High resolution dermal ultrasound (US) combined with superficial radiation therapy (SRT) versus non-image guided SRT or external beam radiotherapy (XRT) in early-stage epithelial cancer: a comparison of studies.

BACKGROUND: To compare the effectiveness of high-resolution dermal ultrasound (US) guided superficial radiotherapy (SRT) to non-image-guided radiotherapy in the treatment of early-stage Non-Melanoma Skin Cancer (NMSC). METHODS: A high-resolution dermal ultrasound (US) image guided form of superficial radiation therapy (designated here as US-SRT) was developed in 2013 where the tumor configuration and depth can be visualized prior to, during, and subsequent to treatments, using a 22 megahertz (MHz) dermal ultrasound (US) with a doppler component. We previously published the results using this technology to treat 2917 early-stage epithelial cancers showing a high local control (LC) rate of 99.3%. We compared these results with similar American studies from a comprehensive literature search used in an article/guideline published by American Society of Radiation Oncology (ASTRO) on curative radiation treatment of basal cell carcinoma (BCC), squamous cell carcinoma (SCC) and squamous cell carcinoma in-situ (SCCIS) lesions from 1988 to 2018. Only U.S. based studies with greater than 100 cases with similar patient/lesion characteristics and stages treated by external beam, electron, or superficial/orthovoltage radiation therapy were included in the criteria for selection. The resultant 4 studies had appropriate comparable cases identified and the data analyzed/calculated with regard to local control. Logistic regression analysis was performed comparing each study to US-SRT individually and collectively with stratification by histology (BCC, SCC, and SCCIS). RESULTS: US-SRT LC was found to be statistically superior to each of the 4 non-image-guided radiation therapy studies individually and collectively (as well as stratified by histology subtype) with p-values ranging from p < 0.0001 to p = 0.0438. CONCLUSIONS: Results of US-SRT in local control were statistically significantly superior across the board versus non-image-guided radiation modalities in treatment of epithelial NMSC and should be considered a new gold standard for treatment of early-stage cutaneous BCC, SCC, and SCCIS.

Analysis of image-guided superficial radiation therapy (IGSRT) on the treatment of early-stage non-melanoma skin cancer (NMSC) in the outpatient dermatology setting.

BACKGROUND: Interest in image-guidance superficial radiation therapy (IGSRT) for the treatment of early-stage non-melanoma skin cancer (NMSC) has resurfaced given its low complication rates, superior cosmesis and local control and cure rates. In addition, it has been recommended by the American Academy of Dermatology (AAD) for early-stage NMSC in patients who are considered poor surgical candidates. METHODS: 1899 NMSC lesions were treated with energies ranging from 50 to 100 kilovoltage (kV), for a mean of 20.2 fractions, and treatment dose of 5364.4 centigray (cGy). Lesions were treated for a mean of 7.5 weeks and followed for 65.5 weeks. SAS studio was used to conduct Kaplan-Meier analysis to calculate local control rates and account for differences in follow-up intervals. A log-rank test was used to calculate statistical differences between histologies. RESULTS: Absolute lesion control was achieved in 99.7% of the patients after an average of 7.5 weeks of treatment, with a stable control rate of 99.6% when the follow-up duration was over 12 months. 95% of lesions with toxicity scoring received a Radiation Treatment Oncology Group Toxicity (RTOG) score of 1 or 2. CONCLUSION: IGSRT has a high safety profile, can achieve superior cosmesis and should be considered first-line for treating early-stage NMSC tumors as cure rates have been shown to be effective in all NMSC on early follow-up.

Local control comparison of early-stage non-melanoma skin Cancer (NMSC) treated by superficial radiotherapy (SRT) and external beam radiotherapy (XRT) with and without dermal image guidance: a meta-analysis.

BACKGROUND: Various treatments exist for non-melanoma skin cancer (NMSC), but the mainstay is surgical removal. Superficial radiotherapy (SRT) is one non-surgical technique that has been used for over a century but fell out of favor due to the advent of Mohs micrographic surgery (MMS). A new technology that combines a 22 megahertz (MHz) dermal ultrasound with SRT (US-SRT) enables tumor visualization before, during, and after treatment, and demonstrates increased cure rates and reduced recurrences. METHODS: We conducted a meta-analysis comparing the local control (LC) of four studies using traditional non-image-guided forms of radiotherapy for NMSC treatment to two seminal studies utilizing high-resolution dermal ultrasound-guided SRT (HRUS-SRT). The four traditional radiotherapy studies were obtained from a comprehensive literature search used in an article published by the American Society of Radiation Oncology (ASTRO) on curative radiation treatment of basal cell carcinoma (BCC), squamous cell carcinoma (SCC) and squamous cell carcinoma in-situ (SCCIS) lesions. The meta-analysis employed a logit as the effect size indicator with Q-statistic to test the null hypothesis. RESULTS: LC rates for the 2 US-SRT studies were statistically superior to the 4 traditional therapies individually and collectively. When stratified by histology, statistically superior outcomes for US-SRT were observed in all subtypes with p-values ranging from p < 0.0001 to p = 0.0438. These results validated an earlier analysis using a logistic regression statistical method showing the same results. CONCLUSION: US-SRT is statistically superior to non-image-guided radiotherapies for NMSC treatment. This modality may represent the future standard of non-surgical treatment for early-stage NMSC.

A Multimodal Transformer: Fusing Clinical Notes with Structured EHR Data for Interpretable In-Hospital Mortality Prediction.

Deep-learning-based clinical decision support using structured electronic health records (EHR) has been an active research area for predicting risks of mortality and diseases. Meanwhile, large amounts of narrative clinical notes provide complementary information, but are often not integrated into predictive models. In this paper, we provide a novel multimodal transformer to fuse clinical notes and structured EHR data for better prediction of in-hospital mortality. To improve interpretability, we propose an integrated gradients (IG) method to select important words in clinical notes and discover the critical structured EHR features with Shapley values. These important words and clinical features are visualized to assist with interpretation of the prediction outcomes. We also investigate the significance of domain adaptive pretraining and task adaptive fine-tuning on the Clinical BERT, which is used to learn the representations of clinical notes. Experiments demonstrated that our model outperforms other methods (AUCPR: 0.538, AUCROC: 0.877, F1:0.490).