Contrastive Multiple Instance Learning: An Unsupervised Framework for Learning Slide-Level Representations of Whole Slide Histopathology Images without Labels

Tavolara, Thomas E.; Gürcan, Metin N.; Niazi, M. Khalid Khan

doi:10.3390/cancers14235778

Cited by 17 publications

(13 citation statements)

References 86 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…At the moment, backbones pretrained on Medical Images perform poorly when compared to those pretrained of ImageNet-1K or ImageNet-22K. This confirms what recent works have found [22,7]. One possible explanation for this is the fact that the best performing available models have been trained in an unsupervised manner for a long time on large GPU clusters [5], whereas those pretrained on Medical Images had access to more limited ressources and less diverse data [7].…”

Section: Backbone Selectionsupporting

confidence: 78%

Classification in Histopathology: A unique deep embeddings extractor for multiple classification tasks

Nivaggioli¹,

Pozin²,

Peyret³

et al. 2023

Preprint

View full text Add to dashboard Cite

Section: Backbone Selectionsupporting

confidence: 78%

Classification in Histopathology: A unique deep embeddings extractor for multiple classification tasks

Nivaggioli¹,

Pozin²,

Peyret³

et al. 2023

Preprint

View full text Add to dashboard Cite

“…Then, a MIL model, MI-LR, was used to predict the patient-level positive probability with the cell vectors belonging to the sample. For the samples of exfoliated margin cells collected from 27 , and selfsupervised CL does not require exogenous labels 54,55 . MICLEAR utilized self-supervised SimCLR for cell embedding in consideration of high intercellular heterogeneity occurring in the positive samples of exfoliated cells, which may introduce huge label noise when using the patient-level labels directly.…”

Section: Discussionmentioning

confidence: 99%

Artificial intelligence-assisted stimulated Raman molecular cytology enables accurate and rapid intraoperative assessment of the surgical margin in pancreatic ductal adenocarcinoma

Fang,

Liu,

Chen

et al. 2023

Preprint

View full text Add to dashboard Cite

Pancreatic ductal adenocarcinoma (PDAC) is a highly mortal cancer with surgical resection the only potentially curative treatment. The accurate intraoperative assessment of PDAC surgical margins is significant in guaranteeing resection adequacy and improving the patient’s survival. However, the commonly used frozen-section pancreatic biopsy is limited by its strict requirement of practitioners and lack of objectivity. Here, we developed the multi-instance cytology with learned Raman embedding (MICLEAR), a label-free cytology based on chemical information reflected by Raman spectra. First, 4085 cells collected from 41 patients were imaged with stimulated Raman scattering (SRS) microscopy. Then, a contrastive learning (CL)-based cell embedding model was obtained to represent each cell with a concise vector that contained its morphological and componential information. Finally, a multi-instance learning (MIL)-based diagnosis model using cell vectors predicted the probability that the margin was positive. MICLEAR reached 80% sensitivity, 94.1% specificity, and 0.86 AUC on the pancreatic neck margin (PNM) samples from 27 patients. It holds promise for rapid and accurate assessment of PDAC surgical margins.

show abstract

“…Indeed, in study [ 32 ], it was shown how a VGG-16 convolutional neural network outperformed (accuracy: 0.841, AUC: 0.903) classical machine learning methods in differentiating between the wo NSCLC subtypes. Furthermore, in study [ 61 ], a self-supervised learning approach reached an AUC that was equal to 0.8641.…”

Section: Limitationmentioning

confidence: 99%

Phenotyping the Histopathological Subtypes of Non-Small-Cell Lung Carcinoma: How Beneficial Is Radiomics?

et al. 2023

View full text Add to dashboard Cite

The aim of this study was to investigate the usefulness of radiomics in the absence of well-defined standard guidelines. Specifically, we extracted radiomics features from multicenter computed tomography (CT) images to differentiate between the four histopathological subtypes of non-small-cell lung carcinoma (NSCLC). In addition, the results that varied with the radiomics model were compared. We investigated the presence of the batch effects and the impact of feature harmonization on the models’ performance. Moreover, the question on how the training dataset composition influenced the selected feature subsets and, consequently, the model’s performance was also investigated. Therefore, through combining data from the two publicly available datasets, this study involves a total of 152 squamous cell carcinoma (SCC), 106 large cell carcinoma (LCC), 150 adenocarcinoma (ADC), and 58 no other specified (NOS). Through the matRadiomics tool, which is an example of Image Biomarker Standardization Initiative (IBSI) compliant software, 1781 radiomics features were extracted from each of the malignant lesions that were identified in CT images. After batch analysis and feature harmonization, which were based on the ComBat tool and were integrated in matRadiomics, the datasets (the harmonized and the non-harmonized) were given as an input to a machine learning modeling pipeline. The following steps were articulated: (i) training-set/test-set splitting (80/20); (ii) a Kruskal–Wallis analysis and LASSO linear regression for the feature selection; (iii) model training; (iv) a model validation and hyperparameter optimization; and (v) model testing. Model optimization consisted of a 5-fold cross-validated Bayesian optimization, repeated ten times (inner loop). The whole pipeline was repeated 10 times (outer loop) with six different machine learning classification algorithms. Moreover, the stability of the feature selection was evaluated. Results showed that the batch effects were present even if the voxels were resampled to an isotropic form and whether feature harmonization correctly removed them, even though the models’ performances decreased. Moreover, the results showed that a low accuracy (61.41%) was reached when differentiating between the four subtypes, even though a high average area under curve (AUC) was reached (0.831). Further, a NOS subtype was classified as almost completely correct (true positive rate ~90%). The accuracy increased (77.25%) when only the SCC and ADC subtypes were considered, as well as when a high AUC (0.821) was obtained—although harmonization decreased the accuracy to 58%. Moreover, the features that contributed the most to models’ performance were those extracted from wavelet decomposed and Laplacian of Gaussian (LoG) filtered images and they belonged to the texture feature class.. In conclusion, we showed that our multicenter data were affected by batch effects, that they could significantly alter the models’ performance, and that feature harmonization correctly removed them. Although wavelet features seemed to be the most informative features, an absolute subset could not be identified since it changed depending on the training/testing splitting. Moreover, performance was influenced by the chosen dataset and by the machine learning methods, which could reach a high accuracy in binary classification tasks, but could underperform in multiclass problems. It is, therefore, essential that the scientific community propose a more systematic radiomics approach, focusing on multicenter studies, with clear and solid guidelines to facilitate the translation of radiomics to clinical practice.

show abstract

Contrastive Multiple Instance Learning: An Unsupervised Framework for Learning Slide-Level Representations of Whole Slide Histopathology Images without Labels

Cited by 17 publications

References 86 publications

Classification in Histopathology: A unique deep embeddings extractor for multiple classification tasks

Classification in Histopathology: A unique deep embeddings extractor for multiple classification tasks

Artificial intelligence-assisted stimulated Raman molecular cytology enables accurate and rapid intraoperative assessment of the surgical margin in pancreatic ductal adenocarcinoma

Phenotyping the Histopathological Subtypes of Non-Small-Cell Lung Carcinoma: How Beneficial Is Radiomics?

Contact Info

Product

Resources

About