Team-Ovomaltine

HS 2023 Presentation: 07/12/2023

Team members

Saengrawi Gogniat,

Sunita Vijay Shelke,

Divya Sarah Abraham,

Aleksandra Iskrzynska.

Coaches

Andreas Martin Charuta Pande

Histopathology department for breast cancer specimen analysis

Aim to improve the process from specimen registration, specimen analysis, and report generation by using digitalization.

Description

This project focuses on transforming the analysis process within the pathology department, specifically for breast biopsy assessments. We aim to revolutionize the conventional approach by integrating advanced technology to expedite the diagnostic process and enhance accuracy. The project incorporates the utilization of an AFM (Atomic Force Microscopy) measurement device, specifically the Artidis, to assess the stiffness of the biopsy sample. Additionally, an image analysis techniques will be employed to automatically classify histopathology slides, reducing the reliance on manual human analysis.

Goals

1. Reducing Waiting Time: The prolonged waiting period for biopsy results often causes distress and hampers timely treatment. By integrating AFM measurement and automated image analysis, we aim to significantly reduce the time required for diagnosis and reporting.

2. Enhancing Accuracy: Human analysis of biopsy samples may lead to errors in diagnosis, resulting in potential overtreating of patients due to incorrect assessments. By introducing additional biomarkers like AFM measurements and automated image classification, we aim to improve diagnostic accuracy and minimize the risk of misdiagnosis.

3. Automation of Analysis: The integration of the Artidis AFM device and image analysis tools will facilitate the automation of certain aspects of the analysis process, reducing the burden on pathologists and enabling more efficient workflows.

Methods

• Implement technology to enhance the analysis process, including Artedis tissue stiffness measurement and machine learning histopathology image recognition. The final result is the integration of all measurements and specimen eligibility status, including the histopathology report from the pathologist, in the final report.
• Integrate Services using HTTP Connector with Python Flask in Deepnote
• Programming language: Python
• Camunda Modeler Platform 7

AS-IS process

TO-BE process

Access to prototypes

Presentation: https://fhnw365-my.sharepoint.com/:p:/r/personal/saengrawi_gogniat_students_fhnw_ch/Documents/DGBP%20project.pptx?d=w6861d6cce5534e1da8f91e9b73d6cb6a&csf=1&web=1&e=YKRNMv

Deepnote link: https://deepnote.com/workspace/l04-7d15acad-6de9-4091-891f-efd145100106/project/OlaAPI-69eb62be-3c1b-49e2-ac1f-d13d4e0bc6d3/notebook/Own%20API%20(Flask)-0fe27415a1ac4ffca2033f4b1f891436#0a4b4afdab044098a22e78ef8fae926a

Github: https://github.com/DigiBP/Team-Ovomaltine

Running the process

Warning: for demo purposes the process can be run for the specimen_IDs from 1 to 20. This restriction is connected to the images of the histopathology slides that are available. To run the process:

1. Run the Deepnote Notebook
2. Open the BPMN diagram via Camunda Modeler: BPMN_Diagram_Improvement.bpmn
3. Deploy the model in Camunda
4. Open Camunda Cockpit and navigate to start a new process: SpecimenJourney_experiment
5. Run through all of the steps of Camunda Tasks

Tools

• Deepnote
• REST API with Flask - http connector
• Camunda Platform 7
• Microsoft Office Tools

Technical Requirements

All required packages for Deepnote can be found in 03_Scripts/01_Deepnote/requirements.txt file in this repository.

Conclusion

Through the integration of advanced technology such as the Artidis AFM device and automated image analysis, significant strides have been made in reducing the overall time required for breast biopsy analysis. Initial results indicate a notable decrease in the time taken from biopsy collection to the final diagnostic report. Despite the incorporation of a new biomarker, the comprehensive approach has expedited the analysis process, enhancing efficiency without compromising accuracy.

Potential for Future Advancements Integrating the Artidis technology marks a pivotal step toward enhancing our diagnostic capabilities. While the incorporation of a new biomarker initially influenced the analysis time, the groundwork laid by introducing this technology opens doors for further advancements. In the future, leveraging the Artidis device may pave the way for additional, rapid diagnostic methods for breast cancer assessment. The potential for an expedited and comprehensive diagnosis process remains a promising prospect, aiming to further streamline and optimize the pathology department's workflow.

References

Dataset source: https://web.inf.ufpr.br/vri/databases/breast-cancer-histopathological-database-breakhis/

[1] Spanhol, F., Oliveira, L. S., Petitjean, C., Heutte, L., A Dataset for Breast Cancer Histopathological Image Classification, IEEE Transactions on Biomedical Engineering (TBME), 63(7):1455-1462, 2016

Plodinec, M., Loparic, M., Monnier, C. et al. The nanomechanical signature of breast cancer. Nature Nanotech 7, 757–765 (2012). https://doi.org/10.1038/nnano.2012.167