This article proposes Orthogonal Time Frequency Space (OTFS)-based In-Phase and Quadrature Index Modulation (IQIM) for multiple-input multiple-output (MIMO) systems, referred to as MOTFS-IQIM, to enhance performance in highmobility channels. MOTFS-IQIM enhances the spectral efficiency (SE) of conventional OTFS-IM by embedding additional index bits through independent activation of the in-phase and quadrature branches. However, this increased bit-embedding capability comes at the cost of exponentially growing detection complexity. When combined with MIMO spatial multiplexing, the joint detection of IQ indices and modulation symbols across multiple antennas further enlarges the search space, making maximumlikelihood (ML) detection computationally prohibitive. To address this challenge, a deep learning (DL)-based detector tailored for MOTFS-IQIM is developed. The proposed network jointly recovers active IQ indices and data symbols while significantly reducing decoding complexity compared with ML detection. Simulations conducted over doubly dispersive channels demonstrate that the MOTFS-IQIM scheme achieves improved SE and biterror-rate performance relative to existing OTFS-IM approaches. Furthermore, the DL-based receiver demonstrates robust performance across various mobility levels and channel conditions, enabling practical and low-complexity implementation of the MOTFS-IQIM system.