NeuroVision-CT: A Five-Class Brain Computed Tomography Dataset for Computer Aided Diagnosis

Description

NeuroVision-CT is a brain Computed Tomography (CT) image dataset developed to support research in artificial intelligence, medical image analysis, deep learning, and computer-aided diagnosis of neurological disorders. The dataset consists of five diagnostic categories: Hydrocephalus, Intracerebral Hemorrhage, Ischemic Stroke, Normal, and Traumatic Skull Fracture. The dataset was collected from Laksam Fair Health Hospital, Laksam, Cumilla, Bangladesh. All CT images were anonymized to protect patient privacy, and no personally identifiable information was retained. Initially, each diagnostic category contained 1,000 images, resulting in a total of 5,000 images. The dataset was divided into training, validation, and test sets using a 70:15:15 ratio. Accordingly, each class contained 700 training images, 150 validation images, and 150 test images. After the dataset split, the 700 training images from each class were augmented to 1,000 images per class. The validation and test sets were kept unchanged, with 150 images per class in each set. Consequently, the final dataset contains 5,000 training images, 750 validation images, and 750 test images, resulting in a total of 6,500 images. All CT images were clinically reviewed and labeled by Dr. Md. Abdul Hannan Mazumder, Emergency Medical Officer (Neurology), Laksam Fair Health Hospital. The images were verified according to their corresponding neurological diagnoses to ensure reliable and consistent class labeling. Dataset Collection Location: Laksam Fair Health Hospital, Laksam, Cumilla, Bangladesh. Submitted Files: NeuroVision-CT Dataset V-1.0 ➜ Training ➜ Validation ➜ Test Metadata: ➜ Metadata_NeuroVision-CT Supporting Documents: ➜ Clinical _Validation_Certificate ➜ Hospital_Permission_Request_Letter ➜ IRB_Approval ➜ IRB_Ethical_Approval_Request_Letter Medical Expert: Dr. Md. Abdul Hannan Mazumder Emergency Medical Officer (Neurology) Laksam Fair Health Hospital Laksam, Cumilla, Bangladesh Email: ahannan.ttt@gmail.com Contact: +8801641726385

Steps to reproduce

1. Collect and anonymize brain CT images from Laksam Fair Health Hospital with required ethical approval. 2. Organize images into five classes: Hydrocephalus, Intracerebral Hemorrhage, Ischemic Stroke, Normal, and Traumatic Skull Fracture. 3. Clinically verify and label the images by a medical expert. 4. Prepare 1,000 images per class and split them into training, validation, and test sets using a 70:15:15 ratio. 5. Augment only the training set from 700 to 1,000 images per class. 6. Keep validation and test sets unchanged and organize the final dataset into Training, Validation, and Test folders.

Institutions

• American International University-Bangladesh

  Dhaka Division, Dhaka
• Daffodil International University

  Dhaka Division, Dhaka

Categories

Licence