3D histology of thyroid cancer and the normal thyroid

Supervisor: Dr. Vincent Detours

Our knowledge of the thyroid gland microanatomy mostly rests on 2D microscopy, with the notable exception of a handful of 3D wax models assembled by manually staking 2D views, back in 1920-1930. The goal of this computational biology project is to produce modern era 3D models of the human thyroid in health and in cancer that resolve and annotate every cell within cubic centimeters volumes.

Our lab has produced a first 3D model of a papillary thyroid cancer. It revealed a fractal-like growth pattern whereby cancer cell infiltrated a highly proliferative stroma, resulting in surprisingly extensive cancer/stroma interface that challenge the textbook concept of a tumor with  an inner core and an invasive front. The project will develop this approach by 1- automating the registration of nearly 1,000 consecutive slices per sample (i.e., gigapixel images from bright field microscope at 40X); 2- automatically annotating the slices with a self-supervised convolutional neural network we developed in-house to recognize epithelium, colloid, and other morphological structures.  3- segmenting all nuclei in all slides; 4- An additional layer of annotation will come from spatial transcriptomic slides evenly spaced across the tissue volume.

The outcome of this project will be a series of fundamental reference models for the 3D histology of the normal thyroid (from transplant donors) and cancer models setting the stage for future pathology.  

The computational group is an interdisciplinary team of wet lab and computational biologists. IRIBHM is an historical player in thyroid research. Successful applicants will be embedded in an environment with strong biomedical expertise and with a deep integration of the experimental and the computational sides of biology.

Applicants have proven programming skills and a background in mathematics, computer science, physics or engineering.

Keywords: anatomy, thyroid, imaging, artificial intelligence, spatial transcriptomics


  1. Saiselet M., Rodrigues-Vitória J., Tourneur A., Craciun L., Spinette A., Larsimont D., Andry G., Lundeberg J., Maenhaut C., Detours V., (2020), Transcriptional output, cell types densities and normalization in spatial transcriptomics, J. Mol. Cell Biol., doi: 10.1093/jmcb/mjaa028
  2. Tarabichi M., Antoniou A., Le Pennec S., Gacquer D., de Saint Aubain N., Craciun L., Cielen T., Laios I., Larsimont D., Andry G., Dumont J.E., Maenhaut C., Detours V., (2018), Distinctive desmoplastic 3D morphology associated with BRAFV600E in papillary thyroid cancers, J Clin Endocrinol Metab. doi:10.1210/jc.2017-02279.

Back to Projects