YOLO AI model based on an automated breast volume scanner for the detection of benign and malignant breast lesions

BackgroundIn recent years, the incidence of breast cancer has been gradually increasing. The three-level prevention strategy of early detection, early diagnosis, and early treatment is particularly important and is the key to improving the survival prognosis of breast cancer. Automated breast volume scanning (ABVS) is a three-dimensional stereoscopic ultrasound imaging technology specifically designed for breast examination. Its unique coronal plane signs provide a new perspective for the diagnosis of breast lesions, allowing for the observation of the boundaries of lesions and their relationship with surrounding tissues, as well as reflecting the morphology and growth patterns of breast nodules. This study takes advantage of the coronal plane advantages of ABVS to analyze the correlation between the “cloudy sign” and molecular biological factors of breast cancer and its distribution characteristics in different molecular subtypes of breast cancer. This study evaluated “cloud sign” in ABVS for differentiating benign and malignant breast lesions.MethodsThis retrospective observational study included patients with breast lesions who underwent ABVS examinations at The Third Affiliated Hospital of Guangzhou Medical University between January 2020 and January 2022. A total of 187 breast lesions from 185 patients were finally included.ResultsThe incidence of the “cloud sign” in the malignant group (90/123, 73.1%) of breast lesions was significantly higher than that in the benign group (73.1% vs. 12.5%, P<0.001). The incidence of the “cloud sign” differs significantly among different age groups (P<0.001) and different tumor size groups (P=0.002). In malignant breast lesions, the “cloud sign” is more frequently presented in lesions with infiltrative nature (P=0.03), positive expression of estrogen receptor (ER) (P<0.001), and positive expression of progesterone receptor (PR) (P=0.011). The area under the curve (AUC) of cloud sign, convergence sign and hyperechoic halo in the diagnosis of malignant lesions were 0.807, 0.668, and 0.778, respectively.ConclusionsThe results suggest that the cloud sign in ABVS may provide useful diagnostic value for breast lesions. The cloud sign observed in ABVS is a characteristic sign of breast lesions, which holds potential diagnostic value for differentiating between benign and malignant lesions.

Objective To evaluate the diagnostic features, characteristics, value, and clinical significance of the automated breast volume scanner (ABVS) in breast lesions. Methods A total of 288 patients with breast lesions diagnosed at the Breast Surgery Department of Peking Union Medical College Hospital between 2011 and 2015 were selected. Diagnostic and image data of preoperative ABVS examinations, hand-held breast ultrasound (HHUS), and surgery or biopsy pathology were collected. Pathology and imaging report results were recorded, accounting for the retraction phenomenon; receiver operating characteristic (ROC) curve analysis was used to calculate the diagnostic performance of the single and combined diagnostic methods. Results (1) A total of 311 breast lesions were found in 288 patients using the ABVS; histopathological diagnosis showed that there were 141 (45.3%) malignant lesions and 170 (54.7%) benign lesions. (2) The detection rates of the retraction phenomenon using the ABVS in malignant and benign lesions were, respectively, 31.2% (44/141) and 1.8% (3/170); the difference was statistically significant (χ2=52.075, P=0.000). The detection rates of the retraction phenomenon using the ABVS in invasive ductal carcinoma (IDC), ductal carcinoma in situ (DCIS), and other types of carcinomas were, respectively, 38.5% (40/104), 10.5% (2/19), and 11.1% (2/18). There were significant differences between IDC and DCIS and between IDC and other types of carcinomas (χ2=5.575, P=0.018; χ2=5.085, P=0.024, respectively). (3) The sensitivity, specificity, and accuracy rates of single ABVS were 89.4%, 80.6%, and 90.1%, respectively, and those of single HHUS were 91.5%, 74.1%, and 91.3%, respectively, for malignant lesion diagnosis. For diagnosis with combined ABVS with HHUS, the sensitivity, specificity, and accuracy rates were 93.6%, 72.9%, and 93.2%, respectively. Sensitivity and specificity rates, and the advantage ratio of the retraction phenomenon were, 31.2%, 98.2%, and 25.251, respectively. Conclusions Use of the ABVS for coronary sections with the retraction phenomenon has important clinical value in identifying malignant breast lesions, especially in identifying IDC, but ABVS cannot completely replace HHUS. ABVS combined with HHUS can improve the diagnostic capacity, and is helpful for early diagnosis of malignant breast lesions. Key words: Breast neoplasms; Ultrasonography, mammary; Automated breast volume scanner

Background & objectives:Breast cancer being one of the most common malignant tumours among women, diagnostic modalities for early detection of the same become of paramount importance. In this context, the hand-held ultrasound (HHUS) and automated breast volume scanner (ABVS) could provide valuable information for clinicians to diagnose breast diseases. This study aimed to compare and evaluate the diagnostic performance of combined use of HHUS and ABVS for the differentiation of benign and malignant breast lesions.Methods:A total of 361 female patients, who underwent both HHUS and ABVS examinations were included in this study. ABVS and HHUS images were interpreted using the American College of Radiology Breast Imaging-Reporting and Data System (BI-RADS). The distributions of the BI-RADS categories and pathology results were shown as specific numbers. Kappa coefficients test (κ) was calculated to compare the diagnostic results amongst the ABVS, HHUS and ABVS combined with HHUS. The sensitivity, specificity, accuracy, positive predictive value (PPV) and negative predictive value (NPV) of the three diagnostic methods were calculated and their respective diagnostic performance was analyzed by receiver operator characteristic curve.Results:Of a total of 431 lesions, 153 (35.5%) were malignant and 278 (64.5%) were benign. With respect to the pathology results, the value of κ was 0.713 (P<0.001) for HHUS, κ=0.765 (P<0.001) for ABVS and κ=0.815 (P<0.001) for HHUS+ABVS. The sensitivity, specificity, accuracy, PPV and NPV for HHUS combined with ABVS were 96.08 (147/153), 88.49 (246/278), 91.18 (393/431), 82.12 (147/179) and 97.62 per cent (246/252) respectively. For HHUS, these were 90.20 (138/153), 84.17 (234/278), 86.31 (372/431), 75.82 (138/182) and 93.98 per cent (234/249) respectively; and for ABVS these were 92.16 (141/153), 87.05 (242/278), 88.86 (383/431), 79.66 (141/177) and 95.28 per cent (242/254), respectively. There was no significant difference amongst these three methods, but the diagnostic performance of HHUS combined with ABVS was better than, or at least equal to, that of HHUS or ABVS alone.Interpretation & conclusions:The results of this study suggest that ABVS is a promising and advantageous modality for breast cancer detection. Furthermore, the combination of HHUS and ABVS showed a more comparable diagnostic performance than HHUS or ABVS alone for distinguishing between benign and malignant breast lesions.

This study aimed to evaluate the diagnostic accuracy (DA) of four artificial intelligence (AI) models compared to logistic regression (LR) in enhancing the performance of the fecal immunochemical test (FIT) for the detection of colorectal carcinoma (CRC). The study cohort comprised 544 patients with colorectal neoplasia (CRN), including 58 CRC and 486 non-CRC cases, recruited from the Barretos Cancer Hospital. Each patient provided three consecutive fecal samples, which were analyzed using two fecal occult blood (FOB) assays: ColonView-FIT (CV) and HemoccultSENSA. Four AI models - gradient boosting machine (GBM), neural network (NN), random forest (RF), and support vector machine (SVM) - were developed, incorporating clinical features and CV results. Diagnostic performance was assessed via hierarchical summary receiver operating characteristic (HSROC) curves. In conventional analysis, the area under the curve (AUC) values for different AI models ranged from 0.926 to 0.977, while the highest AUC values were reached by gradient boosting machine (GBM), neural network (NN), and random forest (RF) models (0.974, 0.976 and 0.977, respectively). In the HSROC analysis, the AUC values for i) 'low risk' variables, ii) 'high risk' variables, and iii) AI models were as follows: i) AUC=0.503 (95% CI=0.390-0.613), ii) AUC=0.773 (95% CI=0.713-0.837), and iii) AUC=0.958 (95% CI=0.930-0.989). In all comparisons of the AUC values, the difference was highly significant (p<0.0001). AI models outperformed conventional LR and non-AI diagnostic features in improving FIT-based CRC screening. This is the first study to show that combining clinical data with FIT results in AI frameworks can significantly improve diagnostic accuracy in CRC screening.

Objective To compare the result of automated breast volume scanner(ABVS)and common ultrasound in diagnosis of breast lesions, and evaluate the value of ABVS in diagnosing breast lesions. Methods One hundred and forty women patients with 172 breast lesions were identified by common ultrasound and ABVS, and their diagnosis efficiency were contrasted and analyzed according to the gold standard of pathology. Results The sensitivity, specificity, accuracy, positive predictive and negative predictive rate of common ultrasound were 89.0%, 75.0%, 83.1%, 83.2%, 83.1%, respectively; those of ABVS were 97.0%, 80.6%, 90.1%, 87.4%, 95.1%, respectively; those of combined two examinations were 99.0%, 88.9%, 94.8%, 92.5%, 98.5%. The sensitivity and accuracy rate of ABVS for diagnosing breast lesions were significantly higher than those of common ultrasound (χ2=4.080, 5.330, P 0.05). The sensitivity, specificity and accuracy rate of retraction phenomenon in coronal plane were 58.0%, 100% and 75.6%. Conclusions The specific coronal plane of ABVS can provide additional information in the diagnosis of breast lesions. Combination of ABVS and common ultrasound can improve the diagnostic accuracy. ABVS plays an important role in diagnosis of breast disease. Key words: Ultrasonography, three-dimensional; Breast neoplasms; Retraction phenomenon; Automated breast volume scanner

This study evaluated the diagnostic accuracy (DA) for colorectal adenomas (CRA), screened by fecal immunochemical test (FIT), using five artificial intelligence (AI) models: logistic regression (LR), support vector machine (SVM), neural network (NN), random forest (RF), and gradient boosting machine (GBM). These models were tested together with clinical features categorized as low-risk (lowR) and high-risk (highR). The colorectal neoplasia (CRN) screening cohort of 5,090 patients included 222 CRA patients and 264 non-CRA patients. Three consecutive fecal samples from each individual were analyzed by two fecal occult blood (FOB) assays. Five AI models including clinical features of CRN patients and CV test results were used to test the DA for CRA measured by receiving operating characteristic (ROC) curves. In conventional ROC analysis, the area under the curve (AUC) values for different AI models ranged from 0.659 and 0.691 (for AIs with LR and SVM), while the highest AUC values were reached by NN, RF, and GBM models (0.809, 0.840, and 0.858, respectively). In the hierarchical summary ROC (HSROC) analysis, the AUC values were as follows: i) with lowR variables, AUC=0.508; ii) with highR variables, AUC=0.566 and iii) with all AI models, AUC= 0.789. The differences in AUC values were: between i) and ii) p=0.008; between i) and iii) p<0.0001 and between ii) and iii) p<0.0001. In detection of CRA, the AI models proved to be superior to the diagnostic features without AI. This is the first study to report that DA in the diagnosis of CRA can be enhanced by AI models that include clinical data of the patients and results of FIT test.

Objective To evaluate the effectiveness of automated breast volume scanner(ABVS) and magnetic resonance imaging(MRI) in the detection and diagnosis of breast cancer and to assess the value of these modalities as well as the joint use of the two. Methods In this study, a total of 50 breast lesions in 37 patients proved by surgery and pathology were included. Before operation, all patients were underwent both ABVS and dynamic contrast-enhanced (DCE-MRI) examinations, and classified into groups according to BI-RADS classification. Then the effectiveness of the two examinations were contrasted, the image features on the two techniques were analyzed, and their differential diagnosis in benign and malignant breast lesions were compared. Results The sensitivity, specificity, accuracy, positive predictive value(PPV) and negative predictive value(NPV) of ABVS and MRI for the diagnosis of breast cancer were 91.67% and 95.83%, 88.46% and 80.77%, 90.00% and 88.00%, 88.00% and 82.14%, 92.00% and 95.45%, respectively, and there was no statistical difference between ABVS and MRI(P>0.05); The sensitivity and specificity of the combination of the two were 99.65% and 97.78%, respectively. The difference of the ABVS coronal features and MRI dynamic enhancement curve types between malignant breast lesions and the benign were statistically significant (P<0.05). Conclusions Both ABVS and MRI are effective to diagnose breast cancer well, while the combination of the two can improve the diagnosis more accurately. Key words: Ultrasonography; Breast neoplasms; Automated breast volume scanner; Magnetic resonance imaging, three-dimensional

PurposeThe aim of this study is to evaluate the clinical utility of automated breast volume scanner (ABVS) for detecting and diagnosing the breast lesions.MethodsFrom December 2010 to January 2012, bilateral whole breast examinations were performed with ABVS for 139 women. Based on the Breast Imaging Reporting and Data System (BI-RADS) categories, the breast lesions were evaluated on coronal multiplanar reconstruction images using the ABVS workstation. Then, the imaging results were compared with those on conventional handheld ultrasound (HHUS) images. Histological diagnoses were performed on BI-RADS category 4 and 5 lesions.ResultsA total of 453 lesions were detected by ABVS. On the HHUS, 33 new lesions were detected but 69 lesions were not detected. BI-RADS category 2 and 3 matched to those on ABVS at 73.5% (61/83) and 85.4% (276/323). In 47 lesions of BI-RADS category 4 or 5, there was an exact match to those on ABVS. In addition, 47 lesions were classified as BI-RADS category 4 and 5, for which an ultrasound-guided core needle biopsy was performed. The malignant lesions of BI-RADS category 4 and 5 showed the following: 2/27 (7.4%) in 4A, 4/5 (80%) in 4B, 2/2 (100%) in 4C, and 13/13 (100%) in 5. The ABVS showed 21 true positives and a positive predictive value of 44.7% (21/47).ConclusionThere was considerable agreement in the assessment of the breast lesions by ABVS and HHUS. The ABVS had advantages of high diagnostic accuracy, examiner-independence, multislice visualization of the whole breast and less time-consuming. Our results indicate that ABVS might be a useful modality in diagnosing breast lesions.

To investigate the inter-rater reliability and agreement of the automated breast volume scanner (ABVS) and the diagnostic accuracy for differentiating malignant and benign lesions. The overall aim was to find out if the ABVS is applicable to daily clinical practice. Qualifying studies were retrieved from Pubmed, EMBASE, Cochrane Library, Biosis Preview, CBM disc and by manual search and reference lists up to 30 September 2014. Pooled sensitivity and specificity of ABVS were calculated and summary receiver operating characteristic curves were drawn. Thirteen studies were included in the meta-analysis of diagnostic accuracy and seven studies were included in the systematic review of inter-rater reliability/agreement of ABVS. For 'diagnostic accuracy', the pooled values of sensitivity, specificity, positive and negative likelihood ratios, and diagnostic odds ratio were 92 % (95 % CI 89.9-93.8), 84.9 % (82.4-87.0), 6.172 (4.364-8.730), 0.101 (0.075-0.136), and 72.226 (39.637-131.61), respectively. For the studies of inter-rater reliability/agreement, the quality was heterogeneous and no evidenced result can be pooled. Sensitivity and specificity of ABVS for differentiating malignant and benign breast lesions were high. More sound studies focusing on inter-rater reliability/agreement of ABVS, which deeply affect the clinical utilization and generalization of ABVS, are urgently needed. • ABVS has high sensitivity and specificity in differentiating malignant and benign breast lesions. • The quality of published inter-rater reliability studies is heterogeneous. • Empirical evidence concerning the inter-rater reliability/agreement for the ABVS is rare. • Comparison studies on ABVS and other medical imaging examinations are warranted.

Differentiation of benign and malignant breast lesions: A comparison between automatically generated breast volume scans and handheld ultrasound examinations

This study aimed to evaluate the diagnostic value of automated breast volume scanner (ABVS) combined with virtual touch tissue quantification (VTQ) in the differential diagnosis of breast lesions.In this retrospective study, 183 patients (mean age, 49.8 ± 8.2 years) with 218 breast lesions underwent ABVS, VTQ, and mammography (MG). All lesions were confirmed by postoperative histopathology. A logistic regression model was constructed to generate a receiver operating characteristic (ROC) curve, calculate the area under the ROC curve (AUC), and compare and evaluate the diagnostic performance of ABVS, VTQ, MG, and ABVS combined with VTQ (ABVS-VTQ).The sensitivity, specificity, and accuracy of ABVS, VTQ, MG, and ABVS-VTQ in diagnosing breast lesions were 94.01% (110/117), 96.03% (97/101), and 94.95% (207/218); 80.34% (94/117), 94.05% (95/101), and 86.69% (189/218); 70.08% (82/117), 68.31% (69/101), and 69.26% (151/218); and 96.58% (113/117), 96.03% (97/101), and 96.33% (210/218), respectively. The AUC of ABVS-VTQ was higher than that of the other examinations alone. The detection rate of ABVS (100%, 218/218) was higher than that of MG (78.89%, 172/218), and the difference was statistically significant (χ2 = 51.426, P < .001).The combined application of ABVS and VTQ can improve the accuracy and specificity of the diagnosis and is a promising ultrasound method for the differential diagnosis of breast lesions.

Objectives To investigate the diagnostic performance of the combination of ultrasound elastography and automated breast volume scanner (ABVS) in differentiation of benign and malignant breast imaging reporting and data system (BI-RADS) 4 breast lesions. Methods Data from 137 breast cancer patients (147 tumors) confirmed pathologically were analyzed. Each tumor was examined by ABVS and ultrasound elastography. All tumors were diagnosed as BI-RADS 4 by ABVS. With final pathology results as the gold standard, the predictive value in differentiating BI-RADS 4 breast lesions between ultrasound elastography and the combination of ultrasound elastography and ABVS were compared. Results There were 54 benign nodules and 93 malignant nodules in this study. The diagnostic sensitivity of ultrasound elastography and the combination of ultrasound elastography and ABVS were 94.6% and 98.9%, the specificity were 57.4% and 57.4%, the accuracy were 81.0% and 83.7%, the area under the curve were 0.858 and 0.965, respectively. The diagnostic performance of ultrasound elastography combined with ABVS was better than that of ultrasound elastography. Conclusions Ultrasound elastography have certain value in differential diagnosis of BI-RADS 4 breast lesions, especially when combining with ABVS, which will improve its diagnostic accuracy. Ultrasound elastography combined with ABVS can improve the detection rate of malignant lesions in BI-RADS 4 breast lesions and reduce the rate of preoperative biopsy, and it has a good application prospect. Key words: Automated breast volume scanner; Elasticity imaging techniques; Breast diseases; Breast imaging reporting and data system

IntroductionImplementing artificial intelligence (AI) to use patient-provided intra-oral photos to detect possible pathologies represents a significant advancement in oral healthcare. AI algorithms can potentially use photographs to remotely detect issues, including caries, demineralisation, and mucosal abnormalities such as gingivitis.AimThis study aims to assess the effectiveness of a newly developed AI model in detecting common oral pathologies from intra-oral images.MethodA unique AI machine-learning model was built using a convolutional neural network (CNN) model and trained using a dataset of over five thousand images. Ninety different unseen images were selected and presented to the AI model to test the accuracy of disease detection. The AI model’s performance was compared with answers provided by fifty-one dentists who reviewed the same ninety images. Both groups identified plaque, calculus, gingivitis, and caries in the images.ResultsAmong the 51 participating dentists, clinicians correctly diagnosed 82.09% of pathologies, while AI achieved 81.11%. Clinician diagnoses matched the AI’s results 81.02% of the time. Statistical analysis using t-tests at 95% and 99% confidence levels yielded p-values of 0.63 and 0.79 for different comparisons, with mean agreement rates of 81.55% and 95.11%, respectively. The findings support the hypothesis that the average AI answers are the same as average answers by dentists, as all p-values exceeded significance thresholds (p > 0.05).ConclusionDespite current limitations, this study highlights the potential of machine learning AI models in the early detection and diagnosis of dental pathologies. AI integration has the scope to enhance clinicians’ diagnostic workflows in dentistry, with advancements in neural networks and machine learning poised to solidify its role as a valuable diagnostic aid.

Background Artificial intelligence (AI) models have improved US assessment of thyroid nodules; however, the lack of generalizability limits the application of these models. Purpose To develop AI models for segmentation and classification of thyroid nodules in US using diverse data sets from nationwide hospitals and multiple vendors, and to measure the impact of the AI models on diagnostic performance. Materials and Methods This retrospective study included consecutive patients with pathologically confirmed thyroid nodules who underwent US using equipment from 12 vendors at 208 hospitals across China from November 2017 to January 2019. The detection, segmentation, and classification models were developed based on the subset or complete set of images. Model performance was evaluated by precision and recall, Dice coefficient, and area under the receiver operating characteristic curve (AUC) analyses. Three scenarios (diagnosis without AI assistance, with freestyle AI assistance, and with rule-based AI assistance) were compared with three senior and three junior radiologists to optimize incorporation of AI into clinical practice. Results A total of 10 023 patients (median age, 46 years [IQR 37-55 years]; 7669 female) were included. The detection, segmentation, and classification models had an average precision, Dice coefficient, and AUC of 0.98 (95% CI: 0.96, 0.99), 0.86 (95% CI: 0.86, 0.87), and 0.90 (95% CI: 0.88, 0.92), respectively. The segmentation model trained on the nationwide data and classification model trained on the mixed vendor data exhibited the best performance, with a Dice coefficient of 0.91 (95% CI: 0.90, 0.91) and AUC of 0.98 (95% CI: 0.97, 1.00), respectively. The AI model outperformed all senior and junior radiologists (P < .05 for all comparisons), and the diagnostic accuracies of all radiologists were improved (P < .05 for all comparisons) with rule-based AI assistance. Conclusion Thyroid US AI models developed from diverse data sets had high diagnostic performance among the Chinese population. Rule-based AI assistance improved the performance of radiologists in thyroid cancer diagnosis. © RSNA, 2023 Supplemental material is available for this article.

Automated Breast Volume Scanning Versus Conventional Ultrasound in Breast Cancer Screening