Automatic Bone Marrow Cellularity Estimation in H&E Stained Whole Slide Images

Abstract

Bone marrow cellularity is an important measure in diagnostic hematopathology. Currently, the gold standard for bone marrow cellularity estimation is manual inspection of hematoxylin and eosin stained whole slide images (H&E WSI) by hematopathologists. However, these assessments are subjective and subject to interobserver and intraobserver variability. This may be reduced by using a computer-assisted estimate of bone marrow cellularity. The aim of this study was to develop a fully automated algorithm to estimate bone marrow cellularity in H&E WSI stains using bone marrow segmentation. Data consisted of eight bone marrow H&E WSIs extracted from eight subjects. An algorithm was developed to estimate the bone marrow cellularity consisting of biopsy segmentation, tissue classification, and bone marrow segmentation. Segmentations of the red and yellow bone marrow (YBM) were used to estimate the bone marrow cellularity within the WSI H&E stains. The DICE coefficient between automatic tissue segmentations and ground truth segmentations conducted by an experienced hematopathologist were used for validation. Furthermore, the agreement between the automatic and two manual cellularity estimates was assessed using Bland-Altman plots and intraclass correlation coefficients (ICC). The validation of the bone marrow segmentation demonstrated an average DICE of 0.901 and 0.920 for the red and YBM, respectively. A mean cellularity estimate difference of −0.552 and − 7.816 was obtained between the automatic cellularity estimates and two manual cellularity estimates, respectively. An ICC of 0.980 (95%CI: 0.925-0.995, P-value: 5.51 × 10 −7 ) was obtained between the automatic and manual cellularity estimates based on manual annotations. The study demonstrated that it was possible to obtain bone marrow cellularity estimates with a good agreement with bone marrow cellularity estimates obtained from an experienced hematopathologist.

Using deep learning for quantification of cellularity and cell lineages in bone marrow biopsies and comparison to normal age-related variation

“…Both studies 4,12 also found inter-rater agreement of visual estimation by pathologists comparable to our study (ICC=0.88-0.91 and 0.870, respectively). Nielsen et al 10 trained a support vector machine to segment biopsies in red (haematopoiesis) and yellow (lipocytes) marrow and used this in a similar way to the present paper to calculate cellularity, with good agreement with visual estimation (ICC=0.799). They reported a higher Dice score for lipocytes (0.9001 versus our 0.88), possibly because they used full segmentation over point annotation.…”

“…Several recent studies reported cellularity measurement using digital image analysis techniques on digitised slides that show good agreement with references of visual estimate or point counting. 4,[10][11][12] These studies have used traditional machine learning techniques, while the field of medical image analysis has shown a shift towards generally better performing deep learning systems. 13 Deep learning has already been applied to the segmentation of erythropoiesis and myelopoiesis, 14 but no work has been published on the simultaneous segmentation of all major cell types in bone marrow, which would allow for a more detailed analysis of the tissue.…”

MarrowQuant 2.0: a digital pathology workflow assisting bone marrow evaluation in clinical and experimental hematology

“…The algorithm should be however used with caution in clinical scenarios where megakaryocyte hyperplasia is expected, including myeloproliferative syndromes. Future versions of the algorithm which will incorporate deep learning and machine learning (35,59) are under development to overcome this problem.…”

“…We tested MarrowQuant 2.0 on H&E BM images from acute myeloid leukemia (AML) and high-risk myelodysplasia (MDS) patients (n = 28) at diagnosis (Dx) and at three timepoints after intensive myeloablative chemotherapy, corresponding to the peak of aplasia of the rst chemotherapy cycle (A, day 17-21), the hematopoietic recovery after the rst cycle of chemotherapy (RC1, day [35][36][37][38][39][40][41][42][43] and the hematopoietic recovery after the second cycle of chemotherapy (RC2), as well as an age-matched control group from hip replacement surgical bone specimens, n = 15 (Fig. 4A).…”

Improved heterogeneous data fusion and multi‐scale feature selection method for lung cancer subtype classification

“…Then the image features can be used to correlate with the genetic characteristics of the tumor 24–27 . There have been many studies on the diagnosis of cancer using pathological imaging and computer technology 28–32 . Liu et al 33 used Inception architecture to classify tumor and segment cancer regions in pathological images.…”