Contexte

Accurate cancer diagnosis often relies on histological examinations, where tissue samples are analyzed to confirm the presence of tumors and characterize lesions [1]. At Primaa, we develop deep learning methods to classify lesions, detect biomarkers, and segment regions of interest in tissue samples.

This internship focuses on applied research for PD-L1 classification in kidney cancer. PD-L1 is a protein often expressed in cancer cells that can inhibit the immune response. Measuring PD-L1 levels, typically assessed using immunohistochemistry (IHC) (Figure 1), helps guide immunotherapy treatments, making accurate classification essential for effective patient care. In addition to IHC data, hematoxylin and eosin (HE) stained slides are also available for each case, providing a complementary view of the tissue structure.

Figure 1. Representative IHC images of different PD-L1 expressions, from low (A) to high expression (D).

[1]https://www.sigmaaldrich.com/FR/fr/applications/clinical-testing-and-diagnostics-manufacturing/histology?srsltid=AfmBOopIdAvPC9UvUjinhpvSgJl2Lq3hnEwdVN41aH-m6eXMbPmhmsvV#staining-in-histopathology
[2] *Weakly supervised multiple instance learning histopathological tumor segmentation, M.Lerousseau, M.Vakalopoulou, M.Classe, J.Adam, E.Battistella, A.Carré, T.Estienne, T.Henry, E.eutsch and N.Paragios
*[3] Bootstrapping Semantic Segmentation with Regional Contrast, S.Liu, S.Zhi, E.Johns and A.J. Davison

Objectives

• Literature Review: Select relevant approaches for PDL1 classification.

• Computer Vision Tasks:

  • Perform classification at the slide level using weak annotations.

  • Conduct patch-level classification or segmentation using full or semi-annotations.

• Model Explainability: Enhance the model’s interpretability to increase clinical usability.

• Product Development: Contribute to the clinical application of the proposed solution.

Practical Informations

• Remuneration: €1000/month + 50% Navigo refund.

• Location: Primaa is based in Paris, 2ème arrondissement, with the possibility of occasional remote work.

• Applied mathematics student.

• Six months internship (start between Feb and Apr 2024)

• Knowledge in python.

• Knowledge in computer vision (image processing, deep learning) : tensorflow and/or pytorch, scikit-image…

Meeting with two data team members.