Radiology reports are one of the main forms of communication between radiologists and other clinicians, and contain important information for patient care. In order to use this information for research and automated patient care programs, it is necessary to convert the raw text into structured data suitable for analysis. State-of-the-art natural language processing (NLP) domain-specific contextual word embeddings have been shown to achieve impressive accuracy for these tasks in medicine, but have yet to be utilized for section structure segmentation. In this work, we pre-trained a contextual embedding BERT model using breast radiology reports and developed a classifier that incorporated the embedding with auxiliary global textual features in order to perform section segmentation. This model achieved 98% accuracy in segregating free-text reports, sentence by sentence, into sections of information outlined in the Breast Imaging Reporting and Data System (BI-RADS) lexicon, which is a significant improvement over the classic BERT model without auxiliary information. We then evaluated whether using section segmentation improved the downstream extraction of clinically relevant information such as modality/procedure, previous cancer, menopausal status, purpose of exam, breast density, and breast MRI background parenchymal enhancement. Using the BERT model pre-trained on breast radiology reports, combined with section segmentation, resulted in an overall accuracy of 95.9% in the field extraction tasks. This is a 17% improvement, compared to an overall accuracy of 78.9% for field extraction with models using classic BERT embeddings and not using section segmentation. Our work shows the strength of using BERT in the analysis of radiology reports and the advantages of section segmentation by identifying the key features of patient factors recorded in breast radiology reports.

Background: Preclinical studies have demonstrated that tumour cell death can be enhanced 10- to 40-fold when radiotherapy is combined with focussed ultrasound-stimulated microbubble (FUS-MB) treatment. The acoustic exposure of microbubbles (intravascular gas microspheres) within the target volume causes bubble cavitation, which induces perturbation of tumour vasculature and activates endothelial cell apoptotic pathways responsible for the ablative effect of stereotactic body radiotherapy. Subsequent irradiation of a microbubble-sensitised tumour causes rapid increased tumour death. The study here presents the mature safety and efficacy outcomes of magnetic resonance (MR)-guided FUS-MB (MRgFUS-MB) treatment, a radioenhancement therapy for breast cancer. Methods and findings: This prospective, single-center, single-arm Phase 1 clinical trial included patients with stages I–IV breast cancer with in situ tumours for whom breast or chest wall radiotherapy was deemed adequate by a multidisciplinary team (clinicaltrials.gov identifier: NCT04431674). Patients were excluded if they had contraindications for contrast-enhanced MR or microbubble administration. Patients underwent 2 to 3 MRgFUS-MB treatments throughout radiotherapy. An MR-coupled focussed ultrasound device operating at 800 kHz and 570 kPa peak negative pressure was used to sonicate intravenously administrated microbubbles within the MR-guided target volume. The primary outcome was acute toxicity per Common Terminology Criteria for Adverse Events (CTCAE) v5.0. Secondary outcomes were tumour response at 3 months and local control (LC). A total of 21 female patients presenting with 23 primary breast tumours were enrolled and allocated to intervention between August/2020 and November/2022. Three patients subsequently withdrew consent and, therefore, 18 patients with 20 tumours were included in the safety and LC analyses. Two patients died due to progressive metastatic disease before 3 months following treatment completion and were excluded from the tumour response analysis. The prescribed radiation doses were 20 Gy/5 fractions (40%, n = 8/20), 30 to 35 Gy/5 fractions (35%, n = 7/20), 30 to 40 Gy/10 fractions (15%, n = 3/20), and 66 Gy/33 fractions (10%, n = 2/20). The median follow-up was 9 months (range, 0.3 to 29). Radiation dermatitis was the most common acute toxicity (Grade 1 in 16/20, Grade 2 in 1/20, and Grade 3 in 2/20). One patient developed grade 1 allergic reaction possibly related to microbubbles administration. At 3 months, 18 tumours were evaluated for response: 9 exhibited complete response (50%, n = 9/18), 6 partial response (33%, n = 6/18), 2 stable disease (11%, n = 2/18), and 1 progressive disease (6%, n = 1/18). Further follow-up of responses indicated that the 6-, 12-, and 24-month LC rates were 94% (95% confidence interval [CI] [84%, 100%]), 88% (95% CI [75%, 100%]), and 76% (95% CI [54%, 100%]), respectively. The study’s limitations include variable tumour sizes and dose fractionation regimens and the anticipated small sample size typical for a Phase 1 clinical trial. Conclusions: MRgFUS-MB is an innovative radioenhancement therapy associated with a safe profile, potentially promising responses, and durable LC. These results warrant validation in Phase 2 clinical trials. Trial registration: clinicaltrials.gov, identifier NCT04431674.

Https://doi.org/10.18632/oncotarget.28002 Christopher Kolios, Lakshmanan Sannachi, Archya Dasgupta, Harini Suraweera, Daniel DiCenzo, Gregory Stanisz, Arjun Sahgal, Frances Wright, Nicole Look-Hong,...

Background: This study was conducted to explore the use of quantitative ultrasound (QUS) in predicting recurrence for patients with locally advanced breast cancer (LABC) early during neoadjuvant chemotherapy (NAC). Methods: Eighty-three patients with LABC were scanned with 7 MHz ultrasound before starting NAC (week 0) and during treatment (week 4). Spectral parametric maps were generated corresponding to tumor volume. Twenty-four textural features (QUS-Tex¹) were determined from parametric maps acquired using grey-level co-occurrence matrices (GLCM) for each patient, which were further processed to generate 64 texture derivatives (QUS-Tex¹-Tex²), leading to a total of 95 features from each time point. Analysis was carried out on week 4 data and compared to baseline (week 0) data. ∆Week 4 data was obtained from the difference in QUS parameters, texture features (QUS-Tex¹), and texture derivatives (QUS-Tex¹-Tex²) of week 4 data and week 0 data. Patients were divided into two groups: recurrence and non-recurrence. Machine learning algorithms using k-nearest neighbor (k-NN) and support vector machines (SVMs) were used to generate radiomic models. Internal validation was undertaken using leave-one patient out cross-validation method. Results: With a median follow up of 69 months (range 7–118 months), 28 patients had disease recurrence. The k-NN classifier was the best performing algorithm at week 4 with sensitivity, specificity, accuracy, and area under curve (AUC) of 87%, 75%, 81%, and 0.83, respectively. The inclusion of texture derivatives (QUS-Tex¹-Tex²) in week 4 QUS data analysis led to the improvement of the classifier performances. The AUC increased from 0.70 (0.59 to 0.79, 95% confidence interval) without texture derivatives to 0.83 (0.73 to 0.92) with texture derivatives. The most relevant features separating the two groups were higher-order texture derivatives obtained from scatterer diameter and acoustic concentration-related parametric images. Conclusions: This is the first study highlighting the utility of QUS radiomics in the prediction of recurrence during the treatment of LABC. It reflects that the ongoing treatment-related changes can predict clinical outcomes with higher accuracy as compared to pretreatment features alone.

(1) Background: Neoadjuvant chemotherapy (NAC) is an integral part of breast cancer management, and response to NAC is an important prognostic factor associated with improved survival outcomes. However, the current standard for response assessment relies on post-surgical histopathological analysis, which limits early therapeutic decision-making and treatment personalization. This study aimed to develop and evaluate a machine learning model that integrates pre-treatment MRI radiomics and clinical features to predict response to NAC in breast cancer patients. (2) Methods: In this study, a machine learning model was developed to predict breast cancer response to NAC using pre-treatment magnetic resonance imaging (MRI) radiomics and clinical data. Radiomic features were extracted from contrast-enhanced T1-weighted (CE-T1) and T2-weighted (T2) MRI sequences using both intratumoral and peritumoral segmentations. Furthermore, this study uniquely examined two response assessment criteria: (1) pathologic complete response (pCR) versus non-pCR, and (2) clinical response versus non-response. A total of 254 patients with biopsy-confirmed breast cancer who completed NAC were included. Radiomic features (n = 400) and clinical features (n = 7) were analyzed to build a predictive model employing the XGBoost classifier. Performance was measured in terms of accuracy, precision, sensitivity, specificity, F1-score, and AUC. (3) Results: The integration of radiomic features with clinical data significantly enhanced the predictive performance. For pCR and non-pCR prediction, the combined features model achieved an accuracy of 80% and AUC of 0.85, outperforming both the clinical features model (Accuracy = 68%, AUC = 0.81) and radiomic features model (Accuracy = 66%, AUC = 0.60). Similarly, for the clinical response and non-response prediction, the combined features model achieved an Accuracy of 74% and AUC of 0.75, outperforming both the clinical features model (Accuracy = 63%, AUC = 0.68) and radiomic features model (Accuracy = 66%, AUC = 0.57). (4) Conclusions: These findings highlight the synergistic effect of integrating clinical data and MRI-based radiomics to improve pre-treatment NAC response prediction, which has the potential to enable more precise and personalized treatment strategies.