Organ segmentation from computed tomography images using the 3D convolutional neural network: a systematic review

Computed tomography images are scans that combine a series of X-rays with computer processing techniques to display organs in the body. Recently, 3D CNN models have become effective in tasks relating to recognition, delineation, and classification. Therefore we propose a review to summarize different 3D CNN algorithms for segmenting organs in computed tomography images. This work systematically applies a two-stage procedure for review. A thorough screening of abstracts and titles to ascertain their relevance was done. Research papers published in the academic repositories were selected, analyzed, and reviewed. Insight relating to 3D organ segmentation is provided, with content such as database usage, disadvantages, and advantages. A comparison of two accuracies was carried out with a graph depicting database categories. Important insights, limitations, observations, and future directions were elucidated. After careful investigation, we observe that the encoder-decoder network is predominant for segmentation. The encoder-decoder framework provides a seamless procedure to segment CT images. A prediction of future trends with insightful recommendations for researchers is proposed. Finally, findings suggest that CNN algorithms produce good accuracies despite their limitations.