Ranking Loss and Sequestering Learning for Reducing Image Search Bias in Histopathology

Mazaheri, Pooria; Bidgoli, Azam Asilian; Rahnamayan, Shahryar; Tizhoosh, Hamid R.

doi:10.2139/ssrn.4216426

Cited by 4 publications

(5 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, based on existing literature, a distance measure is typically used to determine how similar (or dissimilar) 2 images are, a measurement that uses metrics, such as Euclidean distance that quantifies the dissimilarity between 2 given feature vectors representing the 2 images. 45 , 46 In image retrieval, Euclidean distance is favored over metrics, such as cosine similarity, because it accounts for both the magnitude and direction of feature vectors. In addition, it is a more intuitive approach.…”

Section: Resultsmentioning

confidence: 99%

A Preliminary Investigation into Search and Matching for Tumor Discrimination in World Health Organization Breast Taxonomy Using Deep Networks

Shafique,

Gonzalez,

Pantanowitz

et al. 2024

Modern Pathology

View full text Add to dashboard Cite

Section: Resultsmentioning

confidence: 99%

A Preliminary Investigation into Search and Matching for Tumor Discrimination in World Health Organization Breast Taxonomy Using Deep Networks

Shafique,

Gonzalez,

Pantanowitz

et al. 2024

Modern Pathology

View full text Add to dashboard Cite

“…The dropping accuracy of institution classification from 73 to 41% for the lung dataset and from 68 to 43% for brain dataset in the proposed model shows that the ranking loss function reduces bias in the trained model to learn the source site institution. In essence, instead of learning source site institutions during the training, the sequestering with ranking loss model concentrates on salient hospital-agnostic image features [ 29 ].…”

Section: Discussionmentioning

confidence: 99%

Biased data, biased AI: deep networks predict the acquisition site of TCGA images

et al. 2023

Self Cite

View full text Add to dashboard Cite

Background Deep learning models applied to healthcare applications including digital pathology have been increasing their scope and importance in recent years. Many of these models have been trained on The Cancer Genome Atlas (TCGA) atlas of digital images, or use it as a validation source. One crucial factor that seems to have been widely ignored is the internal bias that originates from the institutions that contributed WSIs to the TCGA dataset, and its effects on models trained on this dataset. Methods 8,579 paraffin-embedded, hematoxylin and eosin stained, digital slides were selected from the TCGA dataset. More than 140 medical institutions (acquisition sites) contributed to this dataset. Two deep neural networks (DenseNet121 and KimiaNet were used to extract deep features at 20× magnification. DenseNet was pre-trained on non-medical objects. KimiaNet has the same structure but trained for cancer type classification on TCGA images. The extracted deep features were later used to detect each slide’s acquisition site, and also for slide representation in image search. Results DenseNet’s deep features could distinguish acquisition sites with 70% accuracy whereas KimiaNet’s deep features could reveal acquisition sites with more than 86% accuracy. These findings suggest that there are acquisition site specific patterns that could be picked up by deep neural networks. It has also been shown that these medically irrelevant patterns can interfere with other applications of deep learning in digital pathology, namely image search. Summary This study shows that there are acquisition site specific patterns that can be used to identify tissue acquisition sites without any explicit training. Furthermore, it was observed that a model trained for cancer subtype classification has exploited such medically irrelevant patterns to classify cancer types. Digital scanner configuration and noise, tissue stain variation and artifacts, and source site patient demographics are among factors that likely account for the observed bias. Therefore, researchers should be cautious of such bias when using histopathology datasets for developing and training deep networks.

show abstract

“…To construct the training datasets, we initially integrated the entire TCGA dataset, encompassing the training, validation, and test subsets associated with both LUSC and LUAD cancer subtypes in order to examine the data distribution effect in the biased training procedure. Subsequently, we partitioned each patch into 25 sub-patches in 224x224 pixels 25 , Figure 2. The pretrained KimiaNet structure applied to the TCGA dataset.…”

Section: Feature Extractionmentioning

confidence: 99%

“…Thus, it raises doubts about being biased toward these unrelated signatures for cancer-type detection, which could potentially result in low external validation when dealing with the data collected from unseen data centers. Notably, several studies 25,26 have endeavored slide-based investigations to eliminate these signatures, with the goal of reducing the prediction accuracy of acquisition sites; by doing so, they aim to achieve higher external validation accuracy. However, before proceeding with any further steps, identifying the origin of these signatures as biases not only prevents the occurrence of suddenly biased results in similar histopathology research but also enhances generalization and trustworthiness toward applying AI models for aiding in diagnosis procedures.…”

Section: Introductionmentioning

confidence: 99%

Bias in Histopathology Datasets: A Comprehensive Investigation on Possible Factors

Kheiri,

Rahnamayan,

Makrehchi

et al. 2024

Preprint

View full text Add to dashboard Cite

Deep Neural Networks (DNNs) have demonstrated remarkable capabilities in medical applications, including digital pathology, where they excel at analyzing complex patterns in medical images to assist in accurate disease diagnosis and prognosis. However, concerns have arisen about potential biases in The Cancer Genome Atlas (TCGA) dataset, a comprehensive repository for digitalized histopathology and a validation source for deep models, suggesting that over-optimistic results of model performance may be due to reliance on biased features rather than histological characteristics. Surprisingly, recent studies have confirmed the existence of site-specific bias in the embedded features extracted for cancer-type discrimination, leading to high accuracy in acquisition site classification. This biased behavior motivated us to conduct an in-depth analysis to investigate potential causes behind this unexpected biased ability toward site-specific pattern recognition. The analysis was conducted on two cutting-edge DNN models: KimiaNet, a state-of-the-art DNN trained on TCGA images, and the self-trained EfficientNet. Considering the classification accuracy of acquisition sites as a measure of bias, a series of experiments has been designed to examine the impact of various factors, including patching, data imbalance, and staining.

show abstract

Ranking Loss and Sequestering Learning for Reducing Image Search Bias in Histopathology

Cited by 4 publications

References 50 publications

A Preliminary Investigation into Search and Matching for Tumor Discrimination in World Health Organization Breast Taxonomy Using Deep Networks

A Preliminary Investigation into Search and Matching for Tumor Discrimination in World Health Organization Breast Taxonomy Using Deep Networks

Biased data, biased AI: deep networks predict the acquisition site of TCGA images

Bias in Histopathology Datasets: A Comprehensive Investigation on Possible Factors

Contact Info

Product

Resources

About