Spin-echo multi-spectral imaging (MSI) sequences can overcome the severe susceptibility artifacts caused by metallic hardware [1,2]. While proton density, T₁, T₂, and inversion recovery MSI sequences are now used clinically, diffusion weighting remains challenging. In the presence of metallic hardware, conventional single-shot EPI diffusion sequences experience severe geometric distortions. While these artifacts can be mitigated using FSE-based MSI sequences, the addition of diffusion lobes to an FSE sequence violates the CPMG condition, resulting in a rapid decay of the amplitude of the echo train [3]. In conjunction with a PROPELLER acquisition, echo train stability can be improved by modulating the phase of the refocusing pulses along the X-Y axes [3,4]. Recently, this method was used to create diffusion-weighted images around metal [5]. Alternatively, this stability can also be improved by separating the spin and stimulated echoes [6, 7], which can be efficiently accomplished using the PROPELLER DUO acquisition strategy [8]. Given its scan-time benefits, the work presented here uses the DUO acquisition for the clinical translation of 2D-MSI diffusion-weighted imaging. The PROPELLER-DUO sequence was extended to perform MSI metal artifact imaging by dynamically changing transmit and receive frequencies of the RF pulses to sample multiple spectral-bins. Additionally, the amplitude of the slice-select gradient was flipped relative to the refocusing gradient to excite spatially and spectrally selective bins – which then becomes the 2D MSI approach [9]. To enable additional phase correction of odd and even echoes within the train, alternating echoes were split into orthogonal blades [4,10]. While the split-echo approach can both be acquired using an elongated readout [7], in PROPELLER-DUO they are acquired using two separate readout lobes and split into orthogonal blades. Diffusion weighting was achieved using unipolar trapezoidal gradients around the refocusing pulse. The final pulse sequence is shown in the Figure 1 below. In-vivo data was acquired in subjects with sarcoma in close proximity to metallic hardware. Utilized sequence parameters are: 128 points, ETL = 24 – 32, TE/TR = 55/600-1500ms, BW = 62.5 kHz, NEX = 8 – 10, b-value = 100 – 300 s/mm2, single spin-echo diffusion preparation, centric-blade phase encoding, sinc RF pulse bandwidth = 1.2 kHz, MSI spectral bins = 8 – 10, spectral bin separation = 600 Hz. The PROPELLER raw-data for every bin was reconstructed as detailed by Pipe et al. [3]. Individual bins were subject to a wavelet de-noising procedure and combined in a sum of squares fashion to yield the final artifact minimized 2D-MSI image. With a DUO acquisition mode, the echo train remained largely stable. However, as the spin and stimulated echoes are separated, this mode suffers a ~50% reduction in each echo amplitude. The split-echo mode was insensitive to non-CPMG effects and the images showed few FSE ghosts. Figure 2A shows the different spectral bins acquired for a single-slice, and the corresponding diffusion-weighted images. These different spectral bin images were combined to yield the MSI sum-of-squares (SOS) images shown in Figure 2B. Figure 3 shows the current gold-standard clinical diffusion-weighted image, and the lesion is partially obscured by the artifact caused by the metal implant. Figure 4 shows multiple slices of the final MSI images obtained in a subject with lesion near a total hip replacement. Both subjects had bone lesions proximal to the metallic hardware that show an elevated ADC, suggesting possible necrosis [11]. The 2D MSI approach substantially reduced susceptibility artifacts, allowing robust diffusion weighting around metal implants. Historically, while the XY modulation approach was stable for larger flip-angles, typical MSI acquisitions operate at lower flip-angles (~110°) and also encounter substantial B₁ inhomogeneity near metal implant interfaces. Hence, in spite of a loss in signal amplitude, the echo train stability offered by the DUO mode made­ it the preferred choice for MSI-based diffusion imaging. Additionally, as the DUO mode acquires two echoes for every refocusing pulse, the resulting acquisition time was reduced. However, this scan-time improvement was partially negated with the need for multiple averages to improve the SNR. As a result, the acquisition time was typically between 4 - 7 min. To shorten the scan time, it is anticipated that future clinical implementations of this approach will significantly benefit from simultaneous multi-slice imaging. Additionally, this scan time can be shortened further, by calibrating the spectral coverage needed for every subject [12].