Machine learning analysis of breast ultrasound to classify triple negative and HER2+ breast cancer subtypes.

OBJECTIVES: Early diagnosis of triple-negative (TN) and human epidermal growth factor receptor 2 positive (HER2+) breast cancer is important due to its increased risk of micrometastatic spread necessitating early treatment and for guiding targeted therapies. This study aimed to evaluate the diagnostic performance of machine learning (ML) classification of newly diagnosed breast masses into TN versus non-TN (NTN) and HER2+ versus HER2 negative (HER2-) breast cancer, using radiomic features extracted from grayscale ultrasound (US) b-mode images. MATERIALS AND METHODS: A retrospective chart review identified 88 female patients who underwent diagnostic breast US imaging, had confirmation of invasive malignancy on pathology and receptor status determined on immunohistochemistry available. The patients were classified as TN, NTN, HER2+ or HER2- for ground-truth labelling. For image analysis, breast masses were manually segmented by a breast radiologist. Radiomic features were extracted per image and used for predictive modelling. Supervised ML classifiers included: logistic regression, k-nearest neighbour, and Naïve Bayes. Classification performance measures were calculated on an independent (unseen) test set. The area under the receiver operating characteristic curve (AUC), sensitivity (%), and specificity (%) were reported for each classifier. RESULTS: The logistic regression classifier demonstrated the highest AUC: 0.824 (sensitivity: 81.8%, specificity: 74.2%) for the TN sub-group and 0.778 (sensitivity: 71.4%, specificity: 71.6%) for the HER2 sub-group. CONCLUSION: ML classifiers demonstrate high diagnostic accuracy in classifying TN versus NTN and HER2+ versus HER2- breast cancers using US images. Identification of more aggressive breast cancer subtypes early in the diagnostic process could help achieve better prognoses by prioritizing clinical referral and prompting adequate early treatment.

Extrahepatic metastases occur in a minority of hepatocellular carcinoma patients treated with locoregional therapies: analyzing patterns of progression in 285 patients.

UNLABELLED: Although most cancers are considered predominantly systemic processes, this may not hold true for hepatocellular carcinoma (HCC). The literature regarding patterns of progression of HCC (local versus systemic) has been relatively sparse. Our objectives were to: (1) analyze patterns of progression in HCC patients presenting with intrahepatic disease from initial treatment until death, and (2) identify clinically relevant risk factors for the development of metastases. Over a 9-year period, 285 patients treated with transarterial locoregional therapies underwent scheduled imaging follow-up from treatment until death and were categorized by pattern of progression: (i) intrahepatic (increased tumor enhancement/size, development/progression of vascular invasion, new hepatic lesions) progression or (ii) extrahepatic (adrenal/bone/lung/lymph node) metastases. Uni/multivariate analyses assessing the risk factors for the development of metastases were performed. The median time from last scan to death was 2.4 months (interquartile range: 1.3-4.8 months). The time to development of metastases, vascular invasion, and/or new lesions was 13.8 months (confidence interval: 11.3-17.7 months). Of the 209 patients followed until death, only 50 developed extrahepatic metastases (24%). Multivariate analyses identified age <65 years (P = 0.038), alpha-fetoprotein >200 ng/mL (P = 0.04), and vascular invasion (P = 0.017) as significant predictors of metastases development. CONCLUSION: Knowledge of the risk factors associated with the development of metastases may help guide assessment of patient prognosis. Because 76% of patients presenting with local disease treated with locoregional therapies die without developing extrahepatic metastases, the notion of HCC as a systemic disease, as detected by imaging, may be reconsidered.

Radiographic response to locoregional therapy in hepatocellular carcinoma predicts patient survival times.

BACKGROUND & AIMS: It is not clear whether survival times of patients with hepatocellular carcinoma (HCC) are associated with their response to therapy. We analyzed the association between tumor response and survival times of patients with HCC who were treated with locoregional therapies (LRTs) (chemoembolization and radioembolization). METHODS: Patients received LRTs over a 9-year period (n = 463). Patients with metastases, portal venous thrombosis, or who had received transplants were excluded; 159 patients with Child-Pugh B7 or lower were analyzed. Response (based on European Association for the Study of the Liver [EASL] and World Health Organization [WHO] criteria) was associated with survival times using the landmark, risk-of-death, and Mantel-Byar methodologies. In a subanalysis, survival times of responders were compared with those of patients with stable disease and progressive disease. RESULTS: Based on 6-month data, in landmark analysis, responders survived longer than nonresponders (based on EASL but not WHO criteria: P = .002 and .0694). The risk of death was also lower for responders (based on EASL but not WHO criteria: P = .0463 and .707). Landmark analysis of 12-month data showed that responders survived longer than nonresponders (P < .0001 and .004, based on EASL and WHO criteria, respectively). The risk of death was lower for responders (P = .0132 and .010, based on EASL and WHO criteria, respectively). By the Mantel-Byar method, responders had longer survival than nonresponders, based on EASL criteria (P < .0001; P = .596 with WHO criteria). In the subanalysis, responders lived longer than patients with stable disease or progressive disease. CONCLUSIONS: Radiographic response to LRTs predicts survival time. EASL criteria for response more consistently predicted survival times than WHO criteria. The goal of LRT should be to achieve a radiologic response, rather than to stabilize disease.

Radioembolization results in longer time-to-progression and reduced toxicity compared with chemoembolization in patients with hepatocellular carcinoma.

BACKGROUND & AIMS: Chemoembolization is one of several standards of care treatment for hepatocellular carcinoma (HCC). Radioembolization with Yttrium-90 microspheres is a novel, transarterial approach to radiation therapy. We performed a comparative effectiveness analysis of these therapies in patients with HCC. METHODS: We collected data from 463 patients who were treated with transarterial locoregional therapies (chemoembolization or radioembolization) over a 9-year period. We excluded patients who were not appropriate for comparison and analyzed data from 245 (122 who received chemoembolization and 123 who received radioembolization). Patients were followed for signs of toxicity; all underwent imaging analysis at baseline and follow-up time points. Overall survival was the primary outcome measure. Secondary outcomes included safety, response rate, and time-to-progression. Uni- and multivariate analyses were performed. RESULTS: Abdominal pain and increased transaminase activity were more frequent following chemoembolization (P < .05). There was a trend that patients treated with radioembolization had a higher response rate than with chemoembolization (49% vs 36%, respectively, P = .104). Although time-to-progression was longer following radioembolization than chemoembolization (13.3 months vs 8.4 months, respectively, P = .046), median survival times were not statistically different (20.5 months vs 17.4 months, respectively, P = .232). Among patients with intermediate-stage disease, survival was similar between groups that received chemoembolization (17.5 months) and radioembolization (17.2 months, P = .42). CONCLUSIONS: Patients with HCC treated by chemoembolization or radioembolization with Yttrium-90 microspheres had similar survival times. Radioembolization resulted in longer time-to-progression and less toxicity than chemoembolization. Post hoc analyses of sample size indicated that a randomized study with > 1000 patients would be required to establish equivalence of survival times between patients treated with these two therapies.

Chemoembolization for hepatocellular carcinoma: comprehensive imaging and survival analysis in a 172-patient cohort.

PURPOSE: To determine comprehensive imaging and long-term survival outcome following chemoembolization for hepatocellular carcinoma (HCC). MATERIALS AND METHODS: One hundred seventy-two patients with HCC treated with chemoembolization were studied retrospectively in an institutional review board approved protocol; this study was HIPAA compliant. Baseline laboratory and imaging characteristics were obtained. Clinical and laboratory toxicities following treatment were assessed. Imaging characteristics following chemoembolization were evaluated to determine response rates (size and necrosis) and time to progression (TTP). Survival from the time of first chemoembolization treatment was calculated. Subanalyses were performed by stratifying the population according to Child-Pugh, United Network for Organ Sharing, and Barcelona Clinic for Liver Cancer (BCLC) staging systems. RESULTS: Cirrhosis was present in 157 patients (91%); portal hypertension was present in 139 patients (81%). Eleven patients (6%) had metastases at baseline. Portal vein thrombosis was present in 11 patients (6%). Fifty-five percent of patients experienced some form of toxicity following treatment; 21% developed grade 3 or 4 bilirubin toxicity. Post-chemoembolization response was seen in 31% and 64% of patients according to size and necrosis criteria, respectively. Median TTP was 7.9 months (95% confidence interval: 7.1, 9.4) but varied widely by stage. Median survival was significantly different between patients with BCLC stages A, B, and C disease (stage A, 40.0 months; B, 17.4 months; C, 6.3 months; P < .0001). CONCLUSION: The determination of TTP and survival in patients with HCC is confounded by tumor biology and background cirrhosis; chemoembolization was shown to be a safe and effective therapy in patients with HCC.