QuantiFly: Robust Trainable Software for Automated Drosophila Egg Counting.
QuantiFly: Robust Trainable Software for Automated Drosophila Egg Counting.
复制标题
DOI：
10.1371/journal.pone.0127659
复制
发表时间：
2015
期刊：
影响因子：
3.7
通讯作者：
Piper MD
中科院分区：
文献类型：
--
作者：
Waithe D;Rennert P;Brostow G;Piper MD
我们报告的软件称为QuantiFly的开发和测试：一个自动化的工具，以量化果蝇产卵。许多实验室把果蝇卵作为健康的标志。现有的方法需要实验室研究人员在显微镜下手动计数鸡蛋。这种技术既耗时又繁琐，特别是当实验需要每天计数数百个小瓶时。QuantiFly软件的基础是一种算法，该算法应用并改进了现有的高级模式识别和机器学习例程。在本研究中，通过校正算法输出中观察到的偏倚，进一步提高了基线算法的准确度。QuantiFly软件，其中包括完善的算法，已被设计为通过直观和响应用户友好的图形界面立即访问科学家。该软件也是开源的、独立的、没有依赖性并且易于安装（https：//github.com/dwaithe/quantifly）。与通过数字图像进行的人工鸡蛋计数相比，QuantiFly在透明（定义）和不透明（基于酵母）的苍蝇培养基上产卵的平均准确率分别为94%和85%。因此，该软件能够在大多数实验情况下检测实验差异。值得注意的是，该软件的高级特征识别功能被证明对气泡和裂缝等食品表面伪影具有鲁棒性。用户体验涉及图像采集、通过在一些小瓶的图像中标记蛋的子集来进行算法训练，然后是批量分析模式，其中自动评估新图像的蛋数量。初始训练通常需要大约10分钟，而软件的后续图像评估仅需几秒钟。鉴于手动计数每个小瓶的平均时间约为40秒，我们的软件为从20个小瓶开始的实验带来了省时优势。我们还描述了一个可选的丙烯酸盒，用作数码相机支架，并在图像采集期间提供受控照明，这将保证本研究中使用的条件。
We report the development and testing of software called QuantiFly: an automated tool to quantify Drosophila egg laying. Many laboratories count Drosophila eggs as a marker of fitness. The existing method requires laboratory researchers to count eggs manually while looking down a microscope. This technique is both time-consuming and tedious, especially when experiments require daily counts of hundreds of vials. The basis of the QuantiFly software is an algorithm which applies and improves upon an existing advanced pattern recognition and machine-learning routine. The accuracy of the baseline algorithm is additionally increased in this study through correction of bias observed in the algorithm output. The QuantiFly software, which includes the refined algorithm, has been designed to be immediately accessible to scientists through an intuitive and responsive user-friendly graphical interface. The software is also open-source, self-contained, has no dependencies and is easily installed (https://github.com/dwaithe/quantifly). Compared to manual egg counts made from digital images, QuantiFly achieved average accuracies of 94% and 85% for eggs laid on transparent (defined) and opaque (yeast-based) fly media. Thus, the software is capable of detecting experimental differences in most experimental situations. Significantly, the advanced feature recognition capabilities of the software proved to be robust to food surface artefacts like bubbles and crevices. The user experience involves image acquisition, algorithm training by labelling a subset of eggs in images of some of the vials, followed by a batch analysis mode in which new images are automatically assessed for egg numbers. Initial training typically requires approximately 10 minutes, while subsequent image evaluation by the software is performed in just a few seconds. Given the average time per vial for manual counting is approximately 40 seconds, our software introduces a timesaving advantage for experiments starting with as few as 20 vials. We also describe an optional acrylic box to be used as a digital camera mount and to provide controlled lighting during image acquisition which will guarantee the conditions used in this study.
登录
查看更多内容
影响因子：
48
作者：
Eliceiri, Kevin W.;Berthold, Michael R.;Goldberg, Ilya G.;Ibanez, Luis;Manjunath, B. S.;Martone, Maryann E.;Murphy, Robert F.;Peng, Hanchuan;Plant, Anne L.;Roysam, Badrinath;Stuurmann, Nico;Swedlow, Jason R.;Tomancak, Pavel;Carpenter, Anne E.
通讯作者：
Carpenter, Anne E.
影响因子：
10.6
作者：
Huh, Seungil;Ker, Dai Fei Elmer;Kanade, Takeo
通讯作者：
Kanade, Takeo
影响因子：
7.5
作者：
Geurts, P;Ernst, D;Wehenkel, L
通讯作者：
Wehenkel, L
影响因子：
2.5
作者：
SIMONCELLI, EP;FREEMAN, WT;HEEGER, DJ
通讯作者：
HEEGER, DJ
DOI：
10.1007/3-540-45014-9_1
发表时间：
2000-01-01
期刊：
MULTIPLE CLASSIFIER SYSTEMS
影响因子：
--
作者：
Dietterich, TG
通讯作者：
Dietterich, TG