Automated bone marrow cytology using deep learning to generate a histogram of cell types

Tayebi, Rohollah Moosavi; Mu, Youqing; Dehkharghanian, Taher; Ross, Catherine; Sur, Monalisa; Foley, Ronan; Tizhoosh, Hamid R.; Campbell, Clinton J. V.

doi:10.1038/s43856-022-00107-6

Cited by 54 publications

(39 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Rapidly developing AI object detection technology has been successfully applied to many medical issues. Tayebi et al (2022) applied the YOLO model to automatically identify and detect all bone marrow cells in each region, supporting a more precise hematological diagnosis. Lee et al (2022) proposed the RCNN model to analyze various abnormal teeth types.…”

Section: Discussionmentioning

confidence: 99%

“…The purpose of object detection is to find all objects of interest in an image and determine their positions and sizes ( Mane et al, 2008 ). An object detection diagnostic model plays an essential role in the diagnosis and prognosis of disease, including bone marrow cell automatic detection ( Tayebi et al, 2022 ), tumor region identification in breast cancer samples ( Joseph et al, 2019 ), and tongue cancer diagnosis ( Heo et al, 2022 ). Therefore, object detection techniques may potentially contribute to solving HPCF diagnosis drawbacks.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A study on the diagnosis of the Helicobacter pylori coccoid form with artificial intelligence technology

Zhong

Wang²,

Li³

et al. 2022

Front. Microbiol.

View full text Add to dashboard Cite

BackgroundHelicobacter pylori (H. pylori) is an important pathogenic microorganism that causes gastric cancer, peptic ulcers and dyspepsia, and infects more than half of the world’s population. Eradicating H. pylori is the most effective means to prevent and treat these diseases. H. pylori coccoid form (HPCF) causes refractory H. pylori infection and should be given more attention in infection management. However, manual HPCF recognition on slides is time-consuming and labor-intensive and depends on experienced pathologists; thus, HPCF diagnosis is rarely performed and often overlooked. Therefore, simple HPCF diagnostic methods need to be developed.Materials and methodsWe manually labeled 4,547 images from anonymized paraffin-embedded samples in the China Center for H. pylori Molecular Medicine (CCHpMM, Shanghai), followed by training and optimizing the Faster R-CNN and YOLO v5 models to identify HPCF. Mean average precision (mAP) was applied to evaluate and select the model. The artificial intelligence (AI) model interpretation results were compared with those of the pathologists with senior, intermediate, and junior experience levels, using the mean absolute error (MAE) of the coccoid rate as an evaluation metric.ResultsFor the HPCF detection task, the YOLO v5 model was superior to the Faster R-CNN model (0.688 vs. 0.568, mean average precision, mAP); the optimized YOLO v5 model had a better performance (0.803 mAP). The MAE of the optimized YOLO v5 model (3.25 MAE) was superior to that of junior pathologists (4.14 MAE, p < 0.05), no worse than intermediate pathologists (3.40 MAE, p > 0.05), and equivalent to a senior pathologist (3.07 MAE, p > 0.05).ConclusionHPCF identification using AI has the advantage of high accuracy and efficiency with the potential to assist or replace pathologists in clinical practice for HPCF identification.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A study on the diagnosis of the Helicobacter pylori coccoid form with artificial intelligence technology

Zhong

Wang²,

Li³

et al. 2022

Front. Microbiol.

View full text Add to dashboard Cite

show abstract

“…The purpose of DL is to develop decision-making tools to aid cancer researchers in their studies and health professionals in the clinical care of cancer patients [46]. In another case, Tayebi et al worked on building an end-to-end deep learning-based model for automated bone marrow cytology [47]. A computerized full slide picture of the patient’s blood was used to quickly and automatically determine areas appropriate for cytology, which was followed by the identification and classification of all of the blood cells inside those areas.…”

Section: Related Workmentioning

confidence: 99%

HematoNet: Expert Level Classification of Bone Marrow Cytology Morphology in Hematological Malignancy with Deep Learning

Tripathi

Augustin

Sukumaran

et al. 2022

Preprint

View full text Add to dashboard Cite

There have been few efforts made to automate the cytomorphological categorization of bone marrow cells. For bone marrow cell categorization, deep-learning algorithms have been limited to a small number of samples or disease classifications. In this paper, we proposed a pipeline to classify the bone marrow cells despite these limitations. Data augmentation was used throughout the data to resolve any class imbalances. Then, random transformations such as rotating between 0° to 90°, zooming in/out, flipping horizontally and/or vertically, and translating were performed. The model used in the pipeline was a CoAtNet and that was compared with two baseline models, EfficientNetV2 and ResNext50. We then analyzed the CoAtNet model using SmoothGrad and Grad-CAM, two recently developed algorithms that have been shown to meet the fundamental requirements for explainability methods. After evaluating all three models’ performance for each of the distinct morphological classes, the proposed CoAtNet model was able to outperform the EfficientNetV2 and ResNext50 models due to its attention network property that increased the learning curve for the algorithm which was represented using a precision-recall curve.

show abstract

“…To evaluate the improvement brought about by cells' deep features versus cell counts alone, we also utilised the same WSI retrieval method (Section 7) on rHCT. The rHCT accounts for only the simple counts of cells in each category, and does not account for features in each cell [23], which suggests it should be inferior to the representations from the Hopfield pooling system. As expected, the Hopfield pooling system outperformed this rHCT-based method in four out of five experiments (overall, 0.58 ± 0.02 vs. 0.55 ± 0.01, Fig.…”

Section: Wsi Retrieval Using Wsi Representationmentioning

confidence: 99%

Whole slide image representation in bone marrow cytology

Tizhoosh

Dehkharghanian

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

One of the goals of AI-based computational pathology is to generate compact WSI representations, identifying the essential information required for diagnosis. While such approaches have been applied to histopathology, few applications have been reported in cytology. Bone marrow aspirate cytology is the basis for key clinical decisions in hematology. However, visual inspection of aspirate specimens is a tedious and complex process subject to variation in interpretation, and hematopathology expertise is scarce. The ability to generate a compact representation of an aspirate specimen may form the basis for clinical decision support tools in hematology. We have previously published an end-to-end AI-based system for counting and classifying cells from bone marrow aspirate WSI. Using deep embeddings from this model, we construct bags of individual cell features from each WSI, and apply multiple instance learning to extract vector representations for each WSI. Using these representations in vector search, we achieved 0.58 ± 0.02 mAP@10 in WSI-level image retrieval, which outperforms the Random baseline (0.39 ± 0.1). Using a weighted k-nearest-neighbours (k-NN) model on these slide vectors, we predict five broad diagnostic labels on individual aspirate WSI with a weighted-macro-average F1 score of 0.57 ± 0.03 on the test set of 278 randomly sampled WSIs, which outperforms a classifier using empirical class prior probabilities (0.26 ± 0.02). We present the first example of exploring trainable mechanisms to generate compact, slide-level representations in bone marrow cytology with deep learning. This method has the potential to summarize complex semantic information in WSIs toward improved diagnostics in hematology, and may eventually support AI-assisted computational pathology approaches.

show abstract

Automated bone marrow cytology using deep learning to generate a histogram of cell types

Cited by 54 publications

References 48 publications

A study on the diagnosis of the Helicobacter pylori coccoid form with artificial intelligence technology

A study on the diagnosis of the Helicobacter pylori coccoid form with artificial intelligence technology

HematoNet: Expert Level Classification of Bone Marrow Cytology Morphology in Hematological Malignancy with Deep Learning

Whole slide image representation in bone marrow cytology

Contact Info

Product

Resources

About