AIRIaL: Artificial Intelligence and Resistance Imaging in Lung Cancer

Lung cancer is the most common cause of cancer death world-wide (~1.6 million deaths/year). Unfortunately, of all people who have lung cancer, only 5 out of 100 are living ten years after diagnosis. The main reason for this low survival rate is that standard lung cancer therapies fail due to the presence of cancer that does not respond to these treatments.

In this early-stage project performed in mice, we will develop new medical imaging tests to detect therapy resistance in lung cancer. Using this imaging test, we can predict if the cancer will respond to the prescribed treatment, and if so, how well. In the future, knowing early on about how a patient will respond to treatment will enable the clinical team to switch to alternative options if necessary. This will avoid patients being given potentially toxic treatments that will not benefit them. At present doctors have to wait for CT scans about 3 months after treatment starts to be able to tell if it is working or not.

We will create a computer programme (using artificial intelligence) which is able to automatically read lung scans and more accurately predict whether the cancer is resistant to treatment. Then, we will use engineered materials to deliver drugs only to the resistant cancer cells. At this stage we will test these new drugs and computer program in mice to see if this new approach is effective. By delivering the drug just to the cancer itself, we hope to reduce side effects for patients and significantly improve how well the treatment works. Finally, we will determine if our computer programme can monitor changes in cancer over time, which will provide insight into how treatment resistance occurs as this is currently not well-understood. Ultimately, we hope that the results of this project will enable doctors to tailor the treatment for lung cancer to the individual patient. Moreover, it might lead to the development of new treatments for this devastating disease.

Most lung cancer deaths result from ineffective treatment of stage 3 and 4 disease. Currently, there is no satisfactory way to identify patients that will not respond to standard-of-care treatments. Positron emission tomography (PET) imaging offers a potential solution to this clinical problem through the non-invasive assessment of molecular processes that underpin therapy-resistance. The identification of cancer patients that are refractory to treatment will allow the use of innovative second-line therapies that have the potential to improve patient response and survival.

Here, we will combine expertise in biological, physical, data, and medical sciences to identify and treat therapy-resistant non-small cell lung cancer (NSCLC). We have developed a PET radiotracer, [18F]FSPG, that can non-invasively detect therapy-resistant tumours in multiple models of NSCLC. [18F]FSPG is taken up by the amino acid transporter xCT, which provides the rate-limiting precursor for glutathione biosynthesis and is upregulated >10-fold in therapy-resistant NSCLC. Through innovations in artificial intelligence and machine learning, minimal preclinical [18F]FSPG imaging datasets will be used to identify, segment, and extract quantitative parameters from therapy-resistant tumours; far outperforming conventional methods of annotation. Concurrently, we will develop molecularly imprinted nanoparticles (nanoMIPs) that bind xCT to deliver a targeted payload of drug to these resistant tumours. An iterative selection process will be employed to manufacture double imprinted fluorescent nanoMIPs that bind both xCT and doxorubicin to produce therapeutic agents with excellent selectivity and potency. Finally, these innovative technologies will be combined to image, detect, and treat multifocal therapy-resistant disease in a genetically engineered mouse model of NSCLC. Together, this novel treatment platform will support our mission to provide treatment options where currently there are none.