WATUNet: a deep neural network for segmentation of volumetric sweep imaging ultrasound.

Limited access to breast cancer diagnosis globally leads to delayed treatment. Ultrasound, an effective yet underutilized method, requires specialized training for sonographers, which hinders its widespread use. Volume sweep imaging (VSI) is an innovative approach that enables untrained operators to capture high-quality ultrasound images. Combined with deep learning, like convolutional neural networks, it can potentially transform breast cancer diagnosis, enhancing accuracy, saving time and costs, and improving patient outcomes. The widely used UNet architecture, known for medical image segmentation, has limitations, such as vanishing gradients and a lack of multi-scale feature extraction and selective region attention. In this study, we present a novel segmentation model known as Wavelet_Attention_UNet (WATUNet). In this model, we incorporate wavelet gates and attention gates between the encoder and decoder instead of a simple connection to overcome the limitations mentioned, thereby improving model performance. Two datasets are utilized for the analysis: the public 'Breast Ultrasound Images' dataset of 780 images and a private VSI dataset of 3818 images, captured at the University of Rochester by the authors. Both datasets contained segmented lesions categorized into three types: no mass, benign mass, and malignant mass. Our segmentation results show superior performance compared to other deep networks. The proposed algorithm attained a Dice coefficient of 0.94 and an F1 score of 0.94 on the VSI dataset and scored 0.93 and 0.94 on the public dataset, respectively. Moreover, our model significantly outperformed other models in McNemar's test with false discovery rate correction on a 381-image VSI set. The experimental findings demonstrate that the proposed WATUNet model achieves precise segmentation of breast lesions in both standard-of-care and VSI images, surpassing state-of-the-art models. Hence, the model holds considerable promise for assisting in lesion identification, an essential step in the clinical diagnosis of breast lesions.

No sonographer, no radiologist: Assessing accuracy of artificial intelligence on breast ultrasound volume sweep imaging scans.

Breast ultrasound provides a first-line evaluation for breast masses, but the majority of the world lacks access to any form of diagnostic imaging. In this pilot study, we assessed the combination of artificial intelligence (Samsung S-Detect for Breast) with volume sweep imaging (VSI) ultrasound scans to evaluate the possibility of inexpensive, fully automated breast ultrasound acquisition and preliminary interpretation without an experienced sonographer or radiologist. This study was conducted using examinations from a curated data set from a previously published clinical study of breast VSI. Examinations in this data set were obtained by medical students without prior ultrasound experience who performed VSI using a portable Butterfly iQ ultrasound probe. Standard of care ultrasound exams were performed concurrently by an experienced sonographer using a high-end ultrasound machine. Expert-selected VSI images and standard of care images were input into S-Detect which output mass features and classification as "possibly benign" and "possibly malignant." Subsequent comparison of the S-Detect VSI report was made between 1) the standard of care ultrasound report by an expert radiologist, 2) the standard of care ultrasound S-Detect report, 3) the VSI report by an expert radiologist, and 4) the pathological diagnosis. There were 115 masses analyzed by S-Detect from the curated data set. There was substantial agreement of the S-Detect interpretation of VSI among cancers, cysts, fibroadenomas, and lipomas to the expert standard of care ultrasound report (Cohen's κ = 0.73 (0.57-0.9 95% CI), p<0.0001), the standard of care ultrasound S-Detect interpretation (Cohen's κ = 0.79 (0.65-0.94 95% CI), p<0.0001), the expert VSI ultrasound report (Cohen's κ = 0.73 (0.57-0.9 95% CI), p<0.0001), and the pathological diagnosis (Cohen's κ = 0.80 (0.64-0.95 95% CI), p<0.0001). All pathologically proven cancers (n = 20) were designated as "possibly malignant" by S-Detect with a sensitivity of 100% and specificity of 86%. Integration of artificial intelligence and VSI could allow both acquisition and interpretation of ultrasound images without a sonographer and radiologist. This approach holds potential for increasing access to ultrasound imaging and therefore improving outcomes related to breast cancer in low- and middle- income countries.

Lung ultrasound volume sweep imaging for respiratory illness: a new horizon in expanding imaging access.

BACKGROUND: Respiratory illness is a leading cause of morbidity in adults and the number one cause of mortality in children, yet billions of people lack access to medical imaging to assist in its diagnosis. Although ultrasound is highly sensitive and specific for respiratory illness such as pneumonia, its deployment is limited by a lack of sonographers. As a solution, we tested a standardised lung ultrasound volume sweep imaging (VSI) protocol based solely on external body landmarks performed by individuals without prior ultrasound experience after brief training. Each step in the VSI protocol is saved as a video clip for later interpretation by a specialist. METHODS: Dyspneic hospitalised patients were scanned by ultrasound naive operators after 2 hours of training using the lung ultrasound VSI protocol. Separate blinded readers interpreted both lung ultrasound VSI examinations and standard of care chest radiographs to ascertain the diagnostic value of lung VSI considering chest X-ray as the reference standard. Comparison to clinical diagnosis as documented in the medical record and CT (when available) were also performed. Readers offered a final interpretation of normal, abnormal, or indeterminate/borderline for each VSI examination, chest X-ray, and CT. RESULTS: Operators scanned 102 subjects (0-89 years old) for analysis. Lung VSI showed a sensitivity of 93% and a specificity of 91% for an abnormal chest X-ray and a sensitivity of 100% and a specificity of 93% for a clinical diagnosis of pneumonia. When any cases with an indeterminate rating on chest X-ray or ultrasound were excluded (n=38), VSI lung ultrasound showed 92% agreement with chest X-ray (Cohen's κ 0.83 (0.68 to 0.97, p<0.0001)). Among cases with CT (n=21), when any ultrasound with an indeterminate rating was excluded (n=3), there was 100% agreement with VSI. CONCLUSION: Lung VSI performed by previously inexperienced ultrasound operators after brief training showed excellent agreement with chest X-ray and high sensitivity and specificity for a clinical diagnosis of pneumonia. Blinded readers were able to identify other respiratory diseases including pulmonary oedema and pleural effusion. Deployment of lung VSI could benefit the health of the global community.

Testing telediagnostic right upper quadrant abdominal ultrasound in Peru: A new horizon in expanding access to imaging in rural and underserved areas.

BACKGROUND: Hepatic and biliary diseases are prevalent worldwide, but the majority of people lack access to diagnostic medical imaging for their assessment. The liver and gallbladder are readily amenable to sonographic examination, and ultrasound is a portable, cost-effective imaging modality suitable for use in rural and underserved areas. However, the deployment of ultrasound in these settings is limited by the lack of experienced sonographers to perform the exam. In this study, we tested an asynchronous telediagnostic system for right upper quadrant abdominal ultrasound examination operated by individuals without prior ultrasound experience to facilitate deployment of ultrasound to rural and underserved areas. METHODS: The teleultrasound system utilized in this study employs volume sweep imaging and a telemedicine app installed on a tablet which connects to an ultrasound machine. Volume sweep imaging is an ultrasound technique in which an individual scans the target region utilizing preset ultrasound sweeps demarcated by easily recognized external body landmarks. The sweeps are saved as video clips for later interpretation by an experienced radiologist. Teleultrasound scans from a Peruvian clinic obtained by individuals without prior ultrasound experience were sent to the United States for remote interpretation and quality assessment. Standard of care comparison was made to a same-day ultrasound examination performed by a radiologist. RESULTS: Individuals without prior ultrasound experience scanned 144 subjects. Image quality was rated "poor" on 36.8% of exams, "acceptable" on 38.9% of exams, and "excellent" on 24.3% of exams. Among telemedicine exams of "acceptable" or "excellent" image quality (n = 91), greater than 80% of the liver and gallbladder were visualized in the majority of cases. In this group, there was 95% agreement between standard of care and teleultrasound on whether an exam was normal or abnormal, with a Cohen's kappa of 0.84 (95% CI 0.7-0.98, p <0.0001). Finally, among these teleultrasound exams of "acceptable" or "excellent" image quality, the sensitivity for cholelithiasis was 93% (95% CI 68.1%-99.8%), and the specificity was 97% (95% CI 89.5%-99.6%). CONCLUSION: This asynchronous telediagnostic system allows individuals without prior ultrasound experience to effectively scan the liver, gallbladder, and right kidney with a high degree of agreement with standard of care ultrasound. This system can be deployed to improve access to diagnostic imaging in low-resource areas.

Testing telediagnostic obstetric ultrasound in Peru: a new horizon in expanding access to prenatal ultrasound.

BACKGROUND: Ninety-four percent of all maternal deaths occur in low- and middle-income countries, and the majority are preventable. Access to quality Obstetric ultrasound can identify some complications leading to maternal and neonatal/perinatal mortality or morbidity and may allow timely referral to higher-resource centers. However, there are significant global inequalities in access to imaging and many challenges to deploying ultrasound to rural areas. In this study, we tested a novel, innovative Obstetric telediagnostic ultrasound system in which the imaging acquisitions are obtained by an operator without prior ultrasound experience using simple scan protocols based only on external body landmarks and uploaded using low-bandwidth internet for asynchronous remote interpretation by an off-site specialist. METHODS: This is a single-center pilot study. A nurse and care technician underwent 8 h of training on the telediagnostic system. Subsequently, 126 patients (68 second trimester and 58 third trimester) were recruited at a health center in Lima, Peru and scanned by these ultrasound-naïve operators. The imaging acquisitions were uploaded by the telemedicine platform and interpreted remotely in the United States. Comparison of telediagnostic imaging was made to a concurrently performed standard of care ultrasound obtained and interpreted by an experienced attending radiologist. Cohen's Kappa was used to test agreement between categorical variables. Intraclass correlation and Bland-Altman plots were used to test agreement between continuous variables. RESULTS: Obstetric ultrasound telediagnosis showed excellent agreement with standard of care ultrasound allowing the identification of number of fetuses (100% agreement), fetal presentation (95.8% agreement, κ =0.78 (p < 0.0001)), placental location (85.6% agreement, κ =0.74 (p < 0.0001)), and assessment of normal/abnormal amniotic fluid volume (99.2% agreement) with sensitivity and specificity > 95% for all variables. Intraclass correlation was good or excellent for all fetal biometric measurements (0.81-0.95). The majority (88.5%) of second trimester ultrasound exam biometry measurements produced dating within 14 days of standard of care ultrasound. CONCLUSION: This Obstetric ultrasound telediagnostic system is a promising means to increase access to diagnostic Obstetric ultrasound in low-resource settings. The telediagnostic system demonstrated excellent agreement with standard of care ultrasound. Fetal biometric measurements were acceptable for use in the detection of gross discrepancies in fetal size requiring further follow up.