm shifts in the understanding of macular disease. Although

OCT has continuously undergone hardware improvements since

its inception,

signiﬁcantly less progress has been made in the ﬁeld

of software methods to analyze clinical OCT data. The number of

patients with macular disease requiring efﬁcient disease

management based on OCT in clinical practice continu es to

increase, similarly to the amount of image data produced by

advanced OCT technology such as swept source. Therefore, the

feasibility of manual OCT assessment in clinical practice has

become largely unrealistic. Likewise, poor reproducibility

between OCT assessors, even in a research setting, also has

been reported.

Consequently, automated and reproducible

analysis of clinical OCT data represents an important unmet

need and a promising perspective for clinical practice.

Speciﬁcally, there is a clear need to advance automated analysis

beyond a purely anatomic presence/absence detection to an

accurate measurement of markers for disease activity.

The majority of available analysis software tools for OCT is

limited to the measurement of retinal (layer) thickness, despite the

fact that prior studies demonstrated the limited value of this

biomarker for visual prognosis and disease management.

practice, most physicians use qualitative OCT biomarkers, such

as the presence of intraretinal cystoid ﬂuid (IRC) and subretinal

ﬂuid (SRF), to inform retreatment decisions in individualized

549ª 2017 by the American Academy of Ophthalmology

This is an open access article under the CC BY-NC-ND license

(http://creativecommons.org/licenses/by-nc-nd/4.0/). Published by Elsevier Inc.

https://doi.org/10.1016/j.ophtha.2017.10.031

ISSN 0161-6420/17

therapy of macular diseases. These biomarkers are also

sophisticated prognostic factors for visual outcome, with IRC

representing 1 of the most important variables associated with

vision loss and SRF possibly enhancing visual prognosis.

Prior studies have proposed to detect or quantify macular

ﬂuid in OCT in an automated manner.

However, they are

limited to only being able to detect ﬂuid presence or

absence without measuring its extent and distribution, by

a lack of differentiation between IRC and SRF, or by

evaluation limited to 1 particular OCT device or disease.

In this study, we present a fully automated artiﬁcial intelli-

gence method to detect and quantify IRC and SRF in macular

OCT volume scans. We validate our method on a large dataset of

eyes presenting with the major relevant exudative macular dis-

eases, that is, neovasc ular age-related macular degeneration

(AMD), diabetic macular edema (DME), and macular edema

secondary to retinal vein occlusion (RVO), imaged with the

most commonly used OCT devices (Heidelberg Spectralis,

Heidelberg Engineering, Heidelberg, Germany, and Zeiss

Cirrus, Carl Zeiss Meditec, Dublin, CA).

Methods

Image Dataset

This study followed the tenets set forth in the Declaration of Helsinki,

and approval was obtained by the Ethics Committee of the Medical

University of Vienna. A total of 1200 completely anonymized OCT

volume scans of eyes affected by the major diseases typically causing

macular ﬂuid (AMD, DME, and RVO) were extracted from the Vienna

Reading Center database. Furthermore, for each disease we selected

scans by 2 different OCT devices (Cirrus HD-OCT, Carl Zeiss Meditec,

and Spectralis OCT, Heidelberg Engineering), resulting in 6 distinct

groups (disease  device) of OCT scans.

For detection of ﬂuid, we used 1200 anonymized OCT volumes of

patients with neovascular AMD (n ¼ 400), DME (n ¼ 400), and RVO

(n ¼ 400) acquired with Cirrus (50%) or Spectralis (50%) devices. In

each disease/device group, 50% of the OCT scans showed retinal ﬂuid.

All scans were graded for the presence of IRC and SRF in a quality-

controlled reading center setting by 2 independent readers of the Vienna

Reading Center supervised by experienced retinal specialists. The

grading was performed on full-screen, high-resolution 27-inch monitors

in validated grading software specially designed for the annotation of

these features. In questionable cases, a consensus grading of the readers

and the supervisor was performed. Only scans with a clear consensus

annotation between the readers were taken into the sample. Scans with

primarily low image quality (i.e., scans with sections cut off because of

improper positioning during image acquisition or scans with strong

motion artifacts causing misalignment and blurring of sections) were

excluded. We randomly selected 100 scans with ﬂu id and 100 scans

without ﬂuid per disease/device group from the OCT database, resulting

in a total of 1200 scans.

For evaluation of ﬂuid quantiﬁcation, datasets with complete

manual annotations of IRC and SRF that were available at the Vienna

Reading Center were randomly selected. For neovascular AMD, we

used a Cirrus dataset of 152 scans as reported previously

and a

Spectralis dataset consisting of 60 scans. For DME, we used a dataset

of 16 Cirrus scans and a dataset of 16 Spectralis scans. For RVO, we

used a Cirrus dataset of 100 scans and a Spectralis dataset consisting

of 10 scans. The procedure used for manual annotation of macular

ﬂuid has been reported in detail by Waldstein et al.

Each Cirrus cube

scan consisted of 128 B-scans with a resolution of 512  1024

pixels, with the exception of Cirrus RVO data, which consisted of

200 B-scans with a resolution of 200  1024 pixels. All cube scans

acquired on the Spectralis device consisted of 49 B-scans at a

resolution of 512  496 pixels. Both machines acquire volumes

covering approximately 6  6  2mm

corresponding to a ﬁeld of

view of 20



 20



Description of the Automated Method

We developed a software to perform classiﬁcation among IRC,

SRF, and nonﬂuid regions for each location (¼ pixel) in the OCT

image by using and further developing semantic segmentation,

method based on convolutional neural networks. Speciﬁcally, we

applied deep learning, a state-of-the-art machine learning tech-

nique in the ﬁeld of artiﬁcial intelligence that learns the mapping

from OCT images to pixel-level class labels based on large

amounts of labeled training data. Deep learning models allow one

to learn meaningful abstract data representations. Following the

semantic segmentation approach, the neural network maps an input

image of a speciﬁc size to an image of corresponding class labels of

the same size. The proposed neural network comprises 2 pro-

cessing components, an encoder that transforms an input image

into an abstract representation and a decoder that maps the abstract

representation to an image of clinical class labels assigning each

pixel a class such as normal tissue, IRF, or SRF.

The mapping of the encoder from raw images to abstract repre-

sentations (embeddings) was not computed on the basis of pre-

speciﬁed mathematic descriptions (handcrafted features), but the

encoder parameters were automatically learned solely on the basis of

annotated data used during training. The data embedding learned was

optimized in such a way that it was optimal for the generation of a

corresponding image of class labels. The mapping of the encoder from

raw images to the data embedding needed to generate the label image,

and the mapping of the decoder from the embedding to a full input

resolution label image were learned simultaneously (end-to-end). The

encoder and the decoder comprised a set of computing blocks (layers),

where the layers of the decoder virtually inversed the operations of the

encoder conditioned by the low-dimensional embedding learned by

the encoder. A simpliﬁed overview of the encoder-decoder architec-

ture is shown in Figure 1, and a detailed description of the algorithm

appears in the Appendix (available online at www.aaojournal.org).

The runtime of the computation depended on the resolution of

OCT volumes. For Cirrus scans, the computations took approxi-

mately 70 seconds, whereas for Spectralis scans, computations

took approximately 30 seconds per volume on a TitanX graphical

processing unit (Nvidia, Santa Clara, CA) with NVIDIA Cuda

(version 6.5) and cuDNN library (version 2).

Python, Theano,

and the Lasagne library were used to train and evaluate the deep

learning model.

Experimental Setup

Because of differences in the appearance of Spectralis and Cirrus

scans, we trained separate classiﬁers for Spectralis and Cirrus

scans. We used 70 pixel-wise annotated OCT scans performing

10-fold cross-validation including 7 scans per split for training and

testing of the model on Spectralis data. We performed 4-fold cross-

validation for training and testing of the model on Cirrus data,

where we had 257 pixel-wise annotated OCT scans and used 64

and 65 scans per split. Both datasets comprised neovascular AMD

and RVO cases. When considerably fewer scans were used to train

the model, results showed that the proposed method also achieved

a similar performance. However, the achievable performance, and

at the same time the applicability for clinical routine, increased

with the number of training volumes. When the model trained on

Cirrus data was used on Spectralis data and vice versa, results on

quantifying IRC or SRF showed that the model trained on Cirrus

Ophthalmology Volume 125, Number 4, April 2018

550

剩余9页未读，继续阅读

评论收藏

内容反馈

haimianjie2012

粉丝: 682
资源: 50

Fully-Automated-Detection-and-Quantification-of-Macular-Fluid_20...

最新资源

Fully-Automated-Detection-and-Quantification-of-Macular-Fluid_20...

AFQ_Browser-0.1.2-py3-none-any.whl.zip

AFQ_Browser-0.2.1-py3-none-any.whl.zip

quantification.rar_matlab语音量化_quantification

AFQ_Browser-0.1.2-py2-none-any.whl.zip

AFQ_Browser-0.1-py3-none-any.whl.zip

ExploreASL_Import.m_MR_ROI_matlab_TheInvestigators_ASL_

离散控制Matlab代码-Automatic-Detection-and-Uncertainty-Quantification-of-Land

Quantification.zip_image transform_数据离散化_离散面_面元_面成像

AFQ-Browser-0.3-py3-none-any.whl.zip

A Simple, Rapid and Sensitive Method for the Quantification of Methyl-Beta-Cyclodextrin

Cognitive radio_ brain-empowered wireless communications

UQ_of_ML.7z

DFA.rar_DFA_correlation matlab_time series

4D-CT-lung-deformable-registration-.zip_图形图像处理_Visual_C++_

Channel Coding in Communication Networks

Quantification of Building Seismic Performance Factors

2023年美赛获奖E类论文_2301428.pdf

Raman Spectroscopy a Promising Technique for Quantification and Monitoring Heat-induced Formation of Trans Fatty Acids in Oil

Uncertainty Quantification and Predictive Computational Science 中文 part2

Identification and Quantification of Volatile and Semi-volatile Chemicals by Gas Flow Headspace-Liquid Phase Micro Extraction

Origin软件在福林-酚试剂法测定蛋白含量实验数据处理中的应用.pdf

matlab匹配滤波代码-edge_staining_quantification:MATLAB代码以量化边缘染色

信息安全_数据安全_Machine Learning and Risk Quanti.pdf

Automated Ki-67 Quantification of Immunohistochemical Staining Image of Human Nasopharyngeal Carcinoma Xenografts

YOLOv8-deepsort 实现智能车辆目标检测+车辆跟踪+车辆计数

Transformer模型实现长期预测并可视化结果（附代码+数据集+原理介绍）

最新资源