The manuscript by Misra et al reports the development and initial testing of a deep learning-based multimodal fusion network to segment and classify breast masses using B-mode and elastography ultrasound images.

The authors propose a novel technique to segment and classify multimodal, color-coded imaging information. The newly proposed approach is applied to a retrospective, single-center dataset of 72 patients and compared to previous techniques in breast ultrasound segmentation/ classification. The authors conclude their model to be highly accurate with a sensitivity of 100% and specificity of 94.28% in the test set and thus, may help to improve breast cancer diagnosis.

Although the sample size is low, this is a timely, interesting, and well-written article. The newly proposed deep learning-based multimodal fusion network holds potential to improve breast cancer diagnosis by incorporating multi-modal imaging information. However, some issues need to be clarified:

Abstract
1) Please add absolute numbers for sensitivity and specificity as well as some measurement of uncertainty

Methods
2) Classification: Which probability threshold did you use for the classification and how did you determine this?

Results
3) Please provide absolute numbers alongside the percentages as well as measurements of uncertainty
4) Table 3 compared the authors’ proposed method to previous models. The authors conclude that “ that the performance of the proposed method is significantly improved than other single-modality models and other state-of-the-art multi-modal algorithms”. However, no statistical testing results are presented.
5) Please provide some clinical characteristics of the 72 patients to evaluate representativeness.

Discussion:
6) The low sample size, especially for a deep learning image analysis, needs to be honestly discussed with all its implications on model generalizability (single center dataset, retrospective testing, need for larger, prospective validation). This also needs be mentioned in the abstract

The authors clarified all previous issues.