MacroInsight Builds Clinical Decision Support Systems with Datature

Datature is proud to support MacroInsight's mission of revolutionizing healthcare by developing cutting-edge clinical decision support systems!

Hoki Fung

About MacroInsight

Revolutionizing healthcare through the development of cutting-edge clinical decision support systems


Healthcare, Technology, Clinical, Precision Medicine, Smart Manufacturing


Taipei, Taiwan

MacroInsight’s Challenges and Datature’s Solutions

MacroInsight is dedicated to developing and manufacturing revolutionary AI-based clinical decision support systems (CDSS) that are unmatched in the industry. MacroInsight’s AI-based CDSS can assist healthcare professionals by providing real-time analysis and interpretation of patient data, including medical images, lab results, and electronic health records (EHRs), enhancing healthcare professionals' decision-making capabilities and improving patient outcomes.

However, deploying intelligent systems in real-world clinical settings presents a range of significant challenges. Among the most pressing of these are concerns related to 1) data privacy and security, 2) integration with existing medical systems, such as EHR Systems and Picture Archiving and Communication Systems (PACS), and 3) updates and maintenance of the underlying AI models.

In this user spotlight, we present the various challenges encountered by MacroInsight and the solutions that have been developed by Datature and MacroInsight to address them.

Privacy and Security

Ensuring privacy and security when handling clinical data is of paramount importance. Clinical data such as EHRs and Digital Imaging and Communications in Medicine (DICOM) files from diagnostic imaging scans contain sensitive patient information, and these data must be protected from unauthorized access or breaches. This means MacroInsight has to be extremely cautious when designing and implementing its data pipelines, as well as selecting the technology and services it uses to build its AI capabilities.

Datature’s end-to-end MLOps platform is designed with security as a top priority. It is fully compliant with the Health Insurance Portability and Accountability Act (HIPAA) and incorporates a range of technical and organizational measures to safeguard users' data. These measures include the implementation of data encryption, secure data storage, and robust access controls. Datature’s platform adheres to strict policies and procedures for data handling, and regularly undergoes audits to ensure compliance with industry regulations and standards.

Tailored Solutions: Custom Scripts for Specialized Image Preprocessing

MacroInsight’s CDSS utilizes medical images from various diagnostic imaging exams such as X-Ray, Ultrasound Imaging, Magnetic Resonance Imaging (MRI), Computed Tomography (CT), and Positron Emission Tomography (PET), to deliver diagnostic assessments and prognostic insights to healthcare professionals.

MacroInsight leverage Datature’s Annotator to identify medical images from various diagnostic imaging exams such as X-Ray, Ultrasound Imaging, Magnetic Resonance Imaging (MRI), Computed Tomography (CT), and Positron Emission Tomography (PET).

Incorporating multimodal medical images into AI models can pose several challenges. Firstly, different imaging modalities, especially those obtained from different scanners, have varying image characteristics, such as resolution, contrast, and noise, which can lead to increased heterogeneity in the data. While models trained on heterogeneous data can sometimes reduce the risk of overfitting to specific input patterns hence generalizing better to unseen data during inference predictions, increased heterogeneity in the data can negatively impact the accuracies of the machine learning models if not properly addressed.

MacroInsight leverages on Datature’s domain experts, engineers, and developers to develop specialized scripts for an AI model that were able to recognize different features on different series on multimodal medical images.

Datature’s team of experienced domain experts, engineers, and developers worked closely with MacroInsight to develop specialized scripts aimed at addressing the technical difficulties associated with incorporating heterogeneous, multimodal medical images into an AI model. The scripts streamlined the data integration process, automated the preprocessing of medical images, and ensured that the processed images could be seamlessly integrated into Datature’s end-to-end pipeline for model construction, training, and deployment.

Working with our domain experts and engineers during the early stages of development brought multiple advantages to MacroInsight’s implementation of AI in their clinical decision support systems. The scripts developed by our experts not only incorporated MacroInsight’s multimodal medical images into the AI model effectively, but were also optimized for the deployment and inference prediction processes. Our experts considered crucial factors, such as the user interface and pipeline of MacroInsight's systems, during the development of the scripts to ensure a smooth model deployment and system integration experience.

Tailored Solutions: Advanced Training Options

MacroInsight’s intelligent systems are highly advanced, thus necessitating specialized training methods. One such technique adopted by MacroInsight is Active Learning.

Active learning is a model training technique in AI that allows a model to actively select the data it uses for training, rather than relying on randomly selected or pre-labeled data. The goal of active learning is to focus on the most informative and relevant examples, in order to improve the performance of the model. This is achieved by allowing the model to identify and request labels for specific data points, rather than using a fixed dataset. Active learning can be used to reduce the amount of labeled data required to train a model, and can improve the model's performance on specific tasks.

MacroInsight taps on Active Learning to improve the model’s performance on specific tasks such as showing what’s the inferred tumor type.

Datature’s team of experienced Machine Learning engineers worked closely with MacroInsight to understand the objectives and requirements, and subsequently, designed and implemented custom solutions through the development of specialized scripts and custom deep learning architectures that cater to MacroInsight’s needs.

In the case of active learning, Datature’s implementation incorporates advanced ML techniques such as uncertainty sampling, ensemble methods, and self-supervised learning. Uncertainty sampling is a method where the model can query for labels on the most uncertain examples, in order to reduce the amount of human annotation needed. Ensemble methods involve training multiple models on different subsets of the data, and the final prediction is made by combining the predictions of these models. Self-supervised learning is where the model can learn from the data itself by generating its own labels and adjusting its parameters accordingly using techniques like autoencoder, contrastive learning, and clustering methods.

Model Maintenance

MacroInsight’s intelligence systems can greatly enhance decision making in clinical settings, but as new data arrives and medical knowledge evolves, it is important to ensure that the underlying models powering the systems remain up-to-date and accurate. Regular updating of AI models allows for the incorporation of new data, advances in medical knowledge, and new techniques in model training.

However, this can present a number of challenges. One of the main challenges is ensuring that the updated model performs as well or better than the previous version, effectively mitigating model drift. Model drift refers to the degradation of a model's performance over time or when applied to varying environments or data sets. This degradation can occur due to a plethora of reasons such as variations in new data distributions, changes in data formatting, or alteration of the model's parameters. Model drift can pose significant challenges in real-world applications, as it can lead to poor performance and inaccurate predictions.

In MacroInsight’s case, addressing model drifts caused by new data acquired from 1) clinical scanners with varying contrast factors, 2) new patients, and 3) different scanning protocols can be a challenging task. In MRI, a scanning protocol refers to the specific sequence of steps and parameters used to acquire images from the patient. These protocols include settings such as the type of pulse sequence used, the field of view, the resolution, the slice thickness, and the contrast agents used.

To mitigate this, it is imperative to conduct thorough testing and validation procedures to guarantee the model's performance on new data. Datature’s comprehensive feature set is designed to ensure the updated model is robust to variations introduced by the aforementioned factors. Datature provides MacroInsight with a streamlined process for maintaining and updating their models. MacroInsight has the ability to easily incorporate new data into an existing project, retrain the model using the new data, experiment with different models, evaluate and compare model performances, and smoothly deploy updated models. Additionally, all models are saved, allowing them to make informed decisions when selecting which model to deploy in production.

Tailored Solutions: Integration and Custom APIs

MacroInsight’s systems are designed with the specific needs of the users in mind. MacroInsight understands that healthcare professionals’ time is valuable and that every moment spent navigating and adapting to new technology is a moment taken away from their patients. This means ensuring the process of using MacroInsight’s intelligence systems is as user-friendly and effortless as possible, allowing healthcare professionals to utilize the full capabilities of its systems with minimal interruption to their workflow.

Healthcare professionals are able to automatedly calculate tumor geometry information and utilize the full capabilities of its systems due to Datature’s streamlined process for maintaining and updating MacroInsight’s models.

To address this, Datature worked with MacroInsight to design a secure and seamless workflow and develop the APIs required to streamline the data upload, labeling, and result viewing processes for healthcare professionals.

Datature has simplified our AI model development, specifically our vision model training and deployment. Their MLOps solution has accelerated the progress of smart medical care and we're proud to use their platform, delivering better results with half the effort. - Mingta Tu | Co-Founder at MacroInsight

Datature is proud to support MacroInsight’s mission to revolutionize healthcare with their AI-enabled clinical decision support systems. We take great pride in our role as the driving force behind the AI capabilities of MacroInsight’s intelligent systems, and are excited to see what’s next for MacroInsight!

If you are an innovator looking to unlock the potential of artificial intelligence to transform the healthcare industry, get in touch with us!

About MacroInsight

MacroInsight is an up-and-coming MedTech startup in Taiwan that is making waves in the healthcare industry with its state-of-the-art clinical decision support systems.

The company is dedicated to developing and manufacturing intelligent systems that are unmatched in the industry, enabling healthcare professionals to make diagnoses with confidence and efficiency.

Its flagship product, MacroBrain, leverages the latest AI technology to analyze multimodal, diagnostic images of the brain. By integrating various imaging modalities and knowledge-based algorithms in clinical MRI scans, MacroBrain delivers a thorough assessment that includes brain cancer diagnosis and prognosis.

MacroInsight collaborates closely with leading medical centers in Taiwan to research and develop these advanced clinical decision support systems, ensuring seamless integration and user-friendly interface for healthcare professionals. The company also offers customizable model training services to meet the specific needs of individual hospitals and healthcare organizations.

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 models with the best tools.

develop ml models in minutes with datature