1 D-CryptO Deep learning -based analysis of colon organoid morphology from brightfield images

2025-04-24 0 0 993.12KB 29 页 10玖币

侵权投诉

D-CryptO: Deep learning-based analysis of colon organoid morphology from brightfield

images

Lyan Abdul1, Jocelyn Xu2, Alexander Sotra1, Abbas Chaudary3, Jerry Gao4, Shravanthi

Rajasekar3, Nicky Anvari1, Hamidreza Mahyar5, and Boyang Zhang1,3*

1School of Biomedical Engineering, McMaster University, 1280 Main Street West, Hamilton,

ON, L8S 4L8, Canada

2Faculty of Engineering, McMaster University, 1280 Main Street West, Hamilton, ON, L8S

4L8, Canada

3Department of Chemical Engineering, McMaster University, 1280 Main Street West,

Hamilton, ON, L8S 4L8, Canada

4Faculty of Science, McGill University, 845 Sherbrooke Street West, Montreal, QC H3A

0G4, Canada

5W Booth School of Engineering Practice and Technology, McMaster University, 1280 Main

Street West, Hamilton, ON, L8S 4L8, Canada

*Correspondence to zhangb97@mcmaster.ca

Keywords: deep learning; image analysis; organoids; morphology; drug testing; brightfield

images

Abstract

Stem cell-derived organoids are a promising tool to model native human tissues as they

resemble human organs functionally and structurally compared to traditional monolayer cell-

based assays. For instance, colon organoids can spontaneously develop crypt-like structures

similar to those found in the native colon. While analyzing the structural development of

organoids can be a valuable readout, using traditional image analysis tools makes it

challenging because of the heterogeneities and the abstract nature of organoid morphologies.

To address this limitation, we developed and validated a deep learning-based image analysis

tool, named D-CryptO, for the classification of organoid morphology. D-CryptO can

automatically assess the crypt formation and opacity of colorectal organoids from brightfield

images to determine the extent of organoid structural maturity. To validate this tool, changes

in organoid morphology were analyzed during organoid passaging and short-term forskolin

stimulation. To further demonstrate the potential of D-CryptO for drug testing, organoid

structures were analyzed following treatments with a panel of chemotherapeutic drugs. With

D-CryptO, subtle variations in how colon organoids responded to the different

chemotherapeutic drugs were detected, which suggest potentially distinct mechanisms of

action. This tool could be expanded to other organoid types, like intestinal organoids, to

facilitate 3D tissue morphological analysis.

Introduction

Monolayer cell-based assays are an invaluable tool for studying cellular functions in vitro.

However, these models do not accurately recapitulate in vivo tissue responses. This is largely

because monolayer cell models do not exhibit tissue-specific architecture and lack the

appropriate 3D cellular microenvironment. Stem cell-derived organoids that can

spontaneously differentiate and self-assemble into 3D tissues with structures that resemble

many features of the native organ have emerged as alternative in vitro models.[1] For instance,

colon organoids have been widely used as large intestine models due to their structural and

functional similarities. [2] An important feature of the colon epithelium is the crypt, which are

epithelial invaginations that renew the intestinal lining every 3-5 days. [3] The organization of

the crypt is crucial for the regeneration of the epithelium in vivo. Stem cells at the base of the

crypt are protected from continuous mechanical and chemical stressors, and as a result, can

proliferate and differentiate to regenerate the epithelium. Similarly, colon organoid

morphology reflects the structure and organization of the native colon crypts by exhibiting

budding structures which contain the stem cells that give rise to colon-specific cells. [4,5]

Therefore, analyzing organoid morphology can provide insights into colon physiology and

pathophysiology in vivo.

Qualitative analysis of colon organoid morphology, specifically the opacity and budding of

organoids, has largely been used to assess the maturity of colon organoids. Colon organoids

that are more transparent, have thinner walls, and are cystic are indicative of an earlier

differentiation state.[6] On the other hand, colon organoids have reached a more differentiated

state when they are more opaque due to the thickening of the epithelial wall. Differentiated

colon organoids also exhibit more budding structures that resemble the colon crypt which is

the stem cell niche that controls colonocyte renewal and homeostasis.[2] Previously, the

presence of budding within small intestinal organoids has been used to optimize the

extracellular matrix, study stem cell differentiation, and understand the mechanics of

epithelial folding. [3,7–9] Analysis of budding has also been used to study diseases. For

example, colon organoids from individuals with inflammatory bowel disease or tumour-

derived organoids had lower rates of budding structures.[10,11] However, to assess these

morphological differences, previous work used manual analysis or relied on traditional image

analysis that uses imperfect parameters such as eccentricity to describe organoid shapes.[12–

14]

To facilitate the morphological analysis of organoids with abstract features that are not easily

defined by traditional image analysis parameters, a type of computer vision called deep

learning can be applied. Deep learning refers to an automated method of computer-based

image recognition that relies on using pre-existing data to make predictions on new image

instances.[15] Traditional computer recognition techniques rely on manual feature extraction to

distinguish between the categories of interest. With deep neural networks, both feature

extraction and classification are done automatically without any input from the user. This

provides several advantages. First, colon organoid features are learned directly from the

images without the need for manual feature extraction. Second, analysis of the structures is

not limited to using shape descriptors, so organoid morphology can be characterized despite

the high heterogeneity of colon organoid structure. Third, automatic image analysis can

improve the throughput of morphological analysis. Finally, these models could be trained to

correctly classify between categories despite imaging artifacts. Artificial neural networks have

been previously used to detect and count intestinal organoids and replace immunostaining and

cell viability assays. [16–19] However, deep learning has yet to be used for the morphological

characterization of any type of organoids.

Hence, we used deep learning to characterize the morphological structure of organoids by

developing an analysis tool, D-CryptO, to distinguish between transparent and opaque

organoids, as well as spherical and budding organoids. Collectively, these features reveal the

structural maturity and health of colon organoids. To validate our deep learning model, we

analyzed changes in colon organoid morphologies in (1) organoid passaging, (2) short-term

forskolin stimulation, (3) a drug screening study with a panel of six chemotherapeutic drugs,

and (4) a dose-response study to doxorubicin. We found that morphological analysis allowed

us to capture variations in how colon organoids responded to the different chemotherapeutic

drugs, which provide insights into the potential mechanisms of drug toxicity.

Results

Colon organoid culture and morphological characteristics

Colorectal organoids, derived from primary colon tissue, were embedded in Matrigel, and

cultured for a period of 7 days in a 24-well plate (Figure 1a-b). The primary tissue contains

adult stem cells which proliferate and differentiate to form the colon organoids in vitro.

To determine organoid maturation, we performed histological analysis of the colon organoids.

Organoids expressed villin apically, a marker for microvilli, which is indicative of

differentiated intestinal cells. Furthermore, the expression of ki-67 indicated that stem cells

were also present within the colon organoids (Figure 1c). We observed a spectrum of

morphologies from these colon organoids. Organoids differed in their opacity as well as the

extent of budding (Figure 1d). Colonospheres are transparent with little-to-no budding. On

the other hand, colonoids are more mature organoids that are opaque with a significant

number of budding structures.[20,21] As the proliferating stem cells differentiate into organ-

specific cells, opacity increases due to changes in epithelium thickness. [7] We also observed

organoids that exhibited some characteristics of both colonospheres and colonoids. For

example, some organoids were spherical and opaque while other organoids had buds and were

transparent. Hence, using the parameters of both opacity and budding could give an indication

of the structural maturity of the organoids grown in vitro.

Figure 1. Morphological heterogeneity of colon organoids. a, Illustration of primary cells

embedded in Matrigel that self-assemble into colonospheres and develop into colonoids. b,

Organoids embedded in 50L of Matrigel in a standard 24-well plate. c, Histological

sections of organoids expressing the mature markers of villin, E-cadherin and ki-67. Scale

bar, 100 m. d, Representative images of organoids exhibiting varying levels of opacity and

budding. Scale bar, 200 m

Dataset creation and model training

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

1D-CryptO:Deeplearning-basedanalysisofcolonorganoidmorphologyfrombrightfieldimagesLyanAbdul1,JocelynXu2,AlexanderSotra1,AbbasChaudary3,JerryGao4,ShravanthiRajasekar3,NickyAnvari1,HamidrezaMahyar5,andBoyangZhang1,3*1SchoolofBiomedicalEngineering,McMasterUniversity,1280MainStreetWest,Hamilton,ON,L8S4L...

展开>> 收起<<

1 D-CryptO Deep learning -based analysis of colon organoid morphology from brightfield images.pdf

共29页,预览5页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

1 D-CryptO Deep learning -based analysis of colon organoid morphology from brightfield images

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: