Gene expression assay and Watson for Oncology for optimization of treatment in ER-positive, HER2-negative breast cancer

Use and accuracy of decision support systems using artificial intelligence for tumor diseases: a systematic review and meta-analysis

Background

For therapy planning in cancer patients multidisciplinary team meetings (MDM) are mandatory. Due to the high number of cases being discussed and significant workload of clinicians, Clinical Decision Support System (CDSS) may improve the clinical workflow.

Methods

This review and meta-analysis aims to provide an overview of the systems utilized and evaluate the correlation between a CDSS and MDM.

Results

A total of 31 studies were identified for final analysis. Analysis of different cancers shows a concordance rate (CR) of 72.7% for stage I-II and 73.4% for III-IV. For breast carcinoma, CR for stage I-II was 72.8% and for III-IV 84.1%, P≤ 0.00001. CR for colorectal carcinoma is 63% for stage I-II and 67% for III-IV, for gastric carcinoma 55% and 45%, and for lung carcinoma 85% and 83% respectively, all P>0.05. Analysis of SCLC and NSCLC yields a CR of 94,3% and 82,7%, P=0.004 and for adenocarcinoma and squamous cell carcinoma in lung cancer a CR of 90% and 86%, P=0.02.

Conclusion

CDSS has already been implemented in clinical practice, and while the findings suggest that its use is feasible for some cancers, further research is needed to fully evaluate its effectiveness.

A Deep Neural Network for Estimating Low-Density Lipoprotein Cholesterol From Electronic Health Records: Real-Time Routine Clinical Application

Background

Previously, we constructed a deep neural network (DNN) model to estimate low-density lipoprotein cholesterol (LDL-C).

Objective

To routinely provide estimated LDL-C levels, we applied the aforementioned DNN model to an electronic health record (EHR) system in real time (deep LDL-EHR).

Methods

The Korea National Health and Nutrition Examination Survey and the Wonju Severance Christian Hospital (WSCH) datasets were used as training and testing datasets, respectively. We measured our proposed model’s performance by using 5 indices, including bias, root mean-square error, P10-P30, concordance, and correlation coefficient. For transfer learning (TL), we pretrained the DNN model using a training dataset and fine-tuned it using 30% of the testing dataset.

Results

Based on 5 accuracy criteria, deep LDL-EHR generated inaccurate results compared with other methods for LDL-C estimation. By comparing the training and testing datasets, we found an overfitting problem. We then revised the DNN model using the TL algorithms and randomly selected subdata from the WSCH dataset. Therefore, the revised model (DNN+TL) exhibited the best performance among all methods.

Conclusions

Our DNN+TL is expected to be suitable for routine real-time clinical application for LDL-C estimation in a clinical laboratory.

Multicenter randomized comparative trial of Micromedex, Micromedex with Watson, or Google to answer drug information questions

Objective:

The purpose of this study was to compare pharmacy students’ ability to correctly answer drug information questions using Micromedex with Watson, Micromedex without Watson, or Google.

Methods:

This multicenter randomized trial compared pharmacy student responses to drug information questions using Micromedex with Watson, Micromedex without Watson, or Google from January to March of 2020. First- to fourth-year pharmacy students at two institutions were included. The primary outcome was the number of correct answers. Secondary outcomes were the time taken to answer the questions and differences in number of correct answers by pharmacy student year and institution.

Results:

The analysis included 162 participants: 52 students in the Micromedex group, 51 students in the Watson group, and 59 students in the Google group. There was a significant difference among groups in the total number of questions answered correctly (p=0.02). Post-hoc analysis revealed that participants in the Micromedex group answered more questions correctly than those in the Google group (p=0.015). There were no significant differences between Micromedex and Watson groups (p=0.52) or between Watson and Google groups (p=0.22). There was also no difference in time to complete the questions among groups (p=0.72).

Conclusion:

Utilizing Google did not save students time and led to more incorrect answers. These findings suggest that health care educators and health sciences librarians should further reinforce training on the appropriate use of drug information resources.

Physician Perception and Satisfaction with Artificial Intelligence in Cancer Treatment: The Watson for Oncology Experience and Implications for Low-Middle Income Countries (Preprint)

As technology continues to improve, health care systems have the opportunity to use a variety of innovative tools for decision-making, including artificial intelligence (AI) applications. However, there has been little research on the feasibility and efficacy of integrating AI systems into real-world clinical practice, especially from the perspectives of clinicians who use such tools. In this paper, we review physicians’ perceptions of and satisfaction with an AI tool, Watson for Oncology, which is used for the treatment of cancer. Watson for Oncology has been implemented in several different settings, including Brazil, China, India, South Korea, and Mexico. By focusing on the implementation of an AI-based clinical decision support system for oncology, we aim to demonstrate how AI can be both beneficial and challenging for cancer management globally and particularly for low-middle–income countries. By doing so, we hope to highlight the need for additional research on user experience and the unique social, cultural, and political barriers to the successful implementation of AI in low-middle–income countries for cancer care.

A Roadmap towards Breast Cancer Therapies Supported by Explainable Artificial Intelligence

In recent years personalized medicine reached an increasing importance, especially in the design of oncological therapies. In particular, the development of patients’ profiling strategies suggests the possibility of promising rewards. In this work, we present an explainable artificial intelligence (XAI) framework based on an adaptive dimensional reduction which (i) outlines the most important clinical features for oncological patients’ profiling and (ii), based on these features, determines the profile, i.e., the cluster a patient belongs to. For these purposes, we collected a cohort of 267 breast cancer patients. The adopted dimensional reduction method determines the relevant subspace where distances among patients are used by a hierarchical clustering procedure to identify the corresponding optimal categories. Our results demonstrate how the molecular subtype is the most important feature for clustering. Then, we assessed the robustness of current therapies and guidelines; our findings show a striking correspondence between available patients’ profiles determined in an unsupervised way and either molecular subtypes or therapies chosen according to guidelines, which guarantees the interpretability characterizing explainable approaches to machine learning techniques. Accordingly, our work suggests the possibility to design data-driven therapies to emphasize the differences observed among the patients.

A Proposal of Quantum-Inspired Machine Learning for Medical Purposes: An Application Case

Learning tasks are implemented via mappings of the sampled data set, including both the classical and the quantum framework. Biomedical data characterizing complex diseases such as cancer typically require an algorithmic support for clinical decisions, especially for early stage tumors that typify breast cancer patients, which are still controllable in a therapeutic and surgical way. Our case study consists of the prediction during the pre-operative stage of lymph node metastasis in breast cancer patients resulting in a negative diagnosis after clinical and radiological exams. The classifier adopted to establish a baseline is characterized by the result invariance for the order permutation of the input features, and it exploits stratifications in the training procedure. The quantum one mimics support vector machine mapping in a high-dimensional feature space, yielded by encoding into qubits, while being characterized by complexity. Feature selection is exploited to study the performances associated with a low number of features, thus implemented in a feasible time. Wide variations in sensitivity and specificity are observed in the selected optimal classifiers during cross-validations for both classification system types, with an easier detection of negative or positive cases depending on the choice between the two training schemes. Clinical practice is still far from being reached, even if the flexible structure of quantum-inspired classifier circuits guarantees further developments to rule interactions among features: this preliminary study is solely intended to provide an overview of the particular tree tensor network scheme in a simplified version adopting just product states, as well as to introduce typical machine learning procedures consisting of feature selection and classifier performance evaluation.

Predicting of Sentinel Lymph Node Status in Breast Cancer Patients with Clinically Negative Nodes: A Validation Study

Simple Summary

Sentinel lymph node biopsy procedure is time consuming and expensive, but it is still the intra-operative exam capable of the best performance. However, sometimes, surgery is achieved without a clear diagnosis, so clinical decision support systems developed with artificial intelligence techniques are essential to assist current diagnostic procedures. In this work, we evaluated the usefulness of a CancerMath tool in the sentinel lymph nodes positivity prediction for clinically negative patients. We tested it on 993 patients referred to our institute characterized by sentinel lymph node status, tumor size, age, histologic type, grading, expression of estrogen receptor, progesterone receptor, HER2, and Ki-67. By training the CancerMath (CM) model on our dataset, we reached a sensitivity value of 72%, whereas the online one was 46%, despite a specificity reduction. It was found the addiction of the prognostic factors Her2 and Ki67 could help improve performances on the classification of particular types of patients.

Abstract

In the absence of lymph node abnormalities detectable on clinical examination or imaging, the guidelines provide for the dissection of the first axillary draining lymph nodes during surgery. It is not always possible to arrive at surgery without diagnostic doubts, and machine learning algorithms can support clinical decisions. The web calculator CancerMath (CM) allows you to estimate the probability of having positive lymph nodes valued on the basis of tumor size, age, histologic type, grading, expression of estrogen receptor, and progesterone receptor. We collected 993 patients referred to our institute with clinically negative results characterized by sentinel lymph node status, prognostic factors defined by CM, and also human epidermal growth factor receptor 2 (HER2) and Ki-67. Area Under the Curve (AUC) values obtained by the online CM application were comparable with those obtained after training its algorithm on our database. Nevertheless, by training the CM model on our dataset and using the same feature, we reached a sensitivity median value of 72%, whereas the online one was equal to 46%, despite a specificity reduction. We found that the addition of the prognostic factors Her2 and Ki67 could help improve performances on the classification of particular types of patients with the aim of reducing as much as possible the false positives that lead to axillary dissection. As showed by our experimental results, it is not particularly suitable for use as a support instrument for the prediction of metastatic lymph nodes on clinically negative patients.

Concordance between treatment recommendations provided by IBM Watson for Oncology and a multidisciplinary tumor board for breast cancer in China

OBJECTIVE

Watson for Oncology (WFO), an artificial intelligence from IBM Corporation, can provide a treatment plan by analyzing patient's disease characteristics. The present study was performed to examine the concordance between treatment recommendations proposed by WFO and the multidisciplinary tumor board at our center. The aim was to explore the feasibility of using WFO for breast cancer cases in China and to ascertain the ways to make WFO more suitable for Chinese patients with breast cancer.

METHODS

Data from 302 breast cancer patients treated at the Second Affiliated Hospital of Xi'an Jiaotong University between October 2016 and February 2018 was retrieved and retrospectively analyzed by WFO. The recommendations were divided into 'recommended', 'considered' and 'not recommended' groups. Results were considered concordant when oncologists' recommendations were categorized as 'recommended' or 'for consideration' by WFO.

RESULTS

The concordance rate of 200 subjects with postoperative adjuvant therapy was 77%. However, the rate was 27.5% in the remaining 102 cases with metastatic disease receiving either first-line or no treatment. Further analysis demonstrated that inconsistencies were mainly due to different choices of chemotherapy regimens. Subgroup study indicates that tumor stage, receptor status and age also had influences at the concordance rate.

CONCLUSION

The results of this study suggest that WFO is a promising artificial intelligence system for the treatment of breast cancer. These findings can also serve as a reference framework for the inclusion of artificial intelligence in the ongoing medical reform in China.

Comparison of Annotation Services for Next-Generation Sequencing in a Large-Scale Precision Oncology Program

PURPOSE

Next-generation sequencing (NGS) multigene panel testing has become widespread, including the Veterans Affairs (VA), through the VA National Precision Oncology Program (NPOP). The interpretation of genomic alterations remains a bottleneck for realizing precision medicine. We sought to examine the concordance for pathogenicity determination and clinical actionability of annotation services in NPOP.

METHODS

Unique gene variants were generated from NGS gene panel results using two sequencing services. For each unique gene variant, annotations were provided through N-of-One (NoO), IBM Watson for Genomics (WfG), and OncoKB. Annotations for pathogenicity (all three sources) and actionability (WfG and OncoKB) were examined for concordance. Cohen's kappa statistic was calculated to measure agreement between annotation services.

RESULTS

Among 1,227 NGS results obtained between 2015 and 2017, 1,388 unique variants were identified in 117 genes. The genes with the largest number of variants included TP53 (270), STK11 (92), and CDKN2A (81). The most common cancer type was lung adenocarcinoma (440), followed by colon adenocarcinoma (113). For pathogenic and likely pathogenic variants, there was 30% agreement between WfG and NoO (kappa, -0.26), 76% agreement between WfG and OncoKB (kappa, 0.22), and 42% agreement between NoO and OncoKB (kappa, -0.07). For level 1 drug actionability of gene variant-diagnosis combinations, there was moderate agreement between WfG and OncoKB (96.9%; kappa, 0.44), with 27 combinations identified as level 1 by both services, 58 by WfG alone, and 6 variants by OncoKB alone.

CONCLUSION

There is substantial variability in pathogenicity assessment of NGS variants in solid tumors by annotation services. In addition, there was only moderate agreement in level 1 therapeutic actionability recommendations between WfG and OncoKB. Improvement in the precision of NGS multigene panel annotation is needed.

Concordance Study Between IBM Watson for Oncology and Real Clinical Practice for Cervical Cancer Patients in China: A Retrospective Analysis

Watson for Oncology (WFO) is a artificial intelligence clinical decision-support system with evidence-based treatment options for oncologists. WFO has been gradually used in China, but limited reports on whether WFO is suitable for Chinese patients. This study aims to investigate the concordance of treatment options between WFO and real clinical practice for Cervical cancer patients retrospectively. We retrospectively enrolled 300 cases of cervical cancer patients. WFO provides treatment options for 246 supported cases. Real clinical practice were defined as concordant if treatment options were designated "recommended" or "for consideration" by WFO. Concordance of treatment option between WFO and real clinical practice was analyzed statistically. The treatment concordance between WFO and real clinical practice occurred in 72.8% (179/246) of cervical cancer cases. Logistic regression analysis showed that rural registration residences, advanced age, poor ECOG performance status, stages II-IV disease have a remarkable impact on consistency. The main reasons attributed to the 27.2% (67/246) of the discordant cases were the substitution of nedaplatin for cisplatin, reimbursement plan of bevacizumab, surgical preference, and absence of neoadjuvant/adjuvant chemotherapy and PD-1/PD-L1 antibodies recommendations. WFO recommendations were in 72.8% of concordant with real clinical practice for cervical cancer patients in China. However, several localization and individual factors limit its wider application. So, WFO could be an essential tool but it cannot currently replace oncologists. To be rapidly and fully apply to cervical cancer patients in China, accelerate localization and improvement were needed for WFO.

Artificial Intelligence and Machine Learning: A New Disruptive Force in Orthopaedics

Orthopaedics as a surgical discipline requires a combination of good clinical acumen, good surgical skill, a reasonable physical strength and most of all, good understanding of technology. The last few decades have seen rapid adoption of new technologies into orthopaedic practice, power tools, new implants, CAD-CAM design, 3-D printing, additive manufacturing just to name a few. The new disruption in orthopaedics in the current time and era is undoubtedly the advent of artificial intelligence and robotics. As these technologies take root and innovative applications continue to be incorporated into the main-stream orthopedics, as we know it today, it is imperative to look at and understand the basics of artificial intelligence and what work is being done in the field today. This article takes the form of a loosely structured narrative review and will introduce the reader to key concepts in the field of artificial intelligence as well as some of the directions in application of the same in orthopaedics. Some of the recent work has been summarised and we present our viewpoint at the conclusion as to why we must consider artificial intelligence as a disrupting positive influence on orthopaedic surgery.