Role of Machine Learning and Artificial Intelligence in Interventional Oncology.

PURPOSE OF REVIEW: The purpose of this review is to highlight the current role of machine learning and artificial intelligence and in the field of interventional oncology. RECENT FINDINGS: With advancements in technology, there is a significant amount of research regarding the application of artificial intelligence and machine learning in medicine. Interventional oncology is a field that can benefit greatly from this research through enhanced image analysis and intraprocedural guidance. These software developments can increase detection of cancers through routine screening and improve diagnostic accuracy in classifying tumors. They may also aid in selecting the most effective treatment for the patient by predicting outcomes based on a combination of both clinical and radiologic factors. Furthermore, machine learning and artificial intelligence can advance intraprocedural guidance for the interventional oncologist through more accurate needle tracking and image fusion technology. This minimizes damage to nearby healthy tissue and maximizes treatment of the tumor. While there are several exciting developments, this review also discusses limitations before incorporating machine learning and artificial intelligence in the field of interventional oncology. These include data capture and processing, lack of transparency among developers, validating models, integrating workflow, and ethical challenged. In summary, machine learning and artificial intelligence have the potential to positively impact interventional oncologists and how they provide cancer care treatments.

Quantitative peritumoral magnetic resonance imaging fingerprinting improves machine learning-based prediction of overall survival in colorectal cancer.

Aim: To investigate magnetic resonance imaging (MRI)-based peritumoral texture features as prognostic indicators of survival in patients with colorectal liver metastasis (CRLM). Methods: From 2007-2015, forty-eight patients who underwent MRI within 3 months prior to initiating treatment for CRLM were identified. Clinicobiological prognostic variables were obtained from electronic medical records. Ninety-four metastatic hepatic lesions were identified on T1-weighted post-contrast images and volumetrically segmented. A total of 112 radiomic features (shape, first-order, texture) were derived from a 10 mm region surrounding each segmented tumor. A random forest model was applied, and performance was tested by receiver operating characteristic (ROC). Kaplan-Meier analysis was utilized to generate the survival curves. Results: Forty-eight patients (male:female = 23:25, age 55.3 years ± 18 years) were included in the study. The median lesion size was 25.73 mm (range 8.5-103.8 mm). Microsatellite instability was low in 40.4% (38/94) of tumors, with Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) mutation detected in 68 out of 94 (72%) tumors. The mean survival was 35 months ± 21 months, and local disease progression was observed in 35.5% of patients. Univariate regression analysis identified 42 texture features [8 first order, 5 gray level dependence matrix (GLDM), 5 gray level run time length matrix (GLRLM), 5 gray level size zone matrix (GLSZM), 2 neighboring gray tone difference matrix (NGTDM), and 17 gray level co-occurrence matrix (GLCM)] independently associated with metastatic disease progression (P < 0.03). The random forest model achieved an area under the curve (AUC) of 0.88. Conclusions: MRI-based peritumoral heterogeneity features may serve as predictive biomarkers for metastatic disease progression and patient survival in CRLM.

Disparities in access to endovenous treatment options in chronic lower extremity superficial venous insufficiency: A national 7-year analysis.

OBJECTIVE: The goal of this study was to analyze trends in treatment access for chronic superficial venous disease and to identify disparities in care. METHODS: This retrospective study was exempt from institutional review board approval. The American College of Surgeon National Surgical Quality Improvement Program database was used to identify patients who underwent vein stripping (VS) and endovenous procedures for treatment of chronic superficial venous disease. Endovenous options included radiofrequency ablation (RFA) and laser ablation. Data was available from 2011 to 2018 and demographic information was extracted for each patient identified by Current Procedural Terminology codes. For all racial and ethnic groups, trend lines were plotted, and the relative rate of change was determined within each specified demographic. RESULTS: There were 21,025 patients included in the analysis. The overall mean age was 54.2 years, and the majority of patients were female (64.8%). In total, 27.9%, 55.2%, and 16.9% patients underwent VS, RFA, and laser ablation, respectively. Patients who received laser ablation were older (P < .001). Hispanic ethnicity was associated with significantly lower odds of receiving endovascular thermal ablation (EVTA) over VS (odds ratio [OR], 0.71; 95% confidence interval [CI], 0.64-0.78; P < .001). American Indian/Alaska Native patients were more likely to receive EVTA over VS (OR, 4.02; 95% CI, 2.48-6.86); similarly, Native Hawaiian/Pacific Islander patients were more likely to receive EVTA over VS, although this difference was not statistically significant (OR, 1.44; 95% CI, 0.93-2.27). On multinomial regression, Hispanic patients were less likely to receive RFA over VS, whereas American Indian/Alaskan Native patients were more likely to receive RFA over VS. In all racial and ethnic groups, the percentage of endovenous procedures increased, whereas vein stripping decreased. CONCLUSIONS: Based on a hospital-based dataset, demographic indicators, including age, sex, race, and ethnicity, are associated with differences in endovenous treatments for chronic superficial venous insufficiency suggesting disparities in obtaining minimally invasive treatment options among certain patient groups.

Machine Learning Improves the Prediction of Responses to Immune Checkpoint Inhibitors in Metastatic Melanoma.

Pretreatment LDH is a standard prognostic biomarker for advanced melanoma and is associated with response to ICI. We assessed the role of machine learning-based radiomics in predicting responses to ICI and in complementing LDH for prognostication of metastatic melanoma. From 2008-2022, 79 patients with 168 metastatic hepatic lesions were identified. All patients had arterial phase CT images 1-month prior to initiation of ICI. Response to ICI was assessed on follow-up CT at 3 months using RECIST criteria. A machine learning algorithm was developed using radiomics. Maximum relevance minimum redundancy (mRMR) was used to select features. ROC analysis and logistic regression analyses evaluated performance. Shapley additive explanations were used to identify the variables that are the most important in predicting a response. mRMR selection revealed 15 features that are associated with a response to ICI. The machine learning model combining both radiomics features and pretreatment LDH resulted in better performance for response prediction compared to models that included radiomics or LDH alone (AUC of 0.89 (95% CI: [0.76-0.99]) vs. 0.81 (95% CI: [0.65-0.94]) and 0.81 (95% CI: [0.72-0.91]), respectively). Using SHAP analysis, LDH and two GLSZM were the most predictive of the outcome. Pre-treatment CT radiomic features performed equally well to serum LDH in predicting treatment response.

Machine Learning-Based Radiomic Features on Pre-Ablation MRI as Predictors of Pathologic Response in Patients with Hepatocellular Carcinoma Who Underwent Hepatic Transplant.

BACKGROUND: The aim was to investigate the role of pre-ablation tumor radiomics in predicting pathologic treatment response in patients with early-stage hepatocellular carcinoma (HCC) who underwent liver transplant. METHODS: Using data collected from 2005-2015, we included adult patients who (1) had a contrast-enhanced MRI within 3 months prior to ablation therapy and (2) underwent liver transplantation. Demographics were obtained for each patient. The treated hepatic tumor volume was manually segmented on the arterial phase T1 MRI images. A vector with 112 radiomic features (shape, first-order, and texture) was extracted from each tumor. Feature selection was employed through minimum redundancy and maximum relevance using a training set. A random forest model was developed based on top radiomic and demographic features. Model performance was evaluated by ROC analysis. SHAP plots were constructed in order to visualize feature importance in model predictions. RESULTS: Ninety-seven patients (117 tumors, 31 (32%) microwave ablation, 66 (68%) radiofrequency ablation) were included. The mean model for end-stage liver disease (MELD) score was 10.5 ± 3. The mean follow-up time was 336.2 ± 179 days. Complete response on pathology review was achieved in 62% of patients at the time of transplant. Incomplete pathologic response was associated with four features: two first-order and two GLRM features using univariate logistic regression analysis (p < 0.05). The random forest model included two radiomic features (diagnostics maximum and first-order maximum) and four clinical features (pre-procedure creatinine, pre-procedure albumin, age, and gender) achieving an AUC of 0.83, a sensitivity of 82%, a specificity of 67%, a PPV of 69%, and an NPV of 80%. CONCLUSIONS: Pre-ablation MRI radiomics could act as a valuable imaging biomarker for the prediction of tumor pathologic response in patients with HCC.

Functional Effects of Delivering Human Mesenchymal Stem Cell-Seeded Biological Sutures to an Infarcted Heart.

Stem cell therapy has the potential to improve cardiac function after myocardial infarction (MI); however, existing methods to deliver cells to the myocardium, including intramyocardial injection, suffer from low engraftment rates. In this study, we used a rat model of acute MI to assess the effects of human mesenchymal stem cell (hMSC)-seeded fibrin biological sutures on cardiac function at 1 week after implant. Biological sutures were seeded with quantum dot (Qdot)-loaded hMSCs for 24 h before implantation. At 1 week postinfarct, the heart was imaged to assess mechanical function in the infarct region. Regional parameters assessed were regional stroke work (RSW) and systolic area of contraction (SAC) and global parameters derived from the pressure waveform. MI (n = 6) significantly decreased RSW (0.026 ± 0.011) and SAC (0.022 ± 0.015) when compared with sham operation (RSW: 0.141 ± 0.009; SAC: 0.166 ± 0.005, n = 6) (p < 0.05). The delivery of unseeded biological sutures to the infarcted hearts did not change regional mechanical function compared with the infarcted hearts (RSW: 0.032 ± 0.004, SAC: 0.037 ± 0.008, n = 6). The delivery of hMSC-seeded sutures exerted a trend toward increase of regional mechanical function compared with the infarcted heart (RSW: 0.057 ± 0.011; SAC: 0.051 ± 0.014, n = 6). Global function showed no significant differences between any group (p > 0.05); however, there was a trend toward improved function with the addition of either unseeded or seeded biological suture. Histology demonstrated that Qdot-loaded hMSCs remained present in the infarcted myocardium after 1 week. Analysis of serial sections of Masson's trichrome staining revealed that the greatest infarct size was in the infarct group (7.0% ± 2.2%), where unseeded (3.8% ± 0.6%) and hMSC-seeded (3.7% ± 0.8%) suture groups maintained similar infarct sizes. Furthermore, the remaining suture area was significantly decreased in the unseeded group compared with that in the hMSC-seeded group (p < 0.05). This study demonstrated that hMSC-seeded biological sutures are a method to deliver cells to the infarcted myocardium and have treatment potential.