Talk to SalesSign Up

BrainScanology Optimizes Medical AI Models with Datature

Keechin
August 22, 2022
About

BrainScanology

Revolutionizing shape analysis and image-based diagnostics

Industry

Biopharma, Biotechnology, Health Care, Medical, Software

Headquarters

San Francisco Bay Area, West Coast, Western US

BrainScanology’s Challenge

Brainscanology uses proprietary technology to analyze shapes in medical images and improve image-based diagnostics. Before the shape analysis can take place, however, they first have to locate the region of interest and its boundaries in a medical image. This process is referred to as medical image segmentation. It often requires the annotator to possess significant anatomical and/or medical knowledge. Manual segmentation is also an extremely time-consuming task, even for the most experienced radiologists.

Datature’s Solution

BrainScanology leverages Datature’s platform to annotate medical images and build deep learning segmentation models to automatically segment regions of interest in the images, such as the corpus callosum in a brain MRI scan and a lesion in an X-ray image. 

Using Datature’s Annotator, BrainScanology is able to quickly check and make adjustments to their existing labels, as well as creating new ones. Datature’s Annotator also allows their entire team to make and verify changes collaboratively using different user accounts, drastically speeding up the annotation and QA processes.

Brainscanology leverages Datature's Annotator to automatically segment regions of interest in the images, such as the corpus callosum in a brain MRI scan

Additionally, BrainScanology uses Datature’s Deep Learning Workflow to intricately design and implement data augmentation strategies that can introduce synthetic variations to their deep-learning model. These variations, such as different image orientations and contrasts, can help improve their model's robustness against variations and overfitting during the neural network training phase.

Different image orientations and contrasts can help improve their model's robustness against variations and overfitting during the neural network training phase

Finally, the team at BrainScanology leverages the evaluation preview function under Advanced Evaluation to visually observe how well the model performs at various checkpoints throughout the model training process. Evaluation Previews can provide a lot more valuable information than a typical checkpoint system where only quantitative summaries, such as accuracy and loss, are provided.  

With our evaluation previews, BrainScanology is able to examine 1) whether or not the model is accurately identifying the object of interest, 2) how well the model is segmenting the object of interest, and 3) if the model is improving over time, in real time.

Datature’s evaluation functions do not only help users evaluate the performance of their model, they help users spot potential problems with the input data, augmentation, and model architecture selection, that are not reflected in the quantitative summaries.

Brainscanology leverages the evaluation preview function under Advanced Evaluations to visually observe how well the model performs at various checkpoints

Leveraging Evaluation Previews, the team can also identify images and phenomenons that are hard for the model to converge on. This is where the team can decide on which images/classes to add into the training to improve the model in subsequent trainings.

“I was immediately amazed by how rapid and easy to use Datature’s platform is for segmenting complex shapes in medical images. Its high level of accuracy on the first run, before any additional refinements, blew me away.” - David Nguyen, PhD | Co-Founder & CEO at BrainScanology

Datature is proud to support Brainscanology’s mission to shape the future of diagnostics and we are excited to see what’s next for BrainScanology.

If you are an innovator looking to unlock the potential of deep learning to transform the medical imaging industry, get in touch with us!

About BrainScanology

Co-founded in 2020 by computational biologist David Nguyen, PhD, BrainScanology is a HealthTech Startup in San Francisco that aims to disrupt the healthcare industry with its diagnostic software platform. 

The platform supports the analysis of complex shapes  - not just areas and volumes - in medical images, such as MRI/CT/X-ray/ultrasound scans, using a proprietary algorithm invented by none other than Dr. Nguyen himself.

The algorithm, known as the LCPC Transform, has successfully been applied to detect a wide range of clinical conditions, such as ADHD, Alzheimer’s, and Hill-Sachs lesions. It has also been used in biology research to measure the shapes of sickle cells, cells in 2D cell culture, and 3D Organoids & Microvilli. Their latest use case measures vertebral modic shapes which will revolutionize our understanding of back pain and how to better treat it.

In addition to medicine and research, the technology also has the capability to revolutionize many other industries. While the team is currently focusing on several neurological and psychiatric conditions for now, the potential to apply this technology in solving problems in other industries is limitless.

About Datature

Datature is an end-to-end MLOps platform that allows teams and enterprises to build computer vision models without a single line of code. Teams can manage datasets, annotate, generate synthetic data, train and deploy - all in a single, secure cloud-based platform. With the rise of citizen data scientists, deep tech companies, and enterprises looking to adopt deep-learning - Datature equips these startups / experts with the tools required to build their own capabilities easily within weeks.

Build Your AI Vision Applications for Free

Develop your go-to-market product with our no-code MLOps platform that simplifies how computer vision models are built.

Up to 500 images with model-assisted annotations
Up to 300 minutes of GPU training
Slack and email support from our engineers and product managers