Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This work investigates the transferability of ultrasound encoder representations from a high-resource domain (echocardiography) to a low-resource domain (lung ultrasound for TB screening), finding that encoder architecture and pretraining corpus (echocardiography vs. generic video) have minimal impact on downstream performance. Instead, standardizing features between the encoder and classifier significantly boosts performance across all tested encoders, indicating that feature conditioning is the primary driver of success. The study concludes that the limited size of the labeled TB screening cohort, rather than the encoder's pretraining domain, is the bottleneck for achieving higher accuracy.
Pretraining on echocardiography offers no significant advantage over generic video for lung ultrasound TB screening, but feature conditioning unlocks substantial performance gains.
Lung ultrasound (LUS) is attractive for tuberculosis (TB) screening at primary-care level, but labelled cohorts are small. Echocardiography carries no such constraint, while sharing the same underlying ultrasound imaging physics, signal processing and B-mode appearance as LUS. We ask whether an encoder pretrained on that high-resource ultrasound domain carries representations that remain usable in the low-resource one. Only the encoder varies, across seventeen encoders spanning three architecture families. Among them, a latent-predictive video encoder pretrained on generic video (V-JEPA2-L) and its echocardiography counterpart (EchoJEPA-L) differ in pretraining corpus alone. The choice among these encoders does not resolve the classification, the whole family spanning 2.50 percentage points against a measurement resolution of 2.71. What moves the task instead is feature conditioning. Standardising the features between the encoder and the classifier improves all seventeen encoders by a mean of +1.23 percentage points at $p=1.5\times10^{-5}$. On the held-out test set every encoder selected on the development folds stands above the baseline system by up to +2.57 percentage points of area under the receiver operating characteristic curve (AUROC), and specificity at 90% sensitivity reaches 79.3% against 60.3%. The contrast specified in advance, EchoJEPA-L against V-JEPA2-L, measures -0.16 percentage points at $p=0.926$. We therefore find no evidence that shared ultrasonic physics alone makes echocardiography a more productive pretraining corpus than generic video, and any advantage, if present, is smaller than this cohort can resolve. The video encoders receive replicated still images, however, so whether this absence of an effect reflects the pretraining domain or a video encoder applied to static frames cannot be separated. The limiting factor is the labelled cohort rather than the encoder.