Summary

This article presents 582 bone scan images from 291 adult patients who attended the Nuclear Medicine Service at the Instituto de Investigaciones en Ciencias de la Salud (IICS) of the Universidad Nacional de Asunción (UNA), Paraguay, between 2020 and 2024. The images were acquired using triple modality SPECT-CT-PET equipment. Approximately 20 mCi of technetium-99m methylene diphosphonate (99mTc-MDP) was administered to each patient, producing whole-body planar images in anterior and posterior projections of the axial and appendicular skeleton with a resolution of 256 × 1024 pixels. The images were labeled according to the final diagnosis by a nuclear physician, covering conditions ranging from joint lesions to bone metastases. This dataset will be helpful for researchers working on bone scan image analysis using artificial intelligence techniques to classify bone metastases.

Keywords: Bone scan images, SPECT-CT-PET, Nuclear medicine, Bone metastases classification, Artificial intelligence

Results from AnyScan® DUO SPECT/CT/PET

The acquisition of bone scan studies was performed on a trimodal SPECT-CT-PET equipment, AnyScan® DUO SPECT/CT/PET model from MEDISO brand, under the following protocol:

• Patient preparation: Patients are interviewed, their data are recorded in a clinical file and informed consent is obtained. Metal objects are removed from the areas of the body to be scanned. The patient should be well hydrated and urinate frequently before, during and after the study.
• Administration of the radiopharmaceutical: An intravenous dose of 99mTc-MDP, usually about 20 mCi, is administered.
• Biodistribution of the radiopharmaceutical: An uptake time after injection is expected for the radiopharmaceutical to be adequately distributed in the body. This time is variable, but there is usually a wait of between 2 to 4 hours for the study.
• Bone scans image acquisition:
  • Patient positioning: Passive supine decubitus with arms at the sides, and head and feet secured to prevent movement.
  • Collimator: LEHR (Low Energy High Resolution).
  • Energy window: ±20 % centered at 140 keV.
  • Image matrix: 256 × 1024 pixels. ◦ Image representation: 16-bit grayscale.
• Image processing: MEDISO InterView™ XP Processing Software is used to process whole-body images.
• Medical interpretation: A nuclear physician examines images to locate normal or abnormal bone pathologies, such as tumors, fractures, diseases, or other abnormalities.
• Final report: The nuclear physician interprets the bone scan images and generates a detailed report containing findings, diagnoses, and recommendations.
• Data Organization and Storage:
  • Folder Structure: Bone scan images are classified and grouped into individual folders according to classification labels and the image pairs projection type.
  • Labeling and Recording: Each pair of images is classified according to the diagnosed pathologies and recorded according to the order in the Excel file.

Fig. 1. Bone scan images of a breast cancer patient with no bone metastases. In the anterior (left) and posterior (right) views, a homogeneous distribution of the radiotracer is observed in the skeleton, with no areas of abnormal uptake that would indicate bone lesions or metastases.

Fig. 2. Bone scan images of a breast cancer patient with bone metastases. In the anterior (left) and posterior (right) views, focal areas of increased uptake are identified in the skeleton, suggesting the presence of bone lesions compatible with metastasis.

Fig. 3. The Bone Scan Images Dataset.xlsx’ file contains information for each image pair, including patient number, views, image format, acquisition date, demographic details, and diagnostic classification.

Fig. 4. The comparative graph shows the number of patients by classification and year the study was conducted. A higher prevalence is observed in the no bone metastases classification, peaking in 2023.

Conclusion

This dataset is valuable for researching and training professionals in studying pathologies involving bone metastases. Nuclear physicians can significantly enhance their diagnostic abilities by incorporating the bone scan dataset into their professional training, thereby empowering them to confidently detect bone metastases and other skeletal abnormalities. This dataset is a valuable and scalable resource for digital image processing, as well as for the application and training of artificial intelligence models for bone metastases classification.

Original link ScienceDirect