Digital whole-slide image analysis for automated diatom test in forensic cases of drowning using a convolutional neural network algorithm

Zhou, Yuanyuan; Ji, Zhang; Huang, Jiao; Deng, Kaifei; Zhang, Jianhua; Qin, Zhiqiang; Wang, Zhenyuan; Zhang, Xiaofeng; Tuo, Ya; Chen, Liqin; Chen, Yijiu; Huang, Ping

doi:10.1016/j.forsciint.2019.109922

Cited by 52 publications

(28 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(A) Compared to all experts, the AI system revealed greater efficiency. (B) In the whole process of diatom testing, the AI system could replace the work of manually observing and counting diatoms [33]. Reprinted with permission.…”

Section: Resultsmentioning

confidence: 99%

“…Considering its achievement in image classification, we proved that deep learning, such as CNNs, could realize automatic diatom testing [33]. Specifically, digestive tissue smears, which are similar to traditional histological slides, could also be digitized with a slide scanner ( Figure 3A).…”

Section: The Potential Of Deep Learning In Diatom Testingmentioning

confidence: 99%

“… The fragments of diatom under 400 × light microscopy. Excessive digestion will cause the structure of most diatoms to be destroyed, resulting in an amount of diatom fragments [ 33 ]. Reprinted with permission.…”

Section: Traditional Chemical Digestion For Diatom Detectionmentioning

confidence: 99%

“…Database was divided into training, validation and testing sets. The results were labeled into a pseudo-colour map, showing the site of diatoms[33]. Reprinted with permission.…”

mentioning

confidence: 99%

See 3 more Smart Citations

Research advances in forensic diatom testing

Zhou

Cao

Huang

et al. 2020

Forensic Sciences Research

Self Cite

View full text Add to dashboard Cite

In forensic practice, it is difficult to determine whether a dead body in the water resulted from drowning or from disposal after death. Diatom testing is currently an important supporting technique for the determination of death by drowning and of drowning sites, even though it is a time-consuming and laborious task. This article reviews the development of diatom testing over the decades and discusses a new method for the potential application of deep learning in diatom testing.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: The Potential Of Deep Learning In Diatom Testingmentioning

confidence: 99%

Section: Traditional Chemical Digestion For Diatom Detectionmentioning

confidence: 99%

“…Database was divided into training, validation and testing sets. The results were labeled into a pseudo-colour map, showing the site of diatoms[33]. Reprinted with permission.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Research advances in forensic diatom testing

Zhou

Cao

Huang

et al. 2020

Forensic Sciences Research

Self Cite

View full text Add to dashboard Cite

show abstract

“…Due to the broadening availability of high throughput, in part, in situ, imaging platforms 15 – 18 , and large publicly available image sets 19 , marine plankton has probably been addressed most commonly in such attempts in the aquatic realm 20 – 23 . The attention, however, recently also broadened to fossil foraminifera 24 , radiolarians 25 , as well as diatoms 26 , 27 . Due to the availability of deep learning software libraries 28 , 29 , well performing network architectures pre-trained on massive data sets like ImageNet 30 , and experiences accumulating related to transfer learning, i.e., application of pre-trained networks upon smaller data sets from a specialized image domain, the utilization of deep CNNs for a particular labelled image library is now within reach of taxonomic specialists of individual organismic groups.…”

Section: Introductionmentioning

confidence: 99%

Deep learning-based diatom taxonomy on virtual slides

Kloster

Langenkämper

Zurowietz

et al. 2020

Sci Rep

View full text Add to dashboard Cite

Deep convolutional neural networks are emerging as the state of the art method for supervised classification of images also in the context of taxonomic identification. Different morphologies and imaging technologies applied across organismal groups lead to highly specific image domains, which need customization of deep learning solutions. Here we provide an example using deep convolutional neural networks (CNNs) for taxonomic identification of the morphologically diverse microalgal group of diatoms. Using a combination of high-resolution slide scanning microscopy, web-based collaborative image annotation and diatom-tailored image analysis, we assembled a diatom image database from two Southern Ocean expeditions. We use these data to investigate the effect of CNN architecture, background masking, data set size and possible concept drift upon image classification performance. Surprisingly, VGG16, a relatively old network architecture, showed the best performance and generalizing ability on our images. Different from a previous study, we found that background masking slightly improved performance. In general, training only a classifier on top of convolutional layers pre-trained on extensive, but not domain-specific image data showed surprisingly high performance (F1 scores around 97%) with already relatively few (100-300) examples per class, indicating that domain adaptation to a novel taxonomic group can be feasible with a limited investment of effort. Diatoms are microscopic algae possessing silicate shells called frustules 1. They inhabit marine and freshwater environments as well as terrestrial habitats. Taxonomic composition of their assemblages is routinely assessed using light microscopy in ecological, bioindication and paleoclimate research 2-4. Silicate frustules cleaned of organic material and embedded into high refractive index mountant on cover slips represent the most widely used type of microscopic preparations for such analyses 5,6. Attempts to computerize parts or the whole of this workflow have been made repeatedly, starting with Cairns 7 , and in the most complete manner so far by the ADIAC project 8 , motivated by the desire to speed up the taxonomic enumeration process, to reduce its dependence on highly trained taxonomic experts, and to make identification results more reproducible and transparent. Recently, we described an updated re-implementation of most parts of this workflow, covering high throughput microscopic imaging, segmentation and outline shape feature extraction of diatom specimens 9,10 which we since mainly applied in morphometric investigations 11-14. The missing component in this workflow has been automated or computer-assisted taxonomic identification. For this purpose, i.e. image classification in a taxonomic context, deep convolutional neural networks (CNNs) are currently becoming the state-of-the-art technique. Due to the broadening availability of high throughput, in part, in situ, imaging platforms 15-18 , and large publicly available image sets 19 , marine plankton has pro...

show abstract