This paper proposes an advanced THz multipleinput-multiple-output (MIMO) near-field sparse imaging scheme for target detection that uses single-pass synthetic aperture focusing and multi-pass interferometric synthetic aperture focusing techniques to improve imaging effectiveness and efficiency. It benefits from the MIMO of linear sparse periodic array (SPA) and random sparse imaging to reduce sampling data as well as system cost. Both simulated and proof-of-concept experimental results have verified the scheme, revealing that the single-pass synthetic aperture imaging approach is sufficient to identify pure metallic targets with 3.87% of λ/2 sampling data. And multi-pass interferometric synthetic aperture imaging approach is capable of improving image quality on image signal to noise ratio and contrast, which is ideal for detecting more challenging targets. The random sparse imaging with help of low rank matrix completion (LRMC) technique has shown promising potential of achieving equivalent image quality when further reducing 43.75% data in the experiment of 5-pass imaging on a complex target, this corresponds to 82.51% data reduction compared to λ/2 sampling.