Using Deep Learning to Detect the Presence and Location of Hemoperitoneum on the Focused Assessment with Sonography in Trauma (FAST) Examination in Adults.

Abdominal ultrasonography has become an integral component of the evaluation of trauma patients. Internal hemorrhage can be rapidly diagnosed by finding free fluid with point-of-care ultrasound (POCUS) and expedite decisions to perform lifesaving interventions. However, the widespread clinical application of ultrasound is limited by the expertise required for image interpretation. This study aimed to develop a deep learning algorithm to identify the presence and location of hemoperitoneum on POCUS to assist novice clinicians in accurate interpretation of the Focused Assessment with Sonography in Trauma (FAST) exam. We analyzed right upper quadrant (RUQ) FAST exams obtained from 94 adult patients (44 confirmed hemoperitoneum) using the YoloV3 object detection algorithm. Exams were partitioned via fivefold stratified sampling for training, validation, and hold-out testing. We assessed each exam image-by-image using YoloV3 and determined hemoperitoneum presence for the exam using the detection with highest confidence score. We determined the detection threshold as the score that maximizes the geometric mean of sensitivity and specificity over the validation set. The algorithm had 95% sensitivity, 94% specificity, 95% accuracy, and 97% AUC over the test set, significantly outperforming three recent methods. The algorithm also exhibited strength in localization, while the detected box sizes varied with a 56% IOU averaged over positive cases. Image processing demonstrated only 57-ms latency, which is adequate for real-time use at the bedside. These results suggest that a deep learning algorithm can rapidly and accurately identify the presence and location of free fluid in the RUQ of the FAST exam in adult patients with hemoperitoneum.

Using Wearable Sensors and Machine Learning to Assess Upper Limb Function in Huntington's Disease.

Huntington's disease (HD), like many other neurological disorders, affects both lower and upper limb function that is typically assessed in the clinic - providing a snapshot of disease symptoms. Wearable sensors enable the collection of real-world data that can complement such clinical assessments and provide a more comprehensive insight into disease symptoms. In this context, almost all studies are focused on assessing lower limb function via monitoring of gait, physical activity and ambulation. In this study, we monitor upper limb function during activities of daily living in individuals with HD (n = 16), prodromal HD (pHD, n = 7), and controls (CTR, n = 16) using a wrist-worn wearable sensor, called PAMSys ULM, over seven days. The participants were highly compliant in wearing the sensor with an average daily compliance of 99% (100% for HD, 98% for pHD, and 99% for CTR). Goal-directed movements (GDM) of the hand were detected using a deep learning model, and kinematic features of each GDM were estimated. The collected data was used to predict disease groups (i.e., HD, pHD, and CTR) and clinical scores using a combination of statistical and machine learning-based models. Significant differences in GDM features were observed between the groups. HD participants performed fewer GDMs with long duration (> 7.5 seconds) compared to CTR (p-val = 0.021, d = -0.86). In velocity and acceleration metrics, the highest effect size feature was the entropy of the velocity zero-crossing length segments (HD vs CTR p-val <0.001, d = -1.67; HD vs pHD p-val = 0.043, d=-0.98; CTR vs pHD p-val = 0.046, d=0.96). In addition, this same variable showed a strongest correlation with clinical scores. Classification models achieved good performance in distinguishing HD, pHD and CTR individuals with a balanced accuracy of 67% and a 0.72 recall for the HD group, while regression models accurately predicted clinical scores. Notably the explained variance for the upper extremity function subdomain scale of Unified Huntington's Disease Rating Scale (UHDRS) was the highest, with the model capturing 60% of the variance. Our findings suggest the potential of wearables and machine learning for early identification of phenoconversion, remote monitoring in HD, and evaluating new treatments efficacy in clinical trials and medicine.