SonoRank: calibration-free real-time finger flexion detection from forearm ultrasound sequences

Jul 16, 2026

Powered prosthetic hands are frequently rejected due to limited functionality in devices relying on surface electromyography (sEMG). Sonomyography (ultrasound) offers real-time observation of muscle activity and the potential to control more degrees of freedom, but most existing methods require per-user fine-tuning. SonoRank targets calibration-free finger flexion detection using forearm ultrasound video by learning pairwise sequence ranking per finger and then fine-tuning with a rest reference. On 12-fold leave-one-subject-out validation across twelve subjects, SonoRank improves F1 by 28% over direct classification baselines, supporting subject-independent control and practical deployment.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages8

Reading Time20 min

Introduction

Limitations of sEMG-based prosthetic control

Promise and constraints of ultrasound (sonomyography)

Research question and proposed approach

SonoRank Framework

Stage 1: Pairwise ranking of ultrasound sequences

Stage 2: Fine-tuning for per-finger flexion classification

Real-time operation overview

SonoRank: Towards Calibration-Free Real-Time Finger Flexion Detection from Forearm Ultrasound Sequences Dean Zadok1 , Alon Wolf2 , Alex M. Bronstein 1 , Oren Salzman 1 arXiv :2607 .07542v 1 [ cs .RO] 8 Jul 2026 Abstract—Powered prosthetic hands are frequently abandoned, largely due to the limited functionality of current devices that rely on surface electromyography (sEMG). Sonomyography (ultrasound) has emerged as a promising alternative, owing to its ability to observe muscle activity in real time and control a greater number of degrees of freedom. Yet, existing ultrasound-based methods require per-user fine-tuning, limiting their commercialization. We propose SonoRank, an important step towards calibration-free finger flexion detection from forearm ultrasound video. SonoRank first learns to rank pairs of ultrasound sequences by their relative motion magnitude for each of the five fingers. The learned representations are then fine-tuned to classify whether each finger is actively flexing, using a rest reference that is captured at the beginning of the operation. Under 12-fold leave-one-subject-out crossvalidation on a dataset of twelve subjects with synchronized kinematics, SonoRank achieves a 28% improvement in F1 score over direct classification baselines that skip the ranking stage. These results establish pairwise ranking as an effective pretraining signal for subject-independent control, bringing ultrasound-based prosthetics closer to practical, calibration-free deployment. I. INTRODUCTION Six decades of prosthetics research have transformed artificial hands from simple grippers into mechatronic systems. Modern designs offer dozens of independent degrees of freedom, approaching human-level dexterity [1] . Yet 44% of upper-limb amputees reject their prostheses [2], suggesting a gap in the control interface. Since the mid-twentieth century, surface electromyography (sEMG) has remained the dominant non-invasive sensing modality for prosthetic hands [3], alongside plug-and-play interfaces [4] . However, sEMG captures only noisy electrical activity from superficial muscle fibers, making individual finger discrimination difficult [5] . Montagnani et al. showed that it is precisely finger dexterity that current prostheses lack [6], underscoring that the bottleneck lies mostly in sensing. Ultrasound imaging of the forearm (i.e., sonomyography) offers a richer window into motor intent [7] . B-mode (brightness-mode) images, the grayscale cross-sections produced by conventional ultrasound, capture the spatial structure of deep muscles in real time, enabling the discrimination of finger movements from morphological deformations that sEMG cannot resolve [8], [9] . Yet existing sonomyography methods frame finger motion prediction as either discrete gesture classification [10], [11] or direct joint-angle regres- 1Department of Computer Science, Technion, Haifa, Israel {deanzadok,bron,[osalzman}@cs.technion.ac.il](osalzman}@cs.technion.ac.il) 2Department of Mechanical Engineering, Technion, Haifa, Israel [alonw@me.technion.ac.il](alonw@me.technion.ac.il) | | | --- | | | | | Fig. 1: Ultrasound can see what EMG cannot: B-mode cross-sections of the forearm (top of each image is the skin surface) at rest (top) and during ring-finger flexion (bottom), with the corresponding hand pose on the right. Muscle regions and bones were manually annotated for illustration purposes: the flexor digitorum superficialis (FDS) and profundus (FDP), the two extrinsic muscles that actuate the four fingers, are highlighted alongside the Ulna and Radius (blue) . The dashed circle highlights the boundary between the FDS and FDP. During flexion, this region visibly deforms as the muscles contract and shift relative to each other. This morphological change is the signal that our method is designed to detect. sion [9], [12] . Classification restricts output to a fixed gesture vocabulary, while regression demands subject-specific calibratio

FAQs

What problem does SonoRank address in prosthetic hand control?

SonoRank addresses the limited finger functionality and poor practicality of prosthetic control methods that depend on surface electromyography and require user-specific calibration.

How does SonoRank use forearm ultrasound to detect finger flexion?

SonoRank learns to rank pairs of ultrasound sequences by relative motion magnitude for each of the five fingers, then fine-tunes a model to classify active flexion using a rest reference captured at the start of operation.

What evidence shows SonoRank improves detection quality across subjects?

Under 12-fold leave-one-subject-out cross-validation on a dataset of twelve synchronized subjects, SonoRank achieves a 28% F1 score improvement over direct classification baselines that omit the ranking stage.

Jul 16, 2026

Upload by Evangeline