Computational ultrasound imaging has become a powerful tool in technical and medical applications. It enables visualization through complex scattering media such as skull bone or engineered waveguides. In this context, waveguides for ultrasound imaging in hot melts are necessary to shield the transducers from the hot environment. Published works on the topic of ultrasonic testing usually deal with single-mode waveguides. These waveguides support only a single mode within the submerged environment; therefore, imaging can be performed only by time-consuming point-by-point mechanical scanning. Instead of a single-mode waveguide, a multimode waveguide supports several modes and was proposed for real-time ultrasound imaging without the need for 2-D or 3-D scanning. Meanwhile, the use of multimode waveguides presents several challenges, such as modal dispersion that distorts wavefronts and scrambles spatial information, for which additional signal processing schemes are necessary to compensate for them. The goal of the present work is to create an optimization model for multimodal waveguide-based ultrasound imaging in hot melts. This model allows for a tradeoff between the number of reflected signals returned from the scatterer, attenuation of waves, and the desired bulk wave types, which is governed by the selected depth between the bottom surface of the waveguide and the scatterer position. Therefore, the optimal depth will always depend on the respective fluid-waveguide combination. First, we compute and visualize the paths that return from different positions of the scatterer immersed in the hot melts using a numerical method. Second, the optimal angles of the guided modes in the waveguide are calculated based on the ordinary differential equation (ODE) method and Snell-Descartes law. Accordingly, the guided and bulk waves propagating within the waveguide can be distinguished, where a classification into six types of signals is introduced. Experimental results carried out confirm the feasibility of the proposed optimization model, which achieves a tradeoff between the number of reflected signals returned from the scatterer, attenuation of waves, and the desired bulk wave types to maintain a reasonable signal-to-noise ratio (SNR) value. The present measurement system achieves 0.27 mm axial resolution and 25 ms temporal resolution. The unique advantage of the present physics model of multimode waveguide-based ultrasound imaging is its ability to improve SNR, as well as it can be transferred to several applications such as skull bone imaging in the medical field, where ultrasound imaging of the brain is obstructed by the aberrating skull bone, as well as in energy applications.