LabSwipe is an innovative tool to create custom algorithms for your EggSorter independently and efficiently. It facilitates the labeling and classification of biological images, creating AI algorithms tailored to specific needs. By addressing the variability in manual selection criteria that can affect experimental outcomes, LabSwipe standardizes processes and enhances result reproducibility.
Sorting criteria for your samples may vary depending on each specific project.
Achieving consistent and standardized sorting by hand can be challenging and influenced by various factors, which can affect the reproducibility of your results.
LabSwipe addresses this issue by allowing in a user-friendly way to define sorting criteria, label data, and train algorithms directly on your own devices.
Sorting criteria for your samples may vary depending on each specific project. Achieving consistent and standardized sorting by hand can be challenging and influenced by various factors, which can affect the reproducibility of your results.
LabSwipe addresses this issue by allowing in a user-friendly way to define sorting criteria, label data, and train algorithms directly on your own devices.
Enhance Your Reproducibility
Standard Criteria & AI
Custom image classification algorithms. From recognising different developmental stages of the embryos to the adaptation to new species.
User-Friendly Experience
No prior technical experience in AI is needed to quickly and efficiently label data and train models, speeding up the workflow.
Data Ownership
The intellectual property of the images and data remains with you, as all processing is done locally, ensuring data privacy and control.
Two key functionalities: Labeling and Training
Labeling
The labeling process starts by uploading your images, and defining custom classes to suit your specific project needs.
Within the app you can find various labeling methods, including the unique swipe feature which makes the process intuitive and fun.
LabSwipe will generate at the end of the labelling exportable files to save the progress and review the labeled data efficiently to ensure accuracy.
Training
The training process is straightforward and requires no technical expertise.
LabSwipe uses a Deep Learning model tailored for image classification to train algorithms effectively even with small datasets.
The platform provides clear performance metrics, enabling users to assess the model’s performance on their specific data.
Discover the many applications of LabSwipe
Use Case I: Zebrafish embryos at 6 hpf
Many zebrafish facilities worldwide require the generation of wildtype fish lines. One key step in this process is the surface disinfection of zebrafish embryos. It is common that this disinfection is done when the embryos reach the ‘shield stage’, approximately at 5 to 6 hours post-fertilization (hpf).
A LabSwipe algorithm was developed to identify zebrafish embryos at this specific developmental stage, enabling the EggSorter to automatically select and plate only the appropriate embryos for disinfection, while discarding the rest.
This integration of LabSwipe and the EggSorter can standardise the sorting process with over 95% accuracy.
Zebrafish embryo at the ‘shield stage’ (approx. 6 hpf), classified as good.
Zebrafish embryo at a stage different from the ‘shield stage’, classified as bad.
Use Case I: Zebrafish embryos at 6 hpf
A LabSwipe algorithm was developed to identify zebrafish embryos at 6 hpf, enabling the EggSorter to automatically select and plate only the appropriate embryos for disinfection, while discarding the rest.
This integration of LabSwipe and the EggSorter can standardise the sorting process with over 95% accuracy.
Zebrafish embryo at 6 hpf, classified as good.
Zebrafish embryo at a stage different from 6 hpf, classified as bad.
Use Case II: Killifish embryos
One key developmental stage in killifish growth is the so-called ‘black eyes’ stage, which occurs 7 to 10 days after fertilization. At this stage, it is common practice in killifish husbandry to select and transfer the ‘black eyes’ embryos to the next growing medium.
LabSwipe has been used to create a custom algorithm that will differentiate killifish embryos at the ‘black eye’ stage from the rest.
The algorithm’s performance after the training in this case was 93% accuracy.
Killifish embryo at the ‘black eyes stage’, classified as good.
Killifish embryo at a stage different from the ‘black eyes stage’, classified as bad.
Use Case II: Killifish embryos
LabSwipe has been used to create a custom algorithm that will differentiate killifish embryos at the ‘black eye’ stage from the rest.
The algorithm’s performance after the training in this case was 93% accuracy.
Killifish embryo at the ‘black eyes’ stage, classified as good.
Killifish embryo at a stage different from the ‘black eyes’, classified as bad.
Use Case III: Sea bream embryos
LabSwipe has been used to adapt the EggSorter to work with new species, such as sea bream embryos, a widely used species in aquaculture.
To grow sea bream embryos more efficiently, it is important to ensure that all embryos in the growing tank are at the same developmental stage. For this reason, a LabSwipe algorithm was developed to distinguish between embryos in early developmental stages, dead embryos, and those in later stages.
The algorithm’s accuracy, after the training, to detect alive and young embryos from the rest in this case was 97%.
Sea bream embryo at an early stage of development, classified as good.
Dead sea bream embryo at a later stage of development, classified as bad.
Use Case III: Sea bream embryos
LabSwipe has been used to adapt the EggSorter to work with new species, such as sea bream embryos, a widely used species in aquaculture.
The algorithm’s accuracy, after the training, to detect alive and young embryos from the rest in this case was 97%.
Sea bream embryo at an early stage of development, classified as good.
Dead sea bream embryo at a later stage of development, classified as bad.
Use Case IV: Flower seeds
By using LabSwipe, an algorithm was created to differentiate germinated flower seeds (expressing the fluorescent marker GFP) from not germinated seeds (no fluorescence expression).
The algorithm’s performance after the training was 93% accuracy.
Germinated flower seed. Left: in fluorescence (GFP expression). Right: in brightfield.
Non-germinated flower seed. Left: in fluorescence (no GFP expression). Right: in brightfield.
Use Case IV: Flower seeds
By using LabSwipe, an algorithm was created to differentiate germinated flower seeds (expressing the fluorescent marker GFP) from not germinated seeds (no fluorescence expression).
The algorithm’s performance after the training was 93% accuracy.
Germinated flower seed in fluorescence (GFP expression)
Non-germinated flower seed in fluorescence (no GFP expression).
Customise your screening to fit all your research projects.
Do not hesitate, and contact us to automate your sorting.
Do you want to try it?
Do you want to try it?
Customise your screening to fit all your research projects.
Do not hesitate, and contact us to automate your sorting.