Search

Abstract Invasive breast cancer diagnosis and treatment planning require an accurate assessment of human epidermal growth factor receptor 2 (HER2) expression levels. While immunohistochemical techniques (IHC) are the gold standard for HER2 evaluation, their implementation can be resource-intensive and costly. To reduce these obstacles and expedite the procedure, we present an efficient deep-learning model that generates high-quality IHC-stained images directly from Hematoxylin and Eosin (H&E) stained images. We propose a new IHC-GAN that enhances the Pix2PixHD model into a dual generator module, improving its performance and simplifying its structure. Furthermore, to strengthen feature extraction for HE-stained image classification, we integrate MobileNetV3 as the backbone network. The extracted features are then merged with those generated by the generator to improve overall performance. Moreover, the decoder’s performance is enhanced by providing the related features from the classified labels by incorporating the adaptive instance normalization technique. The proposed IHC-GAN was trained and validated on a comprehensive dataset comprising 4,870 registered image pairs, encompassing a spectrum of HER2 expression levels. Our findings demonstrate promising results in translating H&E images to IHC-equivalent representations, offering a potential solution to reduce the costs associated with traditional HER2 assessment methods. We extensively validate our model and the current dataset. We compare it with state-of-the-art techniques, achieving high performance using different evaluation metrics, showing 0.0927 FID, 22.87 PSNR, and 0.3735 SSIM. The proposed approach exhibits significant enhancements over current GAN models, including an 88% reduction in Frechet Inception Distance (FID), a 4% enhancement in Learned Perceptual Image Patch Similarity (LPIPS), a 10% increase in Peak Signal-to-Noise Ratio (PSNR), and a 45% reduction in Mean Squared Error (MSE). This advancement holds significant potential for enhancing efficiency, reducing manpower requirements, and facilitating timely treatment decisions in breast cancer care.

Abstract Polyps are well-known cancer precursors identified by colonoscopy. However, variability in their size, appearance, and location makes the detection of polyps challenging. Moreover, colonoscopy surveillance and removal of polyps are highly operator-dependent procedures and occur in a highly complex organ topology. There exists a high missed detection rate and incomplete removal of colonic polyps. To assist in clinical procedures and reduce missed rates, automated methods for detecting and segmenting polyps using machine learning have been achieved in past years. However, the major drawback in most of these methods is their ability to generalise to out-of-sample unseen datasets from different centres, populations, modalities, and acquisition systems. To test this hypothesis rigorously, we, together with expert gastroenterologists, curated a multi-centre and multi-population dataset acquired from six different colonoscopy systems and challenged the computational expert teams to develop robust automated detection and segmentation methods in a crowd-sourcing Endoscopic computer vision challenge. This work put forward rigorous generalisability tests and assesses the usability of devised deep learning methods in dynamic and actual clinical colonoscopy procedures. We analyse the results of four top performing teams for the detection task and five top performing teams for the segmentation task. Our analyses demonstrate that the top-ranking teams concentrated mainly on accuracy over the real-time performance required for clinical applicability. We further dissect the devised methods and provide an experiment-based hypothesis that reveals the need for improved generalisability to tackle diversity present in multi-centre datasets and routine clinical procedures.

Abstract Polyps in the colon are widely known cancer precursors identified by colonoscopy. Whilst most polyps are benign, the polyp’s number, size and surface structure are linked to the risk of colon cancer. Several methods have been developed to automate polyp detection and segmentation. However, the main issue is that they are not tested rigorously on a large multicentre purpose-built dataset, one reason being the lack of a comprehensive public dataset. As a result, the developed methods may not generalise to different population datasets. To this extent, we have curated a dataset from six unique centres incorporating more than 300 patients. The dataset includes both single frame and sequence data with 3762 annotated polyp labels with precise delineation of polyp boundaries verified by six senior gastroenterologists. To our knowledge, this is the most comprehensive detection and pixel-level segmentation dataset (referred to as PolypGen) curated by a team of computational scientists and expert gastroenterologists. The paper provides insight into data construction and annotation strategies, quality assurance, and technical validation.

Myositis is a rare muscle disorder resulting from the immune system’s attack on the muscle cells, leading to muscle weaknesses. A precise and early diagnosis of the illness is crucial for effective treatment. Differentiating among various types of Myositis remains a significant challenge, leading to false diagnoses and delayed treatment for IBM patients, which can cause muscular atrophy. This paper introduces a new Hybrid computer-aided model that can effectively differentiate among various categories: PolyMyositis (PM), DermatoMyositis (DM), and Inclusion Body Myositis (IBM). A novel method for detecting and classifying inflammatory myopathies from ultrasound images based on a modified lightweight YOLOv5 Architecture is proposed named YOLO-CSE. The proposed model utilizes the Convolutional Block Attention Module (CBAM) to focus on the significant minor extracted features in the image. Moreover, we reuse the extracted features by incorporating the Spatial Pyramid Pooling-Fast Plus (SPPF $^{+}$ ), strengthening the feature extraction capability while providing maximum information flow in the network. Furthermore, the Exponential Linear Unit (ELU) is integrated to activate the function for more accurate results by naming this layer (SPPF $^{+}$ ELU). The performance is tested on large benchmark datasets consisting of 3214 muscle ultrasound images for 80 patients, divided into 14 DermatoMyositis (DM), 14 PolyMyositis (PM), 19 Inclusion Body Myositis (IBM) and 33 Normal patients. Extensive and detailed analysis has been tested on several stages to clarify the role of the CBAM and SPPF Plus modules in our model. Additionally, we analyze the model’s performance through four different suggested classification strategies (binary and multiclass classification). Data Augmentation was proposed to suppress the dataset’s imbalance and boost the model’s performance. Mosaic data augmentation enhanced the dataset by randomly cropping, arranging and scaling images. Various techniques of augmentation were employed, including flipping, cropping, rotating, Contrast Limited Adaptive Histogram Equalization (CLAHE) and Median Blur filters. Experiments have been conducted to demonstrate the effectiveness of the proposed system. The proposed YOLO-CSE achieved highly acceptable accuracy, precision, recall, and F1-Scores for the different strategies. The suggested model achieved an average accuracy for binary and multiclass classification above 98% and average precision above 96%. The results demonstrate that the proposed model accurately detects and classifies inflammatory myopathies.

Early detection of esophageal abnormalities can help in preventing the progression of the disease into later stages. During esophagus examination, abnormalities are often overlooked due to the irregular shape, variable size, and the complex surrounding area which requires a significant effort and experience. In this paper, a novel deep learning model which is based on faster region-based convolutional neural network (Faster R-CNN) is presented to automatically detect abnormalities in the esophagus from endoscopic images. The proposed detection system is based on a combination of Gabor handcrafted features with the CNN features. The densely connected convolutional networks (DenseNets) architecture is embraced to extract the CNN features providing a strengthened feature propagation between the layers and alleviate the vanishing gradient problem. To address the challenges of detecting abnormal complex regions, we propose fusing extracted Gabor features with the CNN features through concatenation to enhance texture details in the detection stage. Our newly designed architecture is validated on two datasets (Kvasir and MICCAI 2015). Regarding the Kvasir, the results show an outstanding performance with a recall of 90.2% and a precision of 92.1% with a mean of average precision (mAP) of 75.9%. While for the MICCAI 2015 dataset, the model is able to surpass the state-of-the-art performance with 95% recall and 91% precision with mAP value of 84%. The experimental results demonstrate that the system is able to detect abnormalities in endoscopic images with good performance without any human intervention.