A new approach to detect and segment overlapping cells in multi-layer cervical cell volume images

Unbiased estimation of cell number using the automatic optical fractionator

“…The first observation is that total cell counting errors by manual and automatic methods occur in less than 5% of disector volumes, or about 4 or 5 mismatches per 100 disectors. As expected from our previous automatic algorithms to detect stained cells on tissue sections [Mouton and Durgavich, 2005; Chaudhury et al, 2013; Phoulady et al, 2015, 2016 a, b] most residual variance in the correlations of ground truth (manual) and automatic counts was due to the manual data collection. Three-dimensional reconstruction of disector stacks showed more frequent errors by the manual counting (~ 3–4 mismatches per 100 disector stacks) than by the automatic approach (~ 1–2 errors per 100 disector stacks) with most manual counting errors leading to underestimation (false negatives), e.g., sections 1–7 in Figure 4(c).…”

“…The method uses a novel combination of two recent advancements in the field of computer science: extended depth of field (EDF) images that represent 3-D neurons in a disector volume at their optimal plane of focus on a 2-D image [Valdecasas et al, 2001; Bradley and Bamford, 2004; Phoulady et al, 2015, 2016 a, b]; and a combination of segmentation algorithms to automatically count cells visualized by ordinary staining methods in the EDF image. The main innovation lies in the automatic counting of cells in disector volumes that represent a known fraction of the reference space, hence the designation automatic optical fractionator .…”

Segmentation of Cervical Cells for Automated Screening of Cervical Cancer: A Review

“…To refine the boundaries processing steps based on reachability from the nucleus centroid and image gradient information are executed. This improved version of the method proposed in the challenge achieved a Dice Coefficient of 0.861 and False Negative object rate of 0.352 [50]. IV.…”

“…And for more than one nucleus inside the clump an operation presented is performed. The data set provided in the secondoverlapping cervical cytology image segmentation challenge at ISBI 2015 is used for evaluation of propounded method [44]. Niraimathi, Mohideenfatima, and Seenivasagam Vellaichamy in their research proposes a new mathematical model for the segmentation of cervical cell nuclei.…”

Serviks Kanserinin Erken Teşhisi için Çok Katmanlı Sitoloji Küplerinde Çekirdek ve Sitoplazma Bölütlenmesi

“…Bu çalışmalarda klasik bölütleme yöntemleri, piksel tabanlı sınıflandırma ya da ikisinin de kullanıldığı melez/hibrit modeller kullanılmıştır. Örnek olarak, hücre çekirdeklerinin bölütlenmesi için alçak geçirgen gürültü giderici filtrenin ardından iteratif eşikleme [5], yıldız şekil öncülleri temelinde yönlü türevler yardımıyla görüntü işleme tabanlı yaklaşımla bölütlenmesi [6], çok-adımlı aşama seviye kümesi yöntemi yaklaşımları kullanılmıştır [7]. Hücre çekirdeği ve sitoplazmasının bölütlenmesi için piksel yoğunluklarına bağlı Gauss karışım modeli çoklu özellik çıkarım yöntemleri kullanılarak uygulanmıştır [8].…”