Comparing Class-Aware and Pairwise Loss Functions for Deep Metric Learning in Wildlife Re-Identification.

Similarity learning using deep convolutional neural networks has been applied extensively in solving computer vision problems. This attraction is supported by its success in one-shot and zero-shot classification applications. The advances in similarity learning are essential for smaller datasets or datasets in which few class labels exist per class such as wildlife re-identification. Improving the performance of similarity learning models comes with developing new sampling techniques and designing loss functions better suited to training similarity in neural networks. However, the impact of these advances is tested on larger datasets, with limited attention given to smaller imbalanced datasets such as those found in unique wildlife re-identification. To this end, we test the advances in loss functions for similarity learning on several animal re-identification tasks. We add two new public datasets, Nyala and Lions, to the challenge of animal re-identification. Our results are state of the art on all public datasets tested except Pandas. The achieved Top-1 Recall is 94.8% on the Zebra dataset, 72.3% on the Nyala dataset, 79.7% on the Chimps dataset and, on the Tiger dataset, it is 88.9%. For the Lion dataset, we set a new benchmark at 94.8%. We find that the best performing loss function across all datasets is generally the triplet loss; however, there is only a marginal improvement compared to the performance achieved by Proxy-NCA models. We demonstrate that no single neural network architecture combined with a loss function is best suited for all datasets, although VGG-11 may be the most robust first choice. Our results highlight the need for broader experimentation and exploration of loss functions and neural network architecture for the more challenging task, over classical benchmarks, of wildlife re-identification.

A robust model training strategy using hard negative mining in a weakly labeled dataset for lymphatic invasion in gastric cancer.

Gastric cancer is a significant public health concern, emphasizing the need for accurate evaluation of lymphatic invasion (LI) for determining prognosis and treatment options. However, this task is time-consuming, labor-intensive, and prone to intra- and interobserver variability. Furthermore, the scarcity of annotated data presents a challenge, particularly in the field of digital pathology. Therefore, there is a demand for an accurate and objective method to detect LI using a small dataset, benefiting pathologists. In this study, we trained convolutional neural networks to classify LI using a four-step training process: (1) weak model training, (2) identification of false positives, (3) hard negative mining in a weakly labeled dataset, and (4) strong model training. To overcome the lack of annotated datasets, we applied a hard negative mining approach in a weakly labeled dataset, which contained only final diagnostic information, resembling the typical data found in hospital databases, and improved classification performance. Ablation studies were performed to simulate the lack of datasets and severely unbalanced datasets, further confirming the effectiveness of our proposed approach. Notably, our results demonstrated that, despite the small number of annotated datasets, efficient training was achievable, with the potential to extend to other image classification approaches used in medicine.

PseudoCell: Hard Negative Mining as Pseudo Labeling for Deep Learning-Based Centroblast Cell Detection.

Background: Deep learning models for patch classification in whole-slide images (WSIs) have shown promise in assisting follicular lymphoma grading. However, these models often require pathologists to identify centroblasts and manually provide refined labels for model optimization. Objective: To address this limitation, we propose PseudoCell, an object detection framework for automated centroblast detection in WSI, eliminating the need for extensive pathologist's refined labels. Methods: PseudoCell leverages a combination of pathologist-provided centroblast labels and pseudo-negative labels generated from undersampled false-positive predictions based on cell morphology features. This approach reduces the reliance on time-consuming manual annotations. Results: Our framework significantly reduces the workload for pathologists by accurately identifying and narrowing down areas of interest containing centroblasts. Depending on the confidence threshold, PseudoCell can eliminate 58.18-99.35% of irrelevant tissue areas on WSI, streamlining the diagnostic process. Conclusion: This study presents PseudoCell as a practical and efficient prescreening method for centroblast detection, eliminating the need for refined labels from pathologists. The discussion section provides detailed guidance for implementing PseudoCell in clinical practice.

Bone Metastases Lesion Segmentation on Breast Cancer Bone Scan Images with Negative Sample Training.

The use of deep learning methods for the automatic detection and quantification of bone metastases in bone scan images holds significant clinical value. A fast and accurate automated system for segmenting bone metastatic lesions can assist clinical physicians in diagnosis. In this study, a small internal dataset comprising 100 breast cancer patients (90 cases of bone metastasis and 10 cases of non-metastasis) and 100 prostate cancer patients (50 cases of bone metastasis and 50 cases of non-metastasis) was used for model training. Initially, all image labels were binary. We used the Otsu thresholding method or negative mining to generate a non-metastasis mask, thereby transforming the image labels into three classes. We adopted the Double U-Net as the baseline model and made modifications to its output activation function. We changed the activation function to SoftMax to accommodate multi-class segmentation. Several methods were used to enhance model performance, including background pre-processing to remove background information, adding negative samples to improve model precision, and using transfer learning to leverage shared features between two datasets, which enhances the model's performance. The performance was investigated via 10-fold cross-validation and computed on a pixel-level scale. The best model we achieved had a precision of 69.96%, a sensitivity of 63.55%, and an F1-score of 66.60%. Compared to the baseline model, this represents an 8.40% improvement in precision, a 0.56% improvement in sensitivity, and a 4.33% improvement in the F1-score. The developed system has the potential to provide pre-diagnostic reports for physicians in final decisions and the calculation of the bone scan index (BSI) with the combination with bone skeleton segmentation.

Nucleotide-level prediction of CircRNA-protein binding based on fully convolutional neural network.

Introduction: CircRNA-protein binding plays a critical role in complex biological activity and disease. Various deep learning-based algorithms have been proposed to identify CircRNA-protein binding sites. These methods predict whether the CircRNA sequence includes protein binding sites from the sequence level, and primarily concentrate on analysing the sequence specificity of CircRNA-protein binding. For model performance, these methods are unsatisfactory in accurately predicting motif sites that have special functions in gene expression. Methods: In this study, based on the deep learning models that implement pixel-level binary classification prediction in computer vision, we viewed the CircRNA-protein binding sites prediction as a nucleotide-level binary classification task, and use a fully convolutional neural networks to identify CircRNA-protein binding motif sites (CPBFCN). Results: CPBFCN provides a new path to predict CircRNA motifs. Based on the MEME tool, the existing CircRNA-related and protein-related database, we analysed the motif functions discovered by CPBFCN. We also investigated the correlation between CircRNA sponge and motif distribution. Furthermore, by comparing the motif distribution with different input sequence lengths, we found that some motifs in the flanking sequences of CircRNA-protein binding region may contribute to CircRNA-protein binding. Conclusion: This study contributes to identify circRNA-protein binding and provides help in understanding the role of circRNA-protein binding in gene expression regulation.

Lesion-Based Bone Metastasis Detection in Chest Bone Scintigraphy Images of Prostate Cancer Patients Using Pre-Train, Negative Mining, and Deep Learning.

This study aimed to explore efficient ways to diagnose bone metastasis early using bone scintigraphy images through negative mining, pre-training, the convolutional neural network, and deep learning. We studied 205 prostate cancer patients and 371 breast cancer patients and used bone scintigraphy data from breast cancer patients to pre-train a YOLO v4 with a false-positive reduction strategy. With the pre-trained model, transferred learning was applied to prostate cancer patients to build a model to detect and identify metastasis locations using bone scintigraphy. Ten-fold cross validation was conducted. The mean sensitivity and precision rates for bone metastasis location detection and classification (lesion-based) in the chests of prostate patients were 0.72 ± 0.04 and 0.90 ± 0.04, respectively. The mean sensitivity and specificity rates for bone metastasis classification (patient-based) in the chests of prostate patients were 0.94 ± 0.09 and 0.92 ± 0.09, respectively. The developed system has the potential to provide pre-diagnostic reports to aid in physicians' final decisions.