PhD for investigating Robust and reliable Artificial Intelligence for oncology

Early detection, accurate diagnosis/staging of cancer, and subsequent selection of the appropriate treatment are critical factors to improve patient outcomes. Diagnostic imaging plays an integral role in many phases of a patient's care path, but considerable challenges remain regarding the accuracy, efficacy and efficiency. Artificial Intelligence (AI) holds great promise to support image analysis to detect, characterize, and monitoring disease. AI can enable automated segmentation and support in diagnosis (e.g. classifying abnormalities as benign or malignant) and staging (categorizing tumors into pre-defined groups based on expected disease course and treatment strategy). Furthermore, it can support cancer monitoring by capturing image features over time to evaluate the patient's response to treatment.

Many AI solutions for radiology have been developed in research labs, but few have been adopted successfully in clinical practice, limiting their impact on patient outcomes. A well-recognized cause for this poor adoption is a lack of human-centered design of AI, leading to solutions to clinically insignificant problems, too much or too little trust in AI outcomes and poor fit of AI solutions in clinician's ways of working. This project will therefore develop tools, methods and assets to bring AI to clinical practice and evaluate actual use, trust, and experience with AI. The design of optimal physician-AI collaboration combines data science and UI/UX expertise to address technical challenges, such as AI explainability, robustness and uncertainty.

Advancing Cancer care through Interpretable AI

The Eindhoven MedTech Innovation Center (e/MTIC) is a public-private partnership aimed at creating and growing an ecosystem that offers a fast track to high-tech health innovations. Within this ecosystem, we aim to create impactful innovations in the oncology domain, through close multidisciplinary collaboration between an academic partner (Eindhoven University of Technology), a top-clinical hospital (Catharina Hospital Eindhoven) and an industrial partner (Philips). In this triangle, the research project 'Advancing Cancer care through Interpretable AI' (ACACIA) aims to advance clinical decision support for cancer care with Artificial Intelligence (AI), taking a patient-centered and physician-centered approach to optimize physician-AI collaboration and adoption of AI in clinical practice. This project focusses on the integration of AI in CT-based imaging systems and builds on prior work that has been developed within e/MTIC in recent years.

You will be part of an innovative multi-disciplinary team

This PhD position is embedded in a multidisciplinary team of in total 4 PhD positions covering machine learning, product design and clinical expertise. The position advertised here focusses on 'Bringing AI into clinical practice', including clinical data collection, user testing and clinical evaluation of the AI solutions.. Around the core team, a group of domain experts is closely involved in the project, supporting the PhD researchers. This group consists of medical specialists such as radiologists and surgeons, but also experienced research scientists from Philips and Eindhoven University of Technology. The project activities are organized in 3 tightly interlinked workpackages:

WP1. Machine Learning for robust & trustworthy AI

This activity, led by a machine learning PhD student, aims to develop robust and trustworthy
AI in medical imaging. It will address a subset of clinical application areas (e.g., lung cancer and pancreas cancer), by development and maturation of AI methods for the selected clinical use cases. Research topics include, but are not limited to, AI transparency, interpretability and explainability, self-critical AI, confidence quantification and out-of-distribution detection.
A more detailed description is provided below.

WP2. User experience and interface design and evaluation for optimal physician-AI collaboration

This activity, led by a design PhD student, aims to iteratively develop and evaluate UI/UX innovations to shape optimal physician-AI collaboration, as well as methods to evaluate the clinical experience with end-users. This will drive the AI models developed in WP1 and integrate AI-based outputs generated in WP1.

WP3: Clinical data collection, user testing and evaluation

In this activity, led by two clinical PhD students, the AI algorithms developed in WP1 and clinician-AI collaboration innovations developed in WP2 will be evaluated in clinical practice. These comprise the evaluation of AI-based lung nodule assessment, evaluation of AI-based pancreatic cancer detection and resectability assessment and testing scalability of the innovations to a third cancer type (e.g. other abdominal cancer such as kidney cancer). We are interested in what is needed for an optimal physician-AI collaboration and what are benefits of AI solutions in the field; e.g. adoption and appropriate use, impact on patient outcomes, physician experience while working with the AI solution.

The 4 PhD candidates will closely collaborate and share data, AI models, technological platforms and methodologies as complementary focus points of each PhD position.

The clinical PhD position advertised here focuses on the clinical aspects of this project. Specific tasks goal include:

• Collecting clinical data and scans;
• Annotating CT-scans;
• Providing clinical requirements for the AI models and their usage in clinical workflows
• Checking and correcting output of the developed algorithms;
• Measuring benefits of AI-clinician collaboration
• Correlating AI results with follow-up and outcomes