Soil, a critical Earth resource, sustains ecosystems and global food production, serving as a habitat, regulating water, sequestering carbon, and supplying nutrients. Roots play a crucial role in the composition and health of soil. Soil properties and root distribution data provide essential information for land management and agriculture. In this study, we propose an innovative approach combining x-ray computed tomography (CT) scanning, machine learning-based root segmentation, and traditional root analysis methods to investigate plant root distribution comprehensively. Intact soil cores with plant roots were CT scanned to visualize root systems in their natural soil environment. Utilizing a UNET transformer (UNETR) machine learning framework, we achieved automated root segmentation, extracting and differentiating roots from the surrounding soil. Validation against traditional analysis with WinRHIZO and RhizoVision Explorer for root trait measurement showed a strong positive correlation (up to 0.78 Pearson coefficient), affirming the precision of our machine learning method in quantifying root characteristics. This integration of CT scanning and machine learning-based root segmentation provides a non-destructive and efficient method for studying root architecture and distribution. Our research highlights the potential of combining advanced imaging techniques with AI to enhance the understanding of root dynamics and their role in supporting plant growth. The proposed methodology offers a promising toolset for automated root analysis, reducing manual processing time and effort. By shedding light on root-soil interactions, our study contributes to the field of plant root phenotyping and provides valuable insight into the complex world of below-ground plant systems, aligning with scalable and cost-effective monitoring techniques and innovations in remote-sensing-based soil monitoring frameworks