SonoSAMTrack -- Segment and Track Anything on Ultrasound Images
Authors:
Hariharan Ravishankar,
Rohan Patil,
Vikram Melapudi,
Harsh Suthar,
Stephan Anzengruber,
Parminder Bhatia,
Kass-Hout Taha,
Pavan Annangi
Abstract:
In this paper, we present SonoSAMTrack - that combines a promptable foundational model for segmenting objects of interest on ultrasound images called SonoSAM, with a state-of-the art contour tracking model to propagate segmentations on 2D+t and 3D ultrasound datasets. Fine-tuned and tested exclusively on a rich, diverse set of objects from $\approx200$k ultrasound image-mask pairs, SonoSAM demonst…
▽ More
In this paper, we present SonoSAMTrack - that combines a promptable foundational model for segmenting objects of interest on ultrasound images called SonoSAM, with a state-of-the art contour tracking model to propagate segmentations on 2D+t and 3D ultrasound datasets. Fine-tuned and tested exclusively on a rich, diverse set of objects from $\approx200$k ultrasound image-mask pairs, SonoSAM demonstrates state-of-the-art performance on 7 unseen ultrasound data-sets, outperforming competing methods by a significant margin. We also extend SonoSAM to 2-D +t applications and demonstrate superior performance making it a valuable tool for generating dense annotations and segmentation of anatomical structures in clinical workflows. Further, to increase practical utility of the work, we propose a two-step process of fine-tuning followed by knowledge distillation to a smaller footprint model without comprising the performance. We present detailed qualitative and quantitative comparisons of SonoSAM with state-of-the-art methods showcasing efficacy of the method. This is followed by demonstrating the reduction in number of clicks in a dense video annotation problem of adult cardiac ultrasound chamber segmentation using SonoSAMTrack.
△ Less
Submitted 16 November, 2023; v1 submitted 25 October, 2023;
originally announced October 2023.
RISC micrprocessor verification
Authors:
Mitul S Nagar,
Haresh A Suthar,
Chintan Panchal
Abstract:
Today's microprocessors have grown significantly in complexity and functionality. Most of today's processors provide at least three levels of memory hierarchy, are heavily pipelined, and support some sort of cache coherency protocol. These features are extremely complex and sophisticated, and present their own set of unique verification challenges. Verification is clearly not a point tool, but is…
▽ More
Today's microprocessors have grown significantly in complexity and functionality. Most of today's processors provide at least three levels of memory hierarchy, are heavily pipelined, and support some sort of cache coherency protocol. These features are extremely complex and sophisticated, and present their own set of unique verification challenges. Verification is clearly not a point tool, but is part of a process that starts from initial product conception and is to some degrees complete when the product goes to market. Functional verification is necessary to verify the functionality at RTL level. Complex micro-processors like ARM are high performance, low cost and low power 32-bit RISC processors. In our paper complex microprocessor is ARM cortex M3, developed for the embedded applications having low interrupt latency, low gate count, 3- stage pipelining, branch prediction, THUMB and THUMB-2 instruction set. Functional verification is used to verify that the circuit full fills each abstract assertion under the implementation map**. we explore several aspects of processor design, including caches, pipeline depth, ALUs, and bypass logic.The verification was done concurrently with the design implementation of the processor.
△ Less
Submitted 1 September, 2020;
originally announced September 2020.
