Enhancing an Innovative Tumor Detector with Scalable ML Infrastructure

While the latest advancements in AI/ML-based HealthTech promise a wide array of benefits to radiology specialists, in reality the integration process is typically too complicated, time-consuming, and resource-intensive. As a result, many specialists end up not utilizing AI-powered software at all and still rely on less effective assessment techniques including manual.Our client’s goal is to shift the perception of AI-based screening tools and deliver a cutting-edge digital product that would enhance the workflow of a modern radiologist rather than disrupt it. That is why the process of the software development vendor selection was thorough and vigorous. It was absolutely necessary for an IT provider to not only be experienced in the development of modern radiology imaging software but also well-versed in the complexities of the medical workflow.The final product should combine security, full compliance, and accurate analysis with ease of use, intuitive interface, and smooth PACS integration.

Since day one of cooperation on this project, Light IT Global’s goals have been to take all the necessary technical measures to optimize the software’s performance, ensure easy scalability, and modularity through clean architecture and flexible infrastructure. To achieve that, our experts offered and implemented the following procedures:

Careful technical execution that facilitates the smooth operation of the solution’s front-end and back-end was paired with modern intuitive UI/UX and thorough testing. The use of reliable libraries like Material-UI allowed to create a responsive layout and clean UI. Both integration and unit tests (Jest and React Testing) were performed to secure the quality of the final product and the stability/security of all the components.

A custom environment provisioning system created by our experts utilizes a role-based access model, with user roles being user, environment administrator, and reviewer. Each person accessing the environment has a unique ID, username, and email address, which strengthens accountability, simplifies user management, provides operational clarity, and boosts data protection.Each isolated environment is a customised and extended version of the OHIF viewer where users can easily manage studies (upload, view, inspect, delete). File validation mechanisms on the server side ensure that only properly formatted files with correct DICOM tags are accepted. This prevents incorrect display and poor analysis results.The user also gets easy access to a set of tools for CT scans analysis as well as one-button study analysis via the ML script. Once the ML script finishes the analysis, two additional modalities are added to the study:1. Secondary capture — a visual representation of the analysis results
2. Structured Report — a textual summary reportThe report contains a number of actionable insights about all findings (like location, size, characteristics), general information, and comparison with the previous studies.All findings detected by the script can be digitally assessed to safeguard accuracy and proper classification. As the whole process is quick and automated while still providing instant access to all the necessary medical information, radiologists receive a valuable web-based tool that saves time, provides convenience, and enhances the diagnostic process, ensuring nothing is missed.Light IT Global’s ML-focused contributions to the product significantly improved model performance and usability, resulting in a solution ready for real-world clinical environments.

The unique medical device software offered by our client has the potential to revolutionize the way AI/ML-based solutions are used in radiology, in general, and in cancer prevention, in particular.The software is MDR-approved, ensures 96,6% lesion detection sensitivity, and provides a number of benefits to both radiologists and healthcare facilities. The platform provides automatic test fetching and comparison, is fully integrable in the PACS, and offers fully isolated environments for each facility, which combines ease of use and data security. The system’s ability to bring real-life value without any disruption to the existing medical workflows (which may vary across facilities) boosts adoption and increases the commercial potential of the product.The solution is patented and certified. It has already been successfully showcased on multiple radiology-specific congresses across the globe and is currently present in +10 sites across Europe.