The study, led by Qiang Ma and TianXi Liang with corresponding author Yin Wang of Shanghai Pulmonary Hospital, was designed as a retrospective, multicenter analysis. The team assembled ultrasound data from 1,059 patients with subpleural pulmonary lesions, drawn from a primary center between January 2023 and December 2024, with additional external cohorts recruited from two other hospitals between June and December 2024. The patient population had a mean age of roughly 61 years and was predominantly male, with 726 male patients among the 1,059 enrolled. Crucially, the researchers did not simply test their models on data from the same hospital that produced the training set. They held out two entirely separate external test sets, comprising 108 and 94 patients respectively, to see whether the artificial intelligence could generalize to machines, protocols, and patient populations it had never encountered.

The architecture of the framework reflects a logical division of labor. The first stage is segmentation: the computer must find the lesion in the ultrasound image and trace its boundaries precisely. To do this, the team benchmarked three well-established convolutional neural network approaches: nnU-Net, U-Net, and DeepLabv3+. The nnU-Net, a self-configuring variant of the U-Net family that has become something of a gold standard in medical image segmentation, emerged as the clear winner. Its performance was measured with the Dice similarity coefficient, a metric that scores the overlap between the machine-drawn region of interest and the ground-truth annotation, where a value of 1.0 indicates perfect agreement. The nnU-Net achieved a Dice score of 0.920 on the internal tuning-validation set of 257 cases, and, more importantly, held its ground externally, scoring 0.913 on External Test Set 1 and 0.915 on External Test Set 2. Those numbers mean the model’s outlines agreed with expert annotations to within a few percent of perfect overlap, even on unfamiliar data.

With the lesion reliably isolated, the second stage took over: classification. Here the researchers cast a wide net, training and comparing thirteen different candidate architectures drawn from the most influential families of convolutional networks in modern computer vision. The lineup included DenseNet121 and DenseNet201, the EfficientNet family from B0 through B5, three depths of ResNet (18, 50, and 101), Inception-V3, and VGG19. Each architecture brings a different inductive bias to the task. DenseNets, for instance, connect every layer to every subsequent layer, encouraging feature reuse and parameter efficiency; EfficientNets scale depth, width, and image resolution in a carefully balanced way; ResNets use skip connections that allow very deep networks to train without vanishing gradients. After head-to-head evaluation on the internal tuning-validation set, DenseNet121 proved the strongest performer and was selected as the final classification model.

The headline result is the model’s ability to distinguish benign from malignant lesions on images it had never seen. On the two external cohorts, DenseNet121 achieved areas under the receiver operating characteristic curve of 0.889 (95% confidence interval 0.827 to 0.951) and 0.865 (95% confidence interval 0.790 to 0.940). The AUC is a threshold-independent measure of discriminative ability: a value of 0.5 would be no better than a coin flip, while 1.0 would be perfect. Values approaching 0.9 indicate that, across the full range of possible decision thresholds, the model ranks a randomly chosen malignant lesion as more suspicious than a randomly chosen benign one roughly nine times out of ten. For a task that even experienced human readers find difficult on grayscale ultrasound alone, that level of performance is striking, and the fact that it held up across two independent hospitals is what separates this work from many single-center proof-of-concept studies.

Perhaps the most clinically consequential part of the study, however, was not the model’s raw scores but a carefully designed reader study. The researchers recruited six radiologists, three junior and three senior, and compared their diagnostic accuracy under two conditions: reading unaided, and reading with the AI’s output available as an aid. Because each radiologist read the same cases under both conditions, the analysis had to account for the fact that multiple observations come from the same readers and the same patients. The team used case-cluster bootstrap methods to estimate pooled three-reader differences, applying Bonferroni adjustment to control for multiple statistical comparisons, a conservative approach that reduces the risk of false-positive findings.

The outcome was unambiguous for the less experienced readers. Diagnostic accuracy among junior radiologists rose significantly in the AI-assisted reading session in both external test cohorts, with Bonferroni-adjusted P values below 0.001 in both cases. In practical terms, the AI functioned as a leveler of expertise: a junior radiologist armed with the model’s segmentation and classification output approached the kind of judgment that would otherwise require years of accumulated pattern recognition. For senior radiologists, the improvements, while directionally positive, did not reach statistical significance after the multiplicity adjustment. That pattern is common in AI-assisted reading studies and makes intuitive sense. Experts already perform well on their own, leaving less headroom for an algorithm to add measurable value, whereas novices benefit most from a second opinion that never tires and never varies.

The technical choices behind the framework deserve attention because they address two chronic weaknesses of medical deep learning: overfitting to a single center and the black-box problem. By validating on two external cohorts, the authors demonstrated that their pipeline was not merely memorizing the imaging characteristics of one hospital’s ultrasound machines. By comparing thirteen architectures rather than anointing a single favorite in advance, they let the data decide which inductive biases best suited the texture and morphology cues that distinguish benign inflammation from malignancy in subpleural tissue. The study also references gradient-weighted class activation mapping, or Grad-CAM, a visualization technique that highlights which regions of an image drove the model’s decision, and decision curve analysis, which evaluates the net clinical benefit of a diagnostic tool across different threshold preferences. The inclusion of these tools signals an effort to make the framework not only accurate but interpretable and clinically actionable.

The stakes of this problem are easy to underestimate. Subpleural lesions sit at a diagnostic crossroads: they can be early lung cancers, which demand prompt resection or oncological treatment, or they can be benign infectious or inflammatory processes, which demand nothing more invasive than follow-up or antibiotics. Distinguishing between them on grayscale ultrasound alone is genuinely hard, because both categories can produce hypoechoic masses with irregular margins, and the acoustic features that radiologists rely on are subtle and operator-dependent. An error in either direction carries a cost: a missed malignancy forfeits the best window for curative treatment, while an unnecessary biopsy or resection of a benign lesion exposes the patient to procedural risk and expense. A reliable computational second opinion, delivered in seconds from an image that is already being acquired, could shift that balance meaningfully.

The authors are appropriately measured in their conclusions, describing the framework as a potential auxiliary diagnostic tool rather than a replacement for clinical judgment. The study is retrospective, and its models were trained and tested on de-identified data from three Chinese hospitals, so prospective validation in other health systems, with different scanners and patient demographics, remains a necessary next step. The work was supported by the National Natural Science Foundation of China and several Shanghai municipal and institutional research funds, and it was conducted under ethics approvals from Shanghai Pulmonary Hospital, Shanghai Chest Hospital, and Xi’an Chest Hospital in compliance with the Declaration of Helsinki and STARD reporting guidelines. Still, the combination of near-expert segmentation, strong external AUCs, and a demonstrable boost to junior clinicians’ accuracy makes a compelling case that the humble grayscale ultrasound image, long dismissed as too subjective for lung lesion triage, is becoming a far more powerful instrument in the hands of a well-trained neural network. If future prospective trials confirm these findings, the technology could bring expert-level lung lesion assessment to clinics where CT scanners are scarce and thoracic radiologists are scarcer, turning a bedside probe and a laptop into a credible first line of defense against one of the world’s deadliest cancers.

Subject of Research: Deep learning for ultrasound-based segmentation and benign-malignant classification of subpleural pulmonary lesions

Article Title: A grayscale ultrasound-based two-stage deep learning framework for automatic segmentation and benign-malignant differentiation of subpleural pulmonary lesions

Article References: Ma, Q., Liang, T., Yi, J., Bai, H., Li, Y., Shen, M., Bi, K., & Wang, Y. (2026). A grayscale ultrasound-based two-stage deep learning framework for automatic segmentation and benign-malignant differentiation of subpleural pulmonary lesions. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02887-y

Image Credits: AI Generated

DOI: 10.1186/s12880-026-02887-y

Keywords: lung cancer, deep learning, ultrasound, subpleural pulmonary lesions, image segmentation, nnU-Net, DenseNet121, benign-malignant classification, computer-aided diagnosis, radiology, medical imaging, artificial intelligence

The technique at the heart of the consensus, known as US-PLNB, exploits a simple physical fact: sound waves cannot travel through air. That limitation once confined ultrasound to soft tissues like the liver and thyroid, but it becomes an unexpected advantage at the lung’s edge. When a tumor sits directly against the chest wall—the so-called subpleural position—with no aerated lung in between, ultrasound can see it with exquisite clarity, and in real time. Unlike computed tomography, which offers only static snapshots and exposes patients to ionizing radiation, ultrasound shows the needle, the lesion, and the beating, breathing anatomy between them continuously, allowing operators to adjust course millimeter by millimeter. The consensus notes that this real-time capability makes ultrasound particularly valuable for small, mobile lesions near the heart, great vessels, or the lung apex, where a misplaced needle could prove catastrophic.

The document does not pretend ultrasound is a universal solution. Because air blocks sound, the technique applies only to lesions abutting the parietal pleura or to central tumors made visible by surrounding collapse, consolidation, or fluid that creates an acoustic window. CT guidance remains superior for deep and central lesions, offering precise localization at the cost of radiation and longer procedure times. The consensus positions US-PLNB as the first-line method for subpleural lesions, citing reported complication rates of 1.1 to 32.7 percent and safety data consistently favoring ultrasound over CT guidance, including a retrospective series of 3,043 consecutive patients and comparative studies showing shorter wait times and procedure durations.

What distinguishes the new document from prior guidelines, such as those from the British Thoracic Society and the Korean Society of Thoracic Radiology, is its embrace of multimodal ultrasound. Beyond conventional gray-scale imaging, the consensus details how color Doppler imaging maps blood flow within and around a lesion, distinguishing the bidirectional, high-resistance waveform of pulmonary arteries—typical of benign lesions—from the low-resistance bronchial arterial signals and chaotic arteriovenous waveforms that betray malignancy. Elastography adds another dimension: stiff, blue-dominant lesions on strain imaging, or those with shear wave velocities above 2.47 meters per second, are highly likely to be malignant, with reported sensitivity and specificity reaching 98 percent in some series. These tools let operators target the most suspicious, viable regions of a tumor before the needle ever advances.

The most technically sophisticated section concerns contrast-enhanced ultrasound. After an intravenous injection of sulfur hexafluoride microbubbles, malignant lesions typically light up in a snowing-like or curly hair-like pattern as disorganized tumor neovessels fill from the periphery, while benign lesions enhance in a tree-like pattern radiating from the hilum, reflecting preserved pulmonary arterial supply. Because normal lung receives roughly 85 percent of its blood from pulmonary arteries and only 15 percent from bronchial arteries—and because tumors progressively destroy the pulmonary supply in favor of bronchial arteries—the timing of contrast arrival becomes a diagnostic signature. The consensus highlights an arrival-time difference ratio, comparing lesion enhancement against both adjacent lung and chest wall references, that achieved 91 percent diagnostic accuracy at a threshold of 43 percent in published studies. Contrast imaging also reveals necrotic, non-enhancing zones that would yield useless specimens, allowing operators to steer toward viable tissue and reduce false negatives.

Procedural detail in the consensus is granular to the point of being a manual. It specifies that warfarin should be bridged with low-molecular-weight heparin a week before biopsy, aspirin stopped three days ahead, clopidogrel five days ahead, platelets kept above 50 billion per liter, and the international normalized ratio below 1.5. Bevacizumab, the anti-angiogenic drug, must be withheld for roughly four weeks. The document contrasts fine-needle aspiration, which yields cytology with sensitivity of 90 to 95 percent for malignancy, against core needle biopsy, which provides intact histology essential for immunohistochemistry and molecular typing, and recommends combining both to maximize diagnostic yield. Needle selection, firing mechanisms, coaxial techniques that allow multiple samples through a single pleural puncture, and even methods to improve needle visibility on screen—jiggling the needle, injecting saline to outline its track—are all codified.

Complications receive equally rigorous treatment. Hemorrhage, the most common adverse event at 0.3 to 9.9 percent incidence, is managed with a graded protocol from bed rest for minor bleeding through vasopressin and tranexamic acid for moderate hemoptysis to emergency embolization for massive hemorrhage. Pneumothorax, occurring in 0.8 to 11.6 percent of cases, can often be resolved on the spot by aspirating air through the coaxial introducer needle—a maneuver the consensus illustrates with cases in which lesions obscured by escaped air became fully visible again after 20 milliliters were withdrawn. The rarest and deadliest complication, systemic arterial air embolism, carries a mortality of up to 26 percent; the consensus explains that as little as half a milliliter of air in the coronary circulation can be fatal and mandates avoiding aerated bronchi, upright positioning, and positive-pressure ventilation during puncture.

The methodology behind the consensus is as notable as its content. Developed according to the WHO Handbook for Guideline Development and registered on the PREPARE platform, it employed a three-tier structure—a steering committee of five senior experts, a fifteen-member writing group, and a voting panel of nearly one hundred multidisciplinary specialists. Clinical questions were gathered through surveys of more than one hundred physicians, evidence was systematically retrieved from eight Chinese and international databases through August 2025, study quality was assessed with Cochrane, Newcastle-Ottawa, and QUADAS tools, and recommendations were graded using the GRADE framework and ratified through two rounds of anonymous Delphi voting requiring at least 80 percent agreement.

The consensus also looks forward. It endorses image fusion and volume navigation technology, in which electromagnetic sensors track the ultrasound probe and biopsy needle in space, fusing real-time ultrasound with pre-acquired CT volumes to guide biopsies of lesions once considered beyond ultrasound’s reach. The authors acknowledge that large prospective randomized trials directly comparing ultrasound with other guidance modalities are still lacking, and they commit to annual literature monitoring and timely updates as new evidence emerges. For now, the document stands as both a practical manual and a statement of ambition: that a technique once dismissed as too limited for the air-filled lung has matured, with the help of contrast agents, elastography, and navigation systems, into a precise, radiation-free first-line approach for the growing population of patients whose lung lesions sit within reach of a sound wave.

Subject of Research: Development of the first Chinese expert consensus standardizing ultrasound-guided percutaneous lung needle biopsy for subpleural pulmonary lesions

Article Title: Chinese expert consensus on ultrasound-guided percutaneous lung needle biopsy (2025 Edition)

Article References: Wang, Y., Xu, H., & on behalf of Chinese Society of Ultrasound in Medicine, Interventional Physicians Branch of the Chinese Medical Doctor Association, Minimally Invasive Therapy for Oncology Committee of Chinese Anti-Cancer Association, Ultrasound Branch of the Chinese Anti-Tuberculosis Association, Tuberculosis Control and Prevention Branch of the China International Exchange and Promotive Association for Medical and Health Care, Ultrasound Equipment Technology Branch of China Association of Medical Equipment, Ultrasound Su (2026). Chinese expert consensus on ultrasound-guided percutaneous lung needle biopsy (2025 Edition). Clinical Cancer Bulletin, 5(1), Article 4. https://doi.org/10.1007/s44272-026-00055-5

Image Credits: AI Generated

DOI: 10.1007/s44272-026-00055-5

Keywords: ultrasound-guided biopsy, lung biopsy, subpleural pulmonary lesions, expert consensus, contrast-enhanced ultrasound, elastography, interventional ultrasound, pneumothorax, lung cancer diagnosis, image fusion navigation, core needle biopsy, clinical guideline