Leveraging Global Diagnosis For Tumor Localization In Dynamic Cell Imaging Of Breast Cancer Tissue Towards Fast Biopsying

Mandache, Diana; Guillaume, Émilie Benoit à la; Mathieu, Marie Christine; Olivo‐Marin, Jean‐Christophe; Meas-Yedid, Vannary

doi:10.1109/isbi48211.2021.9434110

Cited by 2 publications

(2 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The initial state of the main branch model ϕ 0 is based on a VGG-16 [14] backbone which takes inputs of any size by virtue of its global average pooling (GAP) bottleneck between CONV and FC. This model was trained in [15] in a FS manner on D1 by minimizing the weighted binary cross entropy (wBCE) loss.…”

Section: Mil Architecturementioning

confidence: 99%

“…This work is based on previous work [15] on a breast surgical excisions dataset (D1) comprising 47 samples (11 normal, 36 malignant) coming from 33 patients who have undergone mastectomies. Due to the large amount of tissue, a median of 10 ROIs (between 3 and 16) were acquired per sample summing up to 400 ROIs of 1440 × 1440 px (the size of the DCI field-of-view).…”

Section: Surgical Excisions Dataset (D1)mentioning

confidence: 99%

See 1 more Smart Citation

The Lifecycle of a Neural Network in the Wild: A Multiple Instance Learning Study on Cancer Detection from Breast Biopsies Imaged with Novel Technique

Mandache

Guillaume

Badachi³

et al. 2022

2022 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

In the context of tissue examination for breast cancer assessment, we propose a label-free imaging based on Optical Coherence Tomography (OCT) signal combined with a multiple instance learning (MIL) model to respond to a critical need for fast at point-of-care diagnosis: biopsy or surgery time. This new imaging, Dynamic Cell Imaging (DCI), is the time-resolved variant of Full-Field OCT (FFOCT) and offers an intra-cellular resolution of about 1 micron, together with optical sectioning and an improved cell contrast. In order to tackle the challenges of limited data and annotations, while remaining in the scope of interpretability, we design an instance-level MIL model with a focus on adapted data sampling. The interest of this method is that it incorporates taskspecific feature learning and also produces instance predictions. For a dataset of 150 core-needle biopsies, we achieve a considerable improvement of more than 20 percentage points in specificity and about 10 in accuracy by leveraging intradomain (as compared to extra-domain) pre-training.

show abstract

Section: Mil Architecturementioning

confidence: 99%

Section: Surgical Excisions Dataset (D1)mentioning

confidence: 99%