Dark Mode
Modified Human Sperm Morphology Dataset
Patient Health Records & Digital Health
Tags and Keywords
Trusted By
"No reviews yet"
Free
About
This collection features grayscale images of human sperm intended for machine learning applications in medical diagnostics. The data was gathered from 235 patients dealing with male factor infertility. Experts have meticulously labelled each sperm image to indicate the status of four crucial morphological components: the acrosome, the head, the vacuole, and the tail. The purpose of this data is to advance research in automated sperm morphology analysis, which could ultimately assist in the diagnosis and clinical treatment of male infertility.
Columns
The data is structured as numerical arrays stored in
.npy files, representing both image pixels (x files) and corresponding binary labels (y files). There are no traditional columns as such, but rather key data components:- Image Data (x_ files): Contains the grayscale images of individual sperm cells, available in two resolutions: 128x128 pixels and 64x64 pixels.
- Label Data (y_ files): Contains binary labels (0 for normal, 1 for abnormal) for four separate morphological features: Acrosome, Head, Tail, and Vacuole.
Distribution
The dataset is provided in Python/NumPy format (.npy files) and is partitioned into standard Training, Validation, and Test sets to enable robust model evaluation. The images are grayscale, focusing on a single sperm cell with the head approximately centred, though the sperm tail may not be fully visible in every image.
The image quantities are structured as follows:
- 128x128 resolution: 1000 training images, 240 validation images, and 300 test images.
- 64x64 resolution: 1000 training images, 240 validation images, and 300 test images.
Usage
This data product is suited for several types of automated analysis tasks, including:
- Binary Classification: Developing distinct models to classify the normality or abnormality of each of the four identified morphological features.
- Multi-label Classification: Creating sophisticated models capable of simultaneously predicting the abnormality status across all four features in a single instance.
- Image Analysis: Facilitating research into the visual features most indicative of sperm abnormalities.
- Benchmarking: Serving as a standardised resource for comparing the effectiveness of different computer vision algorithms designed for sperm morphology analysis.
Coverage
The data focuses exclusively on morphological analysis of human sperm samples. The demographic scope includes samples sourced from 235 patients suffering from male factor infertility. Specific geographical origin or time range for the collection is not explicitly detailed.
License
CC BY-NC 3.0
Who Can Use It
This dataset is designed for use by researchers in computer science and biology, particularly those focused on medical imaging, automated diagnostics, and computer vision. Intended users include academics developing new machine learning models for clinical diagnostics and engineers aiming to benchmark image analysis performance.
Dataset Name Suggestions
- MHSMA (Sperm Morphology Analysis Dataset)
- Modified Human Sperm Morphology Dataset
- Automated Male Fertility Diagnosis Images
- Sperm Morphological Abnormality Classifier Data
Attributes
Original Data Source:Modified Human Sperm Morphology Dataset
QUALITYLoading...