Towards precision medicine based on a continuous deep learning optimization and ensemble approach.

We developed a continuous learning system (CLS) based on deep learning and optimization and ensemble approach, and conducted a retrospective data simulated prospective study using ultrasound images of breast masses for precise diagnoses. We extracted 629 breast masses and 2235 images from 561 cases in the institution to train the model in six stages to diagnose benign and malignant tumors, pathological types, and diseases. We randomly selected 180 out of 3098 cases from two external institutions. The CLS was tested with seven independent datasets and compared with 21 physicians, and the system's diagnostic ability exceeded 20 physicians by training stage six. The optimal integrated method we developed is expected accurately diagnose breast masses. This method can also be extended to the intelligent diagnosis of masses in other organs. Overall, our findings have potential value in further promoting the application of AI diagnosis in precision medicine.

Quantitative analysis of contrast-enhanced ultrasound combined with ultrasound in the unifocal papillary thyroid micro-carcinoma.

OBJECTIVE: To evaluate diagnostic value of ultrasound (US) combined with contrast-enhanced ultrasound (CEUS) in the invasiveness of unifocal papillary thyroid micro-carcinoma (UPTMC) without capsule-invasion. METHODS: This retrospective study included data from patients with UPTMC who received US and CEUS examinations in the Ultrasound Department of the Central Hospital of Changsha, China between June 2019 and September 2021. Univariate and multivariate logistic regression analysis were used to evaluate the risk of US and CEUS parameters for UPTMC. Diagnostic performance was estimated by ROC analysis. RESULTS: A total of 136 cases were enrolled, including invasive UPTMC (n = 47) and non-invasive UPTMC (n = 89), which were divided into test set (n = 109) and validation set (n = 27). The occurrence of microcalcification and the ratios (R) of each time-intensity curve (TIC) of CEUS parameter were significantly higher in patients with invasive UTPMC than non-invasive UPTMC (all P < 0.05). Additionally, nodular diameter was significantly longer in the invasive group (P < 0.05). Multivariate analysis showed that microcalcification (OR = 2.917, 95% CI: 1.002-8.491, P = 0.050), R-TTP > 1 (OR = 3.376, 95%CI: 1.267-8.994, P = 0.015), R-DS > 1 (OR = 6.558, 95% CI: 2.358-18.243, P < 0.010) were independently associated with invasive UPTMC. The sensitivities of US, CEUS and their combined application were 82.1%, 46.2% and 79.5%, respectively, and their specificities were 37.1%, 88.6% and 61.4%, respectively. The combination of the two methods had the best diagnostic efficiency (AUC=0.775)compared to US (AUC = 0.596) and CEUS (AUC = 0.750). CONCLUSION: The combination of US and CEUS might have good diagnostic value for UPTMC with capsule non-invasion.

Prediction of Outcomes of Ultrasound-Guided Saline Enema in the Treatment of Pediatric Intussusception: A Retrospective Case-Control Study.

OBJECTIVES: Outcomes of ultrasound-guided saline enema include successful treatment, unsuccessful treatment, or recurrence. This study aimed to investigate the value of ultrasonic parameters of the ileocecal region during hydrostatic reduction to predict enema outcomes. METHODS: Ultrasound images of patients diagnosed with ileocolic intussusception and treated with ultrasound-guided saline enema at two different institutions between January 2019 and April 2021 were retrospectively analyzed to assess ileocecal-valve diameter (ICVD), intussusceptum thickness (IT), and the ratio of IT to ICVD (I/I). Logistic regression analysis was used to explore correlations between ICVD, IT, I/I, and patient characteristics (sex, age, symptom duration, and enema outcome). RESULTS: Of 291 patients with ileocolic intussusception (207 boys; mean ICVD, 8.6 [SD: 0.1] mm; mean IT, 26 [SD: 0.2] mm; mean I/I, 3.0 [SD: 0.01]), 268 had first successful reduction; 23, first failed reduction; 7, final failed reduction; and 41, early recurrence. Significant risk factors for failed reduction included symptom duration >24 hours (odds ratio [OR] = 10, P = .012), ICVD ≤ 8.5 mm (OR = 8, P = .01), and I/I > 3.25 (OR = 16, P < .001). Significant risk factors for early recurrence post-enema included age >1 year (OR = 10, P = .028), ICVD > 8.5 mm (OR = 4, P = .003), and I/I ≤ 2.95 (OR = 6, P < .001). CONCLUSIONS: ICVD and IT measured during ultrasound-guided hydrostatic reduction can predict enema outcomes. The mismatch between IT and ICVD is the primary cause of poor outcomes.

An Intelligent Reminder System Reduces Deficiencies and Errors in Ultrasound Reports.

OBJECTIVES: We designed a computer-based, integrated intelligent reminder system to reduce the deficiencies and errors in ultrasound (US) reports. In this study, we assessed the performance of this system and evaluated its impact on the quality of US reporting. METHODS: Ultrasound reporting deficiencies or errors were divided into 2 categories: missing items (including outpatient or inpatient number and clinical diagnosis) and content errors (including measurement data, sex-related, and laterality errors). The intelligent reminder system was designed in Visual Basic for Applications (Microsoft Corporation, Redmond, WA) and integrated with the US system. It automatically detects reporting errors before printing of the report and provides real-time prompts for correction of the errors. We compared the US reporting deficiencies and errors during the 20 months before and after implementation of the system. RESULTS: Before implementation of the system, deficiencies/errors were detected in 2.26% (8841 of 391,230) of US reports compared with 0.12% (530 of 444,215) of reports after implementation of the system (P < .0001). After adoption of the system, the reported item deficiencies were improved more than the content deficiencies, with the most notable improvement in clinical diagnosis. Sex-related errors were reduced from 7 cases to nil after use of the intelligent reminder system. No laterality errors were found before and after the implementation of the system. CONCLUSIONS: The intelligent reminder system within the US system significantly reduced deficiencies and errors, improving the quality of the report.

Accurately distinguishing pediatric ileocolic intussusception from small-bowel intussusception using ultrasonography.

PURPOSE: The clinical treatment of ileocolic intussusception is different from that of small-bowel intussusception (SBI). This study aimed to analyze the differences between the two groups using ultrasonoscopy to avoid misdiagnosis. METHODS: We conducted a retrospective study of intussusception in patients aged 0-18 years from September 2018 to March 2020. Clinical and ultrasonoscopy data were reviewed. RESULTS: A total of 183 cases of intussusception were included in this study (123 cases of SBI and 60 of ileocolic intussusception). Ultrasonoscopy features that were significantly different between the two groups (p<0.05) included the lesion diameter, fat core thickness, outer wall thickness, lymph nodes inside intussusception, and lesion length. In the SBI group, 85% (104/123) of the normal ileocecal region was found in the right lower quadrant and 98% (120/123) of the normal ascending colon was found in the right side of the abdomen. In the ileocolic intussusception group, none of the normal ileocecal region was found and the ascending colon was found in only 5 cases (8%) (p<0.05). CONCLUSIONS: The use of ultrasonography is the most practical method to distinguish SBI from ileocolic intussusception. In addition to comparing the differences between lesions, conventional probing of the ileocecal region and ascending colon will help to accurately differentiate between the types of intussusception and avoid misdiagnosis. LEVELS OF EVIDENCE: Diagnostic.