Project / product name: PathoX
Link to the project: https://github.com/MaksymPetyak/pathox-ehh23
Team leader: David Koleckar
Challenge: 1. Icebreaker AZ23
Problem: Working with raw pathological data of Czech patients is tedious work due to missing standard formats and processes across laboratories. Aligning such data takes considerable effort even for trained pathologists, leading to demanding manual work in Excel. This raw, unautomated approach is time-consuming and could lead to mistakes. Moreover, there is no unified database it’s hard to get population-level insights and do analytics.
Solution: Our end-to-end solution PathoX is a platform for easy data ingestion, feature extraction and mapping to enhance pathologists’ workflow. An importer workflow supports the mapping of columns and notifies users about missing information. We use FHIR to guarantee future interoperability and compliance with future legislation. Pathologists can also highlight and leave comments on the data, as in their usual workflows, and look at analytics on all files.
Impact: If PathoX is put into production, it will increase the quality of pathological data annotation and analysis, and, more importantly, bring data to one standardised format. This would allow for future analytics and querying on large populations, and it would be the foundation for more advanced future applications, e.g. with AI. It could save pathologists time, reduce mistakes, and ultimately lead to improving the treatment of oncological patients.
Feasibility & financials: Managing a large software system can be difficult, but leveraging the InterSystems FHIR cloud database can significantly aid in scaling. We have designed the application as an end-to-end web solution to run in a browser, making us independent of any third-party software. Since this is a purely software product not relying on GPUs we don’t expect any significant upfront costs (other than the time to develop it). Hence, we see this project as very feasible.
What is new about your solution?: The pathologists' workflows were reenvisioned with the automatic column mapping during the .xlsx files import, as well as workflows for colouring mutation and annotating data. Our solution can be extended to support advanced analytics, including applications of AI. We aim to be the first 21st-century pathology data hub and analysis tool in the Czech market, which is currently missing.
What you have built at the hackathon - text explanation + code (e.g. GitHub link): With close collaboration with domain experts and clinical pathologists, we made a pathological data ingestion and analysis software platform PathoX. It supports import of current excel formats, automates manual work, and provides analysis tools for the stored data. We have created a solid software backbone with FHIR integration. We placed a strong emphasis on considering the future of the project after the EHH23 and its application. https://github.com/MaksymPetyak/pathox-ehh23
What you had before the hackathon, please mention open source as well: Drive to do something big and a vision :] Regarding existing code/solution nothing. We accepted the challenge Friday evening and did everything on the spot at IKEM. We had access to Azure OpenAI which we did not use in our solution. For coding, we have used Pycharm Professional, GitHub, Postman, GIMP.
What comes next and what you wish to achieve: We would love to continue our work to create a fully viable product. We are thrilled! Once the solid software platform is built, we can start focusing on advanced analysis and UI experience features, leveraging our strong machine learning skills. Let's keep the momentum we have built at the EEH and continue our work immediately, working closely with the pathologists and AstraZeneca team, tailoring the app as best we can for Czech customers first, with great potential for expansion abroad.