Article
Computer Science, Interdisciplinary Applications
Lei Fan, Arcot Sowmya, Erik Meijering, Yang Song
Summary: In this paper, a novel framework for cancer prediction is proposed, which utilizes self-supervised learning methods to extract features from histopathological whole slide images and considers the overall survival of multiple patients. Experimental results demonstrate the excellent predictive accuracy of this framework.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Oncology
Thomas E. Tavolara, Metin N. Gurcan, M. Khalid Khan Niazi
Summary: This study proposes an unsupervised method to learn meaningful features from histopathological imaging data. The method achieves high accuracy and correlation in classifying non-small cell lung cancer subtypes and scoring breast cancer proliferation. The significance of this method lies in its ability to learn meaningful features from raw imaging data without slide-level annotations.
Article
Computer Science, Artificial Intelligence
Tiancheng Lin, Zhimiao Yu, Zengchao Xu, Hongyu Hu, Yi Xu, Chang-Wen Chen
Summary: Self-supervised representation learning has limitations when applied to whole-slide pathological images due to their unique characteristics. To address this issue, we propose a novel scheme called Spatial Guided Contrastive Learning (SGCL), which leverages spatial proximity and multi-object priors for stable self-supervision. SGCL outperforms state-of-the-art methods on diverse downstream tasks across multiple datasets.
MEDICAL IMAGE ANALYSIS
(2023)
Article
Oncology
Masayuki Tsuneki, Makoto Abe, Shin Ichihara, Fahdi Kanavati
Summary: Based on the ASCO-endorsed CCO Clinical Practice Guideline, a deep learning model was developed to classify prostate adenocarcinoma into indolent and aggressive types. The model achieved high sensitivity and specificity in classifying core needle biopsy whole slide images.
Article
Engineering, Biomedical
Tingting Zheng, Weixing Chen, Shuqin Li, Hao Quan, Mingchen Zou, Song Zheng, Yue Zhao, Xinghua Gao, Xiaoyu Cui
Summary: This study proposes an innovative weakly supervised deep reinforcement learning framework, FastMDP-RL, for fast and accurate diagnosis of cutaneous melanoma histopathology images. The framework utilizes two deep neural network-based agents to assist decision-making and reduce irrelevant information, thereby improving the efficiency and accuracy of model inference. Experimental results demonstrate that the framework can accurately predict histopathology images quickly even in the absence of pixel-level annotations.
COMPUTERIZED MEDICAL IMAGING AND GRAPHICS
(2023)
Article
Oncology
Pedro C. Neto, Sara P. Oliveira, Diana Montezuma, Joao Fraga, Ana Monteiro, Liliana Ribeiro, Sofia Goncalves, Isabel M. Pinto, Jaime S. Cardoso
Summary: This study proposes an approach to automatically detect and grade lesions in colorectal biopsies, aiming to support pathologists in their diagnosis and provide a second opinion or flag missed details. Given the increasing incidence of colorectal cancer, this AI-based automatic diagnosis method is of great importance in reducing the workload of pathologists.
Article
Oncology
Chong Wang, Xiu-Li Wei, Chen-Xi Li, Yang-Zhen Wang, Yang Wu, Yan-Xiang Niu, Chen Zhang, Yi Yu
Summary: A novel method using deep learning has been developed for fast and accurate diagnosis of hematological disorders. The method converts whole-slide image patches into low-dimensional feature representations and aggregates patch-level features into slide-level representations using an attention-based network. The model achieves high diagnostic performance on bone marrow whole-slide images and microscopy images.
FRONTIERS IN ONCOLOGY
(2022)
Article
Biology
Ziyu Su, Mostafa Rezapour, Usama Sajjad, Metin Nafi Gurcan, Muhammad Khalid Khan
Summary: This research proposes a weakly-supervised multiple instance learning model (SiiMIL) for disease classification in large-scale images. By introducing a novel representation learning method, SiiMIL improves the ratio of tumor to normal instances and achieves better classification performance. The experimental results also show that SiiMIL can generate attention heatmaps that are more interpretable to pathologists.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Computer Science, Artificial Intelligence
Xiyue Wang, Yuexi Du, Sen Yang, Jun Zhang, Minghui Wang, Jing Zhang, Wei Yang, Junzhou Huang, Xiao Han
Summary: This study proposes a robust and accurate WSI-level image retrieval framework called Retrieval with Clustering-guided Contrastive Learning (RetCCL). It integrates novel self-supervised feature learning and global ranking and aggregation algorithms for improved performance. The proposed method utilizes large-scale unlabeled histopathological image data to learn universal features for subsequent WSI retrieval tasks. The framework not only returns similar WSIs to a query WSI but also highlights patches or sub-regions with high similarity, aiding pathologists in interpreting the results.
MEDICAL IMAGE ANALYSIS
(2023)
Article
Oncology
Xiaobo Zhang, Wei Ba, Xiaoya Zhao, Chen Wang, Qiting Li, Yinli Zhang, Shanshan Lu, Lang Wang, Shuhao Wang, Zhigang Song, Danhua Shen
Summary: This study aims to develop a deep learning system for endometrial cancer detection using whole-slide images (WSIs). The model achieved a high degree of accuracy in identifying EC, serving as an assisted diagnostic tool to improve working efficiency for pathologists.
FRONTIERS IN ONCOLOGY
(2022)
Article
Oncology
Fahdi Kanavati, Masayuki Tsuneki
Summary: This study trained deep learning models using transfer learning and weakly-supervised learning to classify breast invasive ductal carcinoma in whole slide images. The models achieved high AUC values in tests and outperformed pre-trained models for adenocarcinoma classification. The results show promise for aiding pathologists in clinical practice.
Article
Biology
Le Li, Yong Liang, Mingwen Shao, Shanghui Lu, Shuilin Liao, Dong Ouyang
Summary: Understanding prognosis and mortality is crucial for evaluating patient treatment plans. With advancements in digital pathology and deep learning, performing survival analysis in whole slide images (WSIs) has become practical. However, current methods using random patch sampling and hand-crafted features or CNNs pre-trained on ImageNet have limitations. To address these challenges, a novel patch sampling strategy based on image information entropy and a Multi-Scale feature Fusion Network (MSFN) based on self-supervised feature extraction are proposed. The method achieves competitive results on popular WSIs survival analysis datasets, TCGA-GBM and TCGA-LUSC.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Oncology
Masayuki Tsuneki, Makoto Abe, Fahdi Kanavati
Summary: In this study, deep learning models were trained to classify TUR-P whole-slide images into prostate adenocarcinoma and benign lesions using transfer and weakly supervised learning. The models achieved good classification performance and demonstrated the potential for practical deployment in a TUR-P histopathological diagnostic workflow system to improve efficiency for pathologists.
Article
Engineering, Biomedical
Ming Y. Lu, Drew F. K. Williamson, Tiffany Y. Chen, Richard J. Chen, Matteo Barbieri, Faisal Mahmood
Summary: The CLAM method utilizes attention-based learning to identify subregions with high diagnostic value for accurate classification of whole-slide images. It can localize well-known morphological features without the need for spatial labels, outperforming standard weakly supervised classification algorithms, and adapt to independent test cohorts, smartphone microscopy, and varying tissue content.
NATURE BIOMEDICAL ENGINEERING
(2021)
Article
Medicine, Research & Experimental
Minji Kim, Hiroaki Sekiya, Gary Yao, Nicholas B. Martin, Monica Castanedes-Casey, Dennis W. Dickson, Tae Hyun Hwang, Shunsuke Koga
Summary: Neuropathologic assessment during autopsy is the gold standard for diagnosing neurodegenerative disorders. In this study, a weakly supervised deep learning-based approach called clustering-constrained-attention multiple-instance learning (CLAM) was used to develop a pipeline for diagnosing Alzheimer's disease (AD) and other tauopathies. The multiattention-branch CLAM model achieved the highest diagnostic accuracy for classifying neurodegenerative disorders. This study supports the feasibility of deep learning-based approaches for the classification of neurodegenerative disorders on whole-slide images (WSIs).
LABORATORY INVESTIGATION
(2023)
Article
Computer Science, Interdisciplinary Applications
Richard J. Chen, Ming Y. Lu, Jingwen Wang, Drew F. K. Williamson, Scott J. Rodig, Neal Lindeman, Faisal Mahmood
Summary: This study proposes an interpretable strategy for multimodal fusion of histology image and genomic features for survival outcome prediction. The results on glioma and clear cell renal cell carcinoma datasets demonstrate that this approach improves the prognostic determinations.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2022)
Article
Biochemistry & Molecular Biology
Jana Lipkova, Tiffany Y. Chen, Ming Y. Lu, Richard J. Chen, Maha Shady, Mane Williams, Jingwen Wang, Zahra Noor, Richard N. Mitchell, Mehmet Turan, Gulfize Coskun, Funda Yilmaz, Derya Demir, Deniz Nart, Kayhan Basak, Nesrin Turhan, Selvinaz Ozkara, Yara Banz, Katja E. Odening, Faisal Mahmood
Summary: A deep learning-based AI system has been developed for automated assessment of gigapixel whole-slide images obtained from endomyocardial biopsy, addressing the detection, subtyping and grading of allograft rejection. The system showed non-inferior performance to conventional assessment, reducing interobserver variability and assessment time.
Correction
Computer Science, Artificial Intelligence
Narmin Ghaffari Laleh, Hannah Sophie Muti, Chiara Maria Lavinia Loeffler, Amelie Echle, Oliver Lester Saldanha, Faisal Mahmood, Ming Y. Lu, Christian Trautwein, Rupert Langer, Bastian Dislich, Roman D. Buelow, Heike Irmgard Grabsch, Hermann Brenner, Jenny Chang-Claude, Elizabeth Alwers, Titus J. Brinker, Firas Khader, Daniel Truhn, Nadine T. Gaisa, Peter Boor, Michael Hoffmeister, Volkmar Schulz, Jakob Nikolas Kather
MEDICAL IMAGE ANALYSIS
(2022)
Article
Biology
Vishhvaan Gopalakrishnan, Dena Crozier, Kyle J. Card, Lacy D. Chick, Nikhil P. Krishnan, Erin McClure, Julia Pelesko, Drew F. K. Williamson, Daniel Nichol, Soumyajit Mandal, Robert A. Bonomo, Jacob G. Scott
Summary: A morbidostat is a bioreactor that uses antibiotics to control bacterial growth, making it suitable for studying antibiotic resistance evolution. We present a low-cost morbidostat called the EVolutionary biorEactor (EVE) that can be constructed by students with minimal engineering and programming experience. We validate EVE in a real classroom setting by evolving replicate Escherichia coli populations under chloramphenicol challenge, providing students the opportunity to learn about bacterial growth and antibiotic resistance.
Editorial Material
Biochemistry & Molecular Biology
Ming Y. Lu, Bowen Chen, Faisal Mahmood
Summary: Researchers have utilized pathology data from Twitter to develop a visual-language model that can classify and retrieve histopathology images. This achievement represents a significant milestone in the advancement of multifunctional foundational artificial intelligence models in computational pathology.
Article
Engineering, Biomedical
Jordan Anaya, John-William Sidhom, Faisal Mahmood, Alexander S. Baras
Summary: This study demonstrates that a weakly supervised multiple-instance learning model can encode and aggregate the local sequence context or genomic position of somatic mutations, providing enhanced explainability for sample-level classification. The model achieves best-in-class performance in tumor type classification and microsatellite status prediction, potentially generating biological insight from genomic datasets.
NATURE BIOMEDICAL ENGINEERING
(2023)
Article
Engineering, Biomedical
Richard J. Chen, Judy J. Wang, Drew F. K. Williamson, Tiffany Y. Chen, Jana Lipkova, Ming Y. Lu, Sharifa Sahai, Faisal Mahmood
Summary: This Perspective discusses the impact of algorithmic biases on healthcare disparities in machine learning. Insufficiently fair AI systems can lead to unequal diagnosis, treatment, and billing of patients. The article explores how biases arise in clinical workflows and outlines emerging technologies such as disentanglement, federated learning, and model explainability to mitigate biases in AI-based medical software development.
NATURE BIOMEDICAL ENGINEERING
(2023)
Meeting Abstract
Oncology
Muhammad Shaban, Ming Y. Lu, Drew F. K. Williamson, Richard J. Chen, Jana Lipkova, Tiffany Y. Chen, Faisal Mahmood
Meeting Abstract
Oncology
Richard J. Chen, Drew F. K. Williamson, Ming Y. Lu, Tiffany Y. Chen, Jana Lipkova, Muhammad Shaban, Faisal Mahmood
CANCER EPIDEMIOLOGY BIOMARKERS & PREVENTION
(2023)