A novel chromosome cluster types identification method using ResNeXt WSL model

Lin, Chengchuang; Zhao, Gansen; Yin, Aihua; Yang, Zhirong; Guo, Li; Chen, Hanbiao; Zhao, Lei; Li, Shuangyin; Luo, Haoyu; Ma, Zhaohui

doi:10.1016/j.media.2020.101943

Cited by 18 publications

(28 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“… • We prove the efficiency of our contributions through the public chromosome cluster types dataset, ChrCluster. We also achieve 94.99% accuracy, which is higher than the 94.09% accuracy proposed by Lin et al (2021) . …”

Section: Introductionmentioning

confidence: 56%

“…In 2021, Lin et al (2021) proposed the chromosome cluster type identification task. In this work, 6,592 chromosome clusters were obtained from the hospital, and they created and made available the first chromosome cluster dataset (ChrCluster for simplicity).…”

Section: Introductionmentioning

confidence: 99%

“…Finally, they propose a classification model as the benchmark of the ChrCluster dataset. To avoid overfitting on the small-scale ChrCluster dataset, Lin et al (2021) takes Instagram weakly supervised learning pretrained weights [ Mahajan et al (2018) ] and the customized ResNeXt [ Xie et al (2017) ] classification model to achieve an accuracy of 94.09%.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

SupCAM: Chromosome cluster types identification using supervised contrastive learning with category-variant augmentation and self-margin loss

Luo

Zhao

2023

Front. Genet.

View full text Add to dashboard Cite

Chromosome segmentation is a crucial analyzing task in karyotyping, a technique used in experiments to discover chromosomal abnormalities. Chromosomes often touch and occlude with each other in images, forming various chromosome clusters. The majority of chromosome segmentation methods only work on a single type of chromosome cluster. Therefore, the pre-task of chromosome segmentation, the identification of chromosome cluster types, requires more focus. Unfortunately, the previous method used for this task is limited by the small-scale chromosome cluster dataset, ChrCluster, and needs the help of large-scale natural image datasets, such as ImageNet. We realized that semantic differences between chromosomes and natural objects should not be ignored, and thus developed a novel two-step method called SupCAM, which could avoid overfitting only using ChrCluster and achieve a better performance. In the first step, we pre-trained the backbone network on ChrCluster following the supervised contrastive learning framework. We introduced two improvements to the model. One is called the category-variant image composition method, which augments samples by synthesizing valid images and proper labels. The other introduces angular margin into large-scale instance contrastive loss, namely self-margin loss, to increase the intraclass consistency and decrease interclass similarity. In the second step, we fine-tuned the network and obtained the final classification model. We validated the effectiveness of modules through massive ablation studies. Finally, SupCAM achieved an accuracy of 94.99% with the ChrCluster dataset, which outperformed the method used previously for this task. In summary, SupCAM significantly supports the chromosome cluster type identification task to achieve better automatic chromosome segmentation.

show abstract

Section: Introductionmentioning

confidence: 56%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

SupCAM: Chromosome cluster types identification using supervised contrastive learning with category-variant augmentation and self-margin loss

Luo

Zhao

2023

Front. Genet.

View full text Add to dashboard Cite

show abstract

“…For example, by using a single convolutional neural network (CNN) algorithm, Esteva et al (2017) demonstrated the classification of skin lesions in performance on par with all tested experts. Lin et al (2020) developed a ResNeXt WSL model that achieved impressive performance (94.09% accuracy, 92.79% sensitivity, and 98.03% specificity) in making chromosome cluster type identification. Actually, simply based on microscopic images, AI algorithms were quite competent in analyzing most, if not all, biological events, such as the early onset of pluripotent stem cell differentiation ( Waisman et al, 2019 ), tumor cell malignancy ( Oei et al, 2019 ), mitosis staging ( Mao et al, 2019 ), and the like.…”

Section: Introductionmentioning

confidence: 99%

AIM-CICs: an automatic identification method for cell-in-cell structures based on convolutional neural network

Tang

Zhao

et al. 2022

Journal of Molecular Cell Biology

View full text Add to dashboard Cite

Whereas biochemical markers are available for most types of cell death, current studies on non-autonomous cell death by entosis relays strictly on the identification of cell-in-cell structures (CICs), a unique morphological readout that can only be quantified manually at present. Moreover, the manual CIC quantification is generally over-simplified as CIC counts, which represents a major hurdle against profound mechanistic investigations. In this study, we take advantage of artificial intelligence (AI) technology to develop an automatic identification method for CICs (AIM-CICs), which performs comprehensive CIC analysis in an automated and efficient way. The AIM-CICs, developed on the algorithm of convolutional neural network (CNN), can not only differentiate between CICs and non-CICs (the area under the receiver operating characteristic curve (AUC) > 0.99), but also accurately categorize CICs into five subclasses based on CIC stages and cell number involved (AUC > 0.97 for all subclasses). The application of AIM-CICs would systemically fuel research on CIC-mediated cell death such as high-throughput screening.

show abstract

“…The network was tested on three publicly available data sets and provided state of the art results. Lin et al (Lin et al, 2021), analyze the difficult problem of Chromosome karyotyping, i.e. the identification and analysis of individual chromosomes.…”

mentioning

confidence: 99%