Study workflow. A Collecting datasets to generate deep features, evaluate feature stability, histological subtype classification and prognosis. b The DPCCA model is used to find shared latent space between CT scans and metabolomics. An enlarged version of this model is shown, with the purple box highlighting the section responsible for generating TMR-CT features. In this model, XA, Xb is the original image associated with metabolomics; from these we create YesA,Yesb image and metabolomics integrations, respectively. The PCCA model then combines them into zA,za B,zb latent variables. Latent variables zA,zb capture view-specific variations while za B captures the covariance. From the latent variables, we use a generative process of model sampling from low-dimensional PCCA to reconstruct image and metabolomics embeddings y^a,y^b{\hat{y}}^{a },{\hat{y }}^{b}. Each integration y^a,y^b{\hat{y}}^{a},{\hat{y}}^{b} is then decoded to produce the X^a,X^b{\hat{X }}^{a},{\hat{X}}^{b} using view-specific decoders. During implementation, when we only have the CT image, we extract the learned representation YesAwhich we defined as TMR-CT features. vs Purple box convolutional neural networks were used to generate TMR-CT features on external datasets and to test feature stability using the RIDER dataset. d TMR-CT features are used for histological subtype classification, along with random forest (RF)-based approaches, as they exhibit superior performance in both tasks. Credit: npj Precision Oncology (2024). DOI: 10.1038/s41698-024-00502-3
Empire researchers used artificial intelligence (AI) to extract information about the chemical composition of lung tumors from medical scans. For the first time, they demonstrated how the combination of medical imaging and AI can be used to provide a “virtual biopsy” for cancer patients.
Their non-invasive method can classify the type of lung cancer a patient has – which is crucial for selecting the right treatment – and predict whether the cancer is likely to progress. According to the researchers, this technique could be used by doctors when it is not possible or appropriate to obtain a biopsy of a patient’s physical tissues.
The study is published in the journal npj Precision Oncology and was led by Imperial College London, alongside collaborators from Cordoba, Spain.
Lead author of the study, Professor Eric Aboagye, from Imperial’s Department of Surgery and Cancer, said: “At present, trying to find detailed information about tissues and tumors requires Invasive biopsies, which may be uncomfortable for the patient, delay treatment. decisions and be costly for health services. Although CT scans are commonly used in the clinic, they fail to offer detailed information on cell type or prognostic information of diseases.
First author and Imperial Ph.D. Candidate Marc Boubnovski Martell adds: “We have developed a system that merges CT scans with the chemical composition of tumors and normal lung tissue. This allows us to classify lung cancer types and, importantly, provide reliable predictions about patient outcomes.
Early detection and diagnosis
Lung cancer is the most common cause of cancer deaths in the UK, with around 35,000 lives lost each year, according to Cancer Research UK. This is partly because symptoms do not appear in the early stages and there is an urgent need for new ways to detect and treat the tumor before it spreads to other parts of the body.
Patients who have symptoms of lung cancer tend to be diagnosed using chest X-rays and computed tomography (CT) scans, which can also show whether the cancer has spread beyond the lungs. If it is safe to obtain a biopsy sample, clinicians then examine the tumor cells under a microscope and classify the type of lung cancer a patient has. This helps doctors decide which treatment would be best.
A relatively new test called metabolomic profiling, which also requires a tissue biopsy, can provide much more detailed information about the chemistry and metabolism of tumor cells and, importantly, how the cancer is likely to progress. However, this operation requires a lot of work and time and is therefore not systematically performed in hospitals.
AI-powered imaging
In recent years, AI has been used to analyze medical exams and look for signs of illness that may not be detected by doctors or might not even be visible to the naked eye. Generative AI, a type of AI that can learn from data to create new content, is currently being researched for multiple applications.
The Imperial team took these ideas further and wondered whether the information about lung tumor chemistry contained in the metabolomic profile could show up in CT scans.
But first, any AI model needs to be trained on existing groups of patients that have a medical exam, a definitive diagnosis, and preferably lots of additional clinical information. The researchers used data from 48 lung cancer patients treated at the Reina Sofia University Hospital (UHRS) in Cordoba, Spain.
Uniquely, all patients underwent a CT scan, as well as a detailed metabolomic profile of their tumor tissue and healthy tissue adjacent to the tumor. Based on this data, the Imperial team developed an AI-powered deep learning assessment tool that they called tissue-metabolomics-radiomics-CT (TMR-CT).
Researchers discovered a significant and powerful correlation between patients’ metabolomic profiles and “deep features” in their CT scans, which appear as lighter or darker areas in the image.
Using this method, the researchers hypothesized that they could bypass the need for physical tissue samples and infer the metabolic characteristics of the tumor from the scan alone.
To test this, they used their TMR-CT model in a separate group of 723 lung cancer patients treated at the Royal Marsden Hospital, Guy and St Thomas’ Hospital or Imperial College NHS Healthcare Trust. All patients had a CT scan, but no metabolomic data was available.
The results showed that TMR-CT skillfully classified lung cancer and, importantly, gave reliable predictions of patient outcomes, outperforming traditional CT-based methods and clinical assessments.
A future clinical tool
The researchers hope to confirm their TMR-CT method in other groups of lung cancer patients as well as potentially in people with brain, ovarian and endometrial cancers, for whom it may also be difficult to obtain biopsies.
In the future, the technique could be incorporated as an algorithm into the software loaded on commercial medical imaging scanners.
Professor Aboagye concludes: “This research shows the potential of using CT scans to gain a deeper and more nuanced understanding of the chemical composition of tissues and tumors, which has until now only been accessible through direct sampling of fabrics. This method could prove particularly beneficial in countries such as the United Kingdom, where the prevalence of lung cancer is high, and could potentially transform diagnosis and treatment protocols. »
More information:
Marc Boubnovski Martell et al, Deep learning of tissue metabolome and CT representations annotates the classification and prognosis of NSCLC, npj Precision Oncology (2024). DOI: 10.1038/s41698-024-00502-3
Provided by Imperial College London
Quote: “Virtual biopsy” uses AI to help doctors assess lung cancer (February 21, 2024) retrieved February 21, 2024 from
This document is subject to copyright. Apart from fair use for private study or research purposes, no part may be reproduced without written permission. The content is provided for information only.
