Peptide receptor radionuclide therapy (PRRT) is an established treatment for metastatic neuroendocrine tumors (NETs), yet long-term disease control occurs only in a subset of patients. Predicting progression-free survival (PFS) could support individualized treatment planning. This study evaluates laboratory, imaging, and multimodal deep learning models for PFS prediction in PRRT-treated patients. In this retrospective, single-center study 116 patients with metastatic NETs undergoing 177Lu-DOTATOC were included. Clinical characteristics, laboratory values, and pretherapeutic somatostatin receptor positron emission tomography/computed tomographies (SR-PET/CT) were collected. Seven models were trained to classify low- vs. high-PFS groups, including unimodal (laboratory, SR-PET, or CT) and multimodal fusion approaches. Explainability was evaluated by feature importance analysis and gradient maps. Forty-two patients (36%) had short PFS (< 1 year), 74 patients long PFS (>1 year). Groups were similar in most characteristics, except for higher baseline chromogranin A (p = 0.003), elevated gamma-GT (p = 0.002), and fewer PRRT cycles (p < 0.001) in short-PFS patients. The Random Forest model trained only on laboratory biomarkers reached an AUROC of 0.59 +- 0.02. Unimodal three-dimensional convolutional neural networks using SR-PET or CT performed worse (AUROC 0.42 +- 0.03 and 0.54 +- 0.01, respectively). A multimodal fusion model laboratory values, SR-PET, and CT -augmented with a pretrained CT branch - achieved the best results (AUROC 0.72 +- 0.01, AUPRC 0.80 +- 0.01). Multimodal deep learning combining SR-PET, CT, and laboratory biomarkers outperformed unimodal approaches for PFS prediction after PRRT. Upon external validation, such models may support risk-adapted follow-up strategies.

The increasing complexity and cost of clinical trials, particularly in the context of oncology and advanced therapies, pose significant challenges for drug development. This study evaluates the predictive capabilities of large language models (LLMs) such as GPT-3.5, GPT-4, and HINT in determining clinical trial outcomes. By leveraging a curated dataset of trials from ClinicalTrials.gov, we compare the models' performance using metrics including balanced accuracy, specificity, recall, and Matthews Correlation Coefficient (MCC). Results indicate that GPT-4o demonstrates robust performance in early trial phases, achieving high recall but facing limitations in specificity. Conversely, the HINT model excels in recognizing negative outcomes, particularly in later trial phases, offering a balanced approach across diverse endpoints. Oncology trials, characterized by high complexity, remain challenging for all models. Additionally, trial duration and disease categories influence predictive performance, with longer durations and complex diseases such as neoplasms reducing accuracy. This study highlights the complementary strengths of LLMs and HINT, providing insights into optimizing predictive tools for clinical trial design and risk management. Future advancements in LLMs are essential to address current gaps in handling negative outcomes and complex domains.

This article was published as a part of the Data Science Blogathon. This problem that we will discuss in this blog comes from the cutting-edge intersection of AI with the drug discovery process, where DataRobot and my team play a very significant role. This blog is focused on an engagement my team, and I did with one of our largest customers, a top-tier pharmaceutical company in the United States. The goal with this type of work my team and I do is to tackle problems that are not standardized, which allows us to learn from them and then cross-functionally work with our Product and Engineering teams to integrate them into the DataRobot Platform, which pushes the boundaries of innovation in AI. In this blog, I consider an example and look into some work I have done in this field where I marry classical approaches in Survival Analysis with modern-day machine learning techniques to improve explainability, improve the accuracy of predicting adverse health events in patients and decrease time to release of the drug in the market.

In the first study of its kind in cancer, researchers have applied artificial intelligence to measure the amount of muscle in patients with brain tumours to help improve prognosis and treatment. Dr. Ella Mi, a clinical research fellow at Imperial College London (UK) will tell the NCRI Virtual Showcase, that using deep learning to evaluate MRI brain scans of a muscle in the head was as accurate and reliable as a trained person, and was considerably quicker. Furthermore, her research showed that the amount of muscle measured in this way could be used to predict how long a patient might survive their disease as it was an indicator of a patient's overall condition. Glioblastoma is an aggressive brain tumour that is very difficult to treat successfully. Average survival after diagnosis is 12-18 months and fewer than 5% of patients are still alive after five years.

In the first study of its kind in cancer, researchers have applied artificial intelligence to measure the amount of muscle in patients with brain tumours to help improve prognosis and treatment. Dr Ella Mi, a clinical research fellow at Imperial College London (UK) told the NCRI Virtual Showcase, that using deep learning to evaluate MRI brain scans of a muscle in the head was as accurate and reliable as a trained person, and was considerably quicker. Furthermore, her research showed that the amount of muscle measured in this way could be used to predict how long a patient might survive their disease as it was an indicator of a patient's overall condition. Glioblastoma is an aggressive brain tumour that is very difficult to treat successfully. Average survival after diagnosis is 12-18 months and fewer than 5% of patients are still alive after five years.