Osteolysis: A Literature Review of Basic Science and Potential Computer-Based Image Processing Detection Methods

Table 2

The names and properties of the potential datasets that can be used for training deep learning neural networks for bone tumor detection and segmentation. Each dataset is also labeled as either public or private.

Name of the dataset	Description	Type of the data	Number of images
CNUH [78] (private)	Provided by the Chonnam National University Hospital (CNUH)	CT datasets focusing on benign and malignant tumors in two regions of knee bone of distal femur and proximal tibia	Benign tumor: 1061, malignant tumor: 134, normal: 381
Shenzhen No. 2 People’s Hospital [71] (private)	Derived from patients diagnosed with bone tumors in the years 2014–2017	CT images stored in DICOM	6422 images
Hokkaido University [67] (private)	Provided by the Hokkaido University in Japan	CT images of metastatic and nonmetastatic tumor images	8790 images
Sichuan University [94] (private)	Provided by the Institutional Ethics Committee of West China Hospital in Sichuan University	Bone scintigraphy (BS) images using SPECT/CT taken from patients diagnosed with bone metastasis and having undergone whole-body BS	13477 images
Public datasets [95] (public)	CT images corresponding to different body parts	CT images manually segmented	270 images
In-house dataset [96] (private)	Containing bone images of different body parts: head, chest, abdomen, neck	CT images	16218 images
Bone tumor [97] (private)	Containing benign and malignant bone tumors	Plain radiography	2899 images
CNUH [98] (private)	Provided by the Chonnam National University Hospital initially for bone segmentation on a deep learning approach containing both malignant and benign tumors	Plain radiography	963 total images, benign tumor: 329, malignant tumor: 134
DIAGNOSTIKO IATRIKI A.E. [63] (private)	Collected from prostate cancer patients with suspected bone metastatic disease, who underwent whole-body scintigraphy	Scintigraphy images (SPECT)	908 total images, 778 bone scan, 328 bone metastasis, 271 benign scan, 179 normal
Chittagong University [99] (private)	Consisting of 60 MRI images of patients diagnosed with bone cancer and their grand truth images	MRI images	60 total images, benign tumor: 30, malignant tumor: 30