BackgroundUrogenital schistosomiasis is considered a Neglected Tropical Disease (NTD) by the World Health Organization (WHO). It is estimated to affect 150 million people worldwide, with a high relevance in resource-poor settings of the African continent. The gold-standard diagnosis is still direct observation of Schistosoma haematobium eggs in urine samples by optical microscopy. Novel diagnostic techniques based on digital image analysis by Artificial Intelligence (AI) tools are a suitable alternative for schistosomiasis diagnosis.MethodologyDigital images of 24 urine sediment samples were acquired in non-endemic settings. S. haematobium eggs were manually labeled in digital images by laboratory professionals and used for training YOLOv5 and YOLOv8 models, which would achieve automatic detection and localization of the eggs. Urine sediment images were also employed to perform binary classification of images to detect erythrocytes/leukocytes with the MobileNetv3Large, EfficientNetv2, and NasNetLarge models. A robotized microscope system was employed to automatically move the slide through the X-Y axis and to auto-focus the sample.ResultsA total number of 1189 labels were annotated in 1017 digital images from urine sediment samples. YOLOv5x training demonstrated a 99.3% precision, 99.4% recall, 99.3% F-score, and 99.4% mAP0.5 for S. haematobium detection. NasNetLarge has an 85.6% accuracy for erythrocyte/leukocyte detection with the test dataset. Convolutional neural network training and comparison demonstrated that YOLOv5x for the detection of eggs and NasNetLarge for the binary image classification to detect erythrocytes/leukocytes were the best options for our digital image database.ConclusionsThe development of low-cost novel diagnostic techniques based on the detection and identification of S. haematobium eggs in urine by AI tools would be a suitable alternative to conventional microscopy in non-endemic settings. This technical proof-of-principle study allows laying the basis for improving the system, and optimizing its implementation in the laboratories.

This data was collected from the distribution range of mangroves along the coast of China and has been randomly selected and manually corrected to be labeled as a crab image analysis dataset. The weights of deep learning models trained on YOLOv5/v8 and EfficientNet are also uploaded simultaneously. Additionally, it includes the necessary test datasets and some test results. Please cite when using this dataset, and contact the administrator if you need help. The specific directories are as follows:

- R-crab: Contains code needed to test model performance, with the subfolder data containing the test dataset required. It includes (1) crab-man as human-marked references, crab-ref as model detection results. (2) luoyuan-burrow for the detection results of crab burrows in the Luoyuan area case study, and luoyuan-crab for crab detection results, including crab classification, localization, and carapace width information. (3) method-test for testing different methods, i.e., whether to use a two-stage detection model. (4) size-conf-test records the model detection results under different image input sizes and confidence threshold levels. (1), (3), and (4) are completed in Out-of-sample data, while (2) is completed in Luoyuan. fig_attr.csv records the test results of image attributes on detection accuracy. label_results.csv records the comparison results between traits measured manually using ImageJ software and our designed model for detecting crab carapace width. luoyuan_list.csv records the numbering information of Luoyuan sampling plots.
- Aiweights: Contains model weights trained based on Object detection data, with cpm-model under v8n-seg-crab.pt for YOLOv8 trained to extract crab carapace width. Sfc-model under adam20.pth is a two-stage detection model trained based on EfficientNet. Trained_weights under burrow-baseline is a crab burrow detection model trained based on our improved YOLOv5 (improvements stored at https://github.com/GuuX29/crab-yolo-add, same below); frame-s6-2560.pt is a plot frame detection model for obtaining standard 50*50cm plot images; s-simam-20.pt and x-simam-20.pt are both for crab detection and classification models, with x having higher accuracy.
- Luoyuan: crop stores standardized processed Luoyuan image data, numbered as above, test stores detection results, same as R-crab.
- Object detection: Contains cropped 640 pixels crab field sampling images, with images in the images folder and bounding box labels in the labels folder, divided into training and testing at a 9:1 ratio, train.txt and val.txt record the allocation information.
- Out-of-sample: Records 100 independent test images not used to train the model, stored in images, crab-man, and crab-ref are the same as in R-crab.
- Segmentation: Stores image data used to test model detection of crab carapace width, with results also stored in R-crab.

We hope this data and method will benefit the progress of research in this field. For any suggestions for improvement and help with usage, please contact the administrator. guuxuan1994@gmail.com