Ordinal Classification and Regression Techniques for Distinguishing Neutrophilic Cell Maturity Stages in Human Bone Marrow
Gräbel P, Crysandt M, Klinkhammer BM, Boor P, Brümmendorf TH, Merhof D (2022)
Publication Type: Conference contribution
Publication year: 2022
Journal
Publisher: Springer Science and Business Media Deutschland GmbH
Book Volume: 13364 LNCS
Pages Range: 186-195
Conference Proceedings Title: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Event location: Paris, FRA
ISBN: 9783031092817
DOI: 10.1007/978-3-031-09282-4_16
Abstract
An automated classification of hematopoietic cells in bone marrow whole slide images would be very beneficial to the workflow of diagnosing diseases such as leukemia. However, the large number of cell types and particularly their continuous maturation process makes this task challenging: the boundaries of cell type classes in this process are fuzzy, leading to inter-rater disagreement and noisy annotations. The data qualifies as ordinal data, as the order of classes is well defined. However, a sensible “distance” between them is difficult to establish. In this work, we propose several classification and regression techniques for ordinal data, which alter the encoding of network output and ground-truth. For classification, we propose using the Gray code or decreasing weights. For regression, we propose encodings inspired by biological properties or characteristics of the dataset. We analyze their performance on a challenging dataset with neutrophilic granulocytes from human bone marrow microscopy images. We show that for a sensible evaluation, it is of utmost importance to take into account the relation between cell types as well as the annotation noise. The proposed methods are straight-forward to implement with any neural network and outperform common classification and regression methods.
Involved external institutions
Rheinisch-Westfälische Technische Hochschule (RWTH) Aachen
Germany (DE)
Universitätsklinikum Aachen (UKA)
Germany (DE)
Germany (DE)
Universitätsklinikum Aachen (UKA)
Germany (DE)
How to cite
APA:
Gräbel, P., Crysandt, M., Klinkhammer, B.M., Boor, P., Brümmendorf, T.H., & Merhof, D. (2022). Ordinal Classification and Regression Techniques for Distinguishing Neutrophilic Cell Maturity Stages in Human Bone Marrow. In Mounîm El Yacoubi, Eric Granger, Pong Chi Yuen, Umapada Pal, Nicole Vincent (Eds.), Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (pp. 186-195). Paris, FRA: Springer Science and Business Media Deutschland GmbH.
MLA:
Gräbel, Philipp, et al. "Ordinal Classification and Regression Techniques for Distinguishing Neutrophilic Cell Maturity Stages in Human Bone Marrow." Proceedings of the 3rd International Conference on Pattern Recognition and Artificial Intelligence, ICPRAI 2022, Paris, FRA Ed. Mounîm El Yacoubi, Eric Granger, Pong Chi Yuen, Umapada Pal, Nicole Vincent, Springer Science and Business Media Deutschland GmbH, 2022. 186-195.
BibTeX: Download