Human Bone Fractures Multi-modal Image Dataset (HBFMID)

Human Bone Fractures Multi-modal Image Dataset (HBFMID) consists of X-ray and MRI images including the elbow, finger, forearm, humerus, shoulder, femur, shinbone, knee, hipbone, wrist, spinal cord, and some of the healthy bones. The dataset primarily used 641 raw images, including 510 for X-ray and 131 for MRI images. After that, we divided our dataset into three sections: training (449 images), validation(128 images), and testing (64 images). We used annotation techniques to label the image data. The auto-orient, resized the image size to 640 X 640, and made contrast adjustments have been applied for image pre-processing. For augmentation, several techniques like scaling, shifting, shearing, cropping, random rotation, and brightness change are applied. In the augmentation technique, the following parameters were used: Flip (Horizontal, Vertical), Rotation (Between -5◦and +5◦), Shear (±2◦Horizontal, ±2◦Vertical), Saturation (Between -5 % and +5 %), Brightness: (Between -10 % and +10 %), and zooming (2%). After augmentation, we received an augmented image comprising about 1347 (449x3) images. In HBFMID dataset a total of 1539 ((1347:training)+(128: validation)+(64: testing)) images are used.