Measurements of Morphometric Parameters of Blood Cells (MPBC) playa key role in haematological examination, and it is considered as one of the principal needs for clinicians in the diagnosis of various diseases in human and animals. Obliviously, the correctness of the diagnosis, and, as a consequence, the effectiveness of clinician actions is highly dependent on the accuracy of MPBC measurements. In this context, deep learning based MPBC measurement systems are a promising solution. In recent studies, researchers have applied semantic segmentation with various backbone networks for white blood cell measurements. Vice versa, few investigations were done about the achieved accuracy. Indeed, accurate segmentation of white blood cell remains a challenging task because of the complex nature of cell images, staining techniques, and imaging conditions which strongly affects the accuracy of the MPBC measurements. This paper presents a comparison among the segmentation performance carried out by U-Net deep learning algorithm with different backbones typically used for MPBC. The goal is to make a first step towards a whole MPBC measurement system capable of evaluating the effects of the influencing magnitudes, attenuate them (if possible), and evaluate the accuracy of the measurements. The aims are to increase measurement reliability and to give clinicians further information to take their decisions.
Comparison of U -NET backbones for morphometric measurements of white blood cell
Carni D. L.;Lamonaca F.
2022-01-01
Abstract
Measurements of Morphometric Parameters of Blood Cells (MPBC) playa key role in haematological examination, and it is considered as one of the principal needs for clinicians in the diagnosis of various diseases in human and animals. Obliviously, the correctness of the diagnosis, and, as a consequence, the effectiveness of clinician actions is highly dependent on the accuracy of MPBC measurements. In this context, deep learning based MPBC measurement systems are a promising solution. In recent studies, researchers have applied semantic segmentation with various backbone networks for white blood cell measurements. Vice versa, few investigations were done about the achieved accuracy. Indeed, accurate segmentation of white blood cell remains a challenging task because of the complex nature of cell images, staining techniques, and imaging conditions which strongly affects the accuracy of the MPBC measurements. This paper presents a comparison among the segmentation performance carried out by U-Net deep learning algorithm with different backbones typically used for MPBC. The goal is to make a first step towards a whole MPBC measurement system capable of evaluating the effects of the influencing magnitudes, attenuate them (if possible), and evaluate the accuracy of the measurements. The aims are to increase measurement reliability and to give clinicians further information to take their decisions.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.