Article
Biology
Md Ziaul Hoque, Anja Keskinarkaus, Pia Nyberg, Taneli Mattila, Tapio Seppanen
Summary: Medical image registration and fusion is an effective application for disease tracking and treatment decision-making. However, challenges such as image appearance variations and large image size exist in digital pathology. In this paper, a whole slide image registration algorithm is proposed, which utilizes adaptive smoothing and feature matching to improve matching accuracy.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Multidisciplinary Sciences
Lukasz Roszkowiak, Anna Korzynska, Krzysztof Siemion, Jakub Zak, Dorota Pijanowska, Ramon Bosch, Marylene Lejeune, Carlos Lopez
Summary: This study introduces CHISEL, an end-to-end system for quantitative evaluation of benign and malignant digitized tissue samples with immunohistochemical nuclear staining. It incorporates seamless segmentation based on regions of interest cropping and nuclei cluster splitting, utilizing machine learning and recursive local processing to eliminate distorted outlines. Validation with labeled datasets showed comparable or better results than state-of-the-art methods, with a focus on achieving best results for DAB&H-stained breast cancer tissue samples.
SCIENTIFIC REPORTS
(2021)
Article
Anatomy & Morphology
Flavio Santos da Silva, Natalia Caroline Santos Aquino de Souza, Marcus Vinicius de Moraes, Bento Joao Abreu, Moacir Franco de Oliveira
Summary: The study presents a macro called CmyoSize, which accurately measures the transnuclear cross-sectional size of cardiomyocytes in H&E images. It is a fully automated and standardized method that achieves high precision and is much faster than manual tracing. The results show that CmyoSize is feasible, accurate, and time-efficient for cardiomyocyte size quantification.
ANNALS OF ANATOMY-ANATOMISCHER ANZEIGER
(2022)
Article
Computer Science, Artificial Intelligence
Alfonso Vizcaino, Hermilo Sanchez-Cruz, Humberto Sossa, J. Luis Quintanar
Summary: This study proposes an innovative deep learning model for neural cell counting in densely populated areas, achieving state-of-the-art results, and also presents an improved image treatment method that can enhance the performance of other DL models.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Biology
Amirreza Mahbod, Gerald Schaefer, Benjamin Bancher, Christine Loew, Georg Dorffner, Rupert Ecker, Isabella Ellinger
Summary: This paper introduces CryoNuSeg, the first fully annotated FS-derived cryosectioned and H&E-stained nuclei instance segmentation dataset, containing images from 10 human organs with three manual mark-ups for measuring intraobserver and inter-observer variabilities. The effects of tissue fixation/embedding protocol on automatic nuclei instance segmentation performance are investigated, providing a baseline segmentation benchmark for future research. The dataset and detailed information are available to researchers at https://github.com/masih4/CryoNuSeg.
COMPUTERS IN BIOLOGY AND MEDICINE
(2021)
Article
Gastroenterology & Hepatology
Na Cheng, Yong Ren, Jing Zhou, Yiwang Zhang, Deyu Wang, Xiaofang Zhang, Bing Chen, Fang Liu, Jin Lv, Qinghua Cao, Sijin Chen, Hong Du, Dayang Hui, Zijin Weng, Qiong Liang, Bojin Su, Luying Tang, Lanqing Han, Jianning Chen, Chunkui Shao
Summary: This study developed a deep learning diagnostic model for hepatocellular nodular lesions which improved the histopathologic diagnosis of these lesions, particularly for early hepatocellular carcinoma and risk stratification of patients. The model showed significant advantages in patch-level recognition, especially for fragmentary or scarce biopsy specimens.
Article
Multidisciplinary Sciences
Zahra Mousavi Kouzehkanan, Sepehr Saghari, Sajad Tavakoli, Peyman Rostami, Mohammadjavad Abaszadeh, Farzaneh Mirzadeh, Esmaeil Shahabi Satlsar, Maryam Gheidishahran, Fatemeh Gorgi, Saeed Mohammadi, Reshad Hosseini
Summary: Accurate and early detection of anomalies in peripheral white blood cells is crucial for individual well-being evaluation and diagnosis of hematologic diseases. This study introduces a free access dataset called Raabin-WBC, containing a large number of images of normal white blood cells. The dataset provides diversity for machine learning experiments and tasks in health-related fields.
SCIENTIFIC REPORTS
(2022)
Article
Chemistry, Multidisciplinary
Tasleem Kausar, Adeeba Kausar, Muhammad Adnan Ashraf, Muhammad Farhan Siddique, Mingjiang Wang, Muhammad Sajid, Muhammad Zeeshan Siddique, Anwar Ul Haq, Imran Riaz
Summary: This paper presents a novel machine learning-based model, called SA-GAN, for color stain normalization in histopathology images. The model is trained using the distributions of the entire dataset and the color statistics of a single target image. Evaluation results on four different histopathology datasets demonstrate the effectiveness of SA-GAN in acclimating stain contents and enhancing normalization quality. Additionally, the proposed method achieves a 6.9% improvement in accuracy for multiclass cancer type classification.
APPLIED SCIENCES-BASEL
(2022)
Article
Medicine, General & Internal
Min Du, Yu-Meng Cai, Yu-Lei Yin, Li Xiao, Yuan Ji
Summary: This study aims to evaluate the prognostic significance of tumor-infiltrating lymphocytes (TILs) in H&E-stained slides of hepatocellular carcinoma (HCC). The results indicate that HCC patients with high infiltrating lymphocytes tend to have a lower recurrence rate and less microvascular invasion.
WORLD JOURNAL OF CLINICAL CASES
(2022)
Article
Optics
Jiansheng Wang, Xintian Mao, Yan Wang, Xiang Tao, Junhao Chu, Qingli Li
Summary: Artificial intelligence is widely used in digital pathology diagnosis, but it heavily relies on high-quality annotated datasets, which are time-consuming and expensive to create. This paper proposes a new strategy using hyperspectral images to generate annotated pathological benchmark datasets. The method uses a Spatial-Spectral Hyperspectral GAN to transform hyperspectral images into standard histological images, and combines gradient boosting decision tree and graph-cut method to automatically generate annotations. The results show promising performance in generating completely annotated pathology benchmark datasets.
OPTICS AND LASER TECHNOLOGY
(2023)
Article
Biology
Shyam Lal, Devikalyan Das, Kumar Alabhya, Anirudh Kanfade, Aman Kumar, Jyoti Kini
Summary: This study presents NucleiSegNet, a deep learning network architecture for nuclei segmentation in H&E stained liver cancer histopathology images. The architecture includes robust residual block, bottleneck block, and attention decoder block, and outperforms state-of-the-art methods in nuclei segmentation tasks. Additionally, a new liver dataset (KMC liver dataset) was introduced for further research.
COMPUTERS IN BIOLOGY AND MEDICINE
(2021)
Article
Biology
Iqra Kiran, Basit Raza, Areesha Ijaz, Muazzam A. Khan
Summary: Cancer is the second deadliest disease globally, and early detection is crucial. This study proposes a model for nuclei segmentation, combining U-Net and DenseRes-Unet to effectively extract features. The distance map and binary threshold techniques enhance the interior and contour information of nuclei in the images. Evaluation on publicly available datasets shows that the proposed model achieves high accuracy and performance.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Engineering, Biomedical
Emrah Hancer, Mohamed Traore, Refik Samet, Zeynep Yildirim, Nooshin Nemati
Summary: A key step in computational pathology is to automate the manual nuclei segmentation process in H&E stained whole slide images. This paper introduces an imbalance-aware nuclei segmentation methodology that outperforms recent studies in terms of evaluation metrics on histopathology images.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2023)
Article
Computer Science, Artificial Intelligence
Pengshuai Yang, Xiaoxu Yin, Haiming Lu, Zhongliang Hu, Xuegong Zhang, Rui Jiang, Hairong Lv
Summary: This paper proposes a hybrid self-supervised learning method CS-CO tailored for histopathological images. It makes good use of domain-specific knowledge and requires no side information, showing good rationality and versatility. Experimental results demonstrate the effectiveness and robustness of CS-CO on common computational histopathology tasks, and ablation studies prove the complementarity and enhancement of cross-staining prediction and contrastive learning in CS-CO.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Chemistry, Multidisciplinary
Ludovic Venet, Sarthak Pati, Michael D. Feldman, MacLean P. Nasrallah, Paul Yushkevich, Spyridon Bakas
Summary: The proposed two-step diffeomorphic registration method accurately aligns differently stained histology slices and was evaluated on a diverse dataset, demonstrating robustness and accuracy while maintaining computational efficiency. This approach shows promise for enhancing histological image alignment and understanding tissue structures.
APPLIED SCIENCES-BASEL
(2021)
