Sweden
April 1, 2022Corti
April 1, 2022Immunology
Artificial intelligence , and machine learning are rapidly changing the study of cancer immunology, harboring great implications for the future of cancer treatment...
Artificial Intelligence to Predict Treatment Success from Early CT scans
Researchersat Columbia University’s oncology department have developed a way to predict immunotherapy outcomes in patients with melanoma using artificial intelligence. They combined AI with standard-of-care imaging using a large database to develop a machine learning algorithm that creates a “radiomic” signature based on a patients’ Computed Tomography (CT). This algorithm considers many radiological factors like tumor volume, heterogeneity, distribution in diseased sites, and pixel intensity variation in the image. The researchers validated the algorithm with data from 287 patients with advanced melanoma who participated in different multicenter clinical trials with various immunotherapy protocols. The radiomic signature, which used CT images received at baseline and 3-month follow-up, estimated overall survival at six months with high accuracy. It outperformed the traditional method based on tumor diameter, commonly used in clinical trials to assess treatment efficacy.(1)
Researchersat Columbia University’s oncology department have developed a way to predict immunotherapy outcomes in patients with melanoma using artificial intelligence. They combined AI with standard-of-care imaging using a large database to develop a machine learning algorithm that creates a “radiomic” signature based on a patients’ Computed Tomography (CT). This algorithm considers many radiological factors like tumor volume, heterogeneity, distribution in diseased sites, and pixel intensity variation in the image.
The researchers validated the algorithm with data from 287 patients with advanced melanoma who participated in different multicenter clinical trials with various immunotherapy protocols. The radiomic signature, which used CT images received at baseline and 3-month follow-up, estimated overall survival at six months with high accuracy. It outperformed the traditional method based on tumor diameter, commonly used in clinical trials to assess treatment efficacy.(1)
A common cancer immunotherapy target might also hold the key to tackling obesity
The PD-L1 protein is a popular target for cancer therapies as its expression in tumors allows them to silence immune responses. Guided by AI, researchers from the University of Erlangen-Nuremberg (FAU) and Trinity College Dublin in Germany and Ireland have unraveled new evidence that the protein may also defend against obesity. This research demonstrated that PD-L1 expression on dendritic cells controls immune responses and limits diet-induced obesity. PD-L1 was removed from a mouse model of high-calorie diet-induced obesity. Compared with the group of normal mice, the engineered mice became severely obese and had increased insulin resistance.
Another important finding is that PD-L1 was noted to be increased in the visceral fat of obese human subjects. PD-L1’s role in fat tissue translates to important implications for patients with obesity and cancer receiving PD-1/PD-L1 checkpoint inhibitors. These implications include the effect on immune response and side effects caused by the modulation of immune cells in fat tissue.(2)
The PD-L1 protein is a popular target for cancer therapies as its expression in tumors allows them to silence immune responses. Guided by AI, researchers from the University of Erlangen-Nuremberg (FAU) and Trinity College Dublin in Germany and Ireland have unraveled new evidence that the protein may also defend against obesity.
This research demonstrated that PD-L1 expression on dendritic cells controls immune responses and limits diet-induced obesity. PD-L1 was removed from a mouse model of high-calorie diet-induced obesity. Compared with the group of normal mice, the engineered mice became severely obese and had increased insulin resistance.The researchers validated the algorithm with data from 287 patients with advanced melanoma who participated in different multicenter clinical trials with various immunotherapy protocols. The radiomic signature, which used CT images received at baseline and 3-month follow-up, estimated overall survival at six months with high accuracy. It outperformed the traditional method based on tumor diameter, commonly used in clinical trials to assess treatment efficacy.(1)Another important finding is that PD-L1 was noted to be increased in the visceral fat of obese human subjects. PD-L1’s role in fat tissue translates to important implications for patients with obesity and cancer receiving PD-1/PD-L1 checkpoint inhibitors. These implications include the effect on immune response and side effects caused by the modulation of immune cells in fat tissue.(2)
AI tool as novel biomarker for immunotherapy response
Lunit, a South Korea-based company devoted to developing advanced software for medical data analysis, has announced the publication of its study validating a new marker. The investigators proved the effectiveness of Lunit’s AI biomarker, Lunit SCOPE IO, in predicting clinical outcomes and immunotherapy results in patients with advanced non-small-cell lung cancer. Dr. Tony Mok from the Chinese University of Hong Kong, a co-author of this study, mentioned that they adopted AI technology to determine the tumor immune phenotype and predict treatment outcomes in two large cohorts of patients with advanced non-small-cell lung cancer.(3)
Lunit, a South Korea-based company devoted to developing advanced software for medical data analysis, has announced the publication of its study validating a new marker. The investigators proved the effectiveness of Lunit’s AI biomarker, Lunit SCOPE IO, in predicting clinical outcomes and immunotherapy results in patients with advanced non-small-cell lung cancer.
Dr. Tony Mok from the Chinese University of Hong Kong, a co-author of this study, mentioned that they adopted AI technology to determine the tumor immune phenotype and predict treatment outcomes in two large cohorts of patients with advanced non-small-cell lung cancer.(3)
Artificial intelligence platform shows potential for thyroid cancer screening
The American Society for Radiation Oncology (ASTRO) announced an artificial intelligence (AI) model that accurately detects thyroid cancer and predicts pathological, genomic, and immunological outcomes. This AI model is based on the study of routine ultrasound images and incorporates multiple machine learning methods. The new AI model could present a low-cost, non-invasive option for screening, staging, and personalized immunology treatment planning for the disease.
Researchers obtained around 1,346 thyroid nodule images through routine diagnostic ultrasound from 784 patients to train and validate the AI platform. The ultrasound images were divided into datasets, one for internal training and validation and one for external validation. Malignancy was confirmed with different samples obtained by fine needle biopsy. Pathological staging and mutational status were established with operative reports and genomic sequencing. A multimodal platform utilizing these techniques accurately predicted 98.7% of thyroid nodule malignancies in the internal dataset, significantly outperforming individual AI modalities.(4)
The American Society for Radiation Oncology (ASTRO) announced an artificial intelligence (AI) model that accurately detects thyroid cancer and predicts pathological, genomic, and immunological outcomes. This AI model is based on the study of routine ultrasound images and incorporates multiple machine learning methods.
The new AI model could present a low-cost, non-invasive option for screening, staging, and personalized immunology treatment planning for the disease.Researchers obtained around 1,346 thyroid nodule images through routine diagnostic ultrasound from 784 patients to train and validate the AI platform. The ultrasound images were divided into datasets, one for internal training and validation and one for external validation. Malignancy was confirmed with different samples obtained by fine needle biopsy. Pathological staging and mutational status were established with operative reports and genomic sequencing. A multimodal platform utilizing these techniques accurately predicted 98.7% of thyroid nodule malignancies in the internal dataset, significantly outperforming individual AI modalities.(4)
Bacteria Identified in Seconds With "Fingerprint" Machine Learning Technique
To date, bacterial identification can take hours, creating a diagnostic delay with sometimes extremely harmful infections. Researchers at KAIST (Korea Advanced Institute of Science and Technology) have developed a faster and more accurate identification method than current practices. They trained a deep learning algorithm to identify the “fingerprint” spectra of the molecular elements of various bacteria.This tool allowed the researchers to classify various bacteria in different media with up to 98% accuracy. Raman spectroscopy sends light through a sample to analyze how it scatters. The results reveal structural information about the piece -the spectral fingerprint- allowing researchers to identify its molecules.
However, many factors, such as spectral signals of the surrounding media and overlapping bacterial structures, make understanding these fingerprints very difficult. The researchers solved this problem by creating a separate deep learning model to analyze the fingerprints. They demonstrated that the platform is a simple, fast, and helpful route to classify the signals of two common bacteria and their preferred media without any break procedures. The researchers intend to use their platform to study more bacteria and media types to build a training data library. This registry aims for decreased collection and detection times for new samples.(5)
To date, bacterial identification can take hours, creating a diagnostic delay with sometimes extremely harmful infections. Researchers at KAIST (Korea Advanced Institute of Science and Technology) have developed a faster and more accurate identification method than current practices. They trained a deep learning algorithm to identify the “fingerprint” spectra of the molecular elements of various bacteria.
This tool allowed the researchers to classify various bacteria in different media with up to 98% accuracy. Raman spectroscopy sends light through a sample to analyze how it scatters. The results reveal structural information about the piece -the spectral fingerprint- allowing researchers to identify its molecules. However, many factors, such as spectral signals of the surrounding media and overlapping bacterial structures, make understanding these fingerprints very difficult. The researchers solved this problem by creating a separate deep learning model to analyze the fingerprints. They demonstrated that the platform is a simple, fast, and helpful route to classify the signals of two common bacteria and their preferred media without any break procedures. The researchers intend to use their platform to study more bacteria and media types to build a training data library. This registry aims for decreased collection and detection times for new samples.(5)
Driving Progress in Personalized Cancer Therapy
In collaboration with colleagues at the University of Arkansas, engineers at Johns Hopkins University, Maryland, created the first non-invasive optical probe to understand the complex changes in tumors after immunotherapy. The researchers probed colon cancer tumors in mice treated with two types of immune checkpoint inhibitors. The researchers used Raman spectroscopy to identify the molecular composition of the tissue and enhanced their analysis using AI. Raman spectroscopy produces an excellent molecular characterization and is well-suited for discovering the compositional changes in the tumor microenvironment. The algorithm used about 7,500 spectral data points from 25 tumors to identify the range of features induced by immunotherapy.(6)
In collaboration with colleagues at the University of Arkansas, engineers at Johns Hopkins University, Maryland, created the first non-invasive optical probe to understand the complex changes in tumors after immunotherapy. The researchers probed colon cancer tumors in mice treated with two types of immune checkpoint inhibitors. The researchers used Raman spectroscopy to identify the molecular composition of the tissue and enhanced their analysis using AI.
Raman spectroscopy produces an excellent molecular characterization and is well-suited for discovering the compositional changes in the tumor microenvironment. The algorithm used about 7,500 spectral data points from 25 tumors to identify the range of features induced by immunotherapy.(6)
UK becomes first country to design a customized cancer vaccine
Mr.Graham Booth was diagnosed with head and neck cancer in 2011, and; despite treatment, he relapsed four more times. However, he is now participating in a new clinical trial to prevent his cancer from returning. The Clatterbridge Cancer Centre, based in Liverpool, UK, is testing a new vaccine called TG4050 to treat head and neck cancers. The Biotechnology company Transgene, based in France, manufactures the treatment.
The corporation uses an AI model to discover new neo genes to design and invent novel targeted immunotherapies to treat cancer. Nevertheless, the injection that Clatterbridge is trialing is distinct because it is custom-made to the person’s DNA. It is impossible to target all neoantigens at once, and the trial uses artificial intelligence algorithms to decide which neoantigens to prioritize.
These algorithms are oriented on vast amounts of historical data to produce an individualized list of the most immunogenic series for each patient. Regarding the proprietary nature of the AI model used in this study and the data used to train it, it is necessary to assess whether this vaccine is effective at the end of the trial.(7)
Mr.Graham Booth was diagnosed with head and neck cancer in 2011, and; despite treatment, he relapsed four more times. However, he is now participating in a new clinical trial to prevent his cancer from returning. The Clatterbridge Cancer Centre, based in Liverpool, UK, is testing a new vaccine called TG4050 to treat head and neck cancers. The Biotechnology company Transgene, based in France, manufactures the treatment.
The corporation uses an AI model to discover new neo genes to design and invent novel targeted immunotherapies to treat cancer. Nevertheless, the injection that Clatterbridge is trialing is distinct because it is custom-made to the person’s DNA. It is impossible to target all neoantigens at once, and the trial uses artificial intelligence algorithms to decide which neoantigens to prioritize.These algorithms are oriented on vast amounts of historical data to produce an individualized list of the most immunogenic series for each patient. Regarding the proprietary nature of the AI model used in this study and the data used to train it, it is necessary to assess whether this vaccine is effective at the end of the trial.(7)