Hasil untuk "Neoplasms. Tumors. Oncology. Including cancer and carcinogens"

Fatemeh Ansarizadeh, Tonghua Zhang

This study presents a mathematical model that captures the interactions among tumor cells, healthy cells, and immune cells in a tumor-bearing host, with a specific focus on breast cancer. Incorporating the concept of delay, the model consists of four differential equations to analyze these cellular dynamics. The findings demonstrate the superior efficacy of metronomic chemotherapy compared to the maximum tolerated dose (MTD) method and underscore the necessity of adjunct therapies. Oscillatory tumor cell dynamics revealed by the model highlight the challenges of achieving complete tumor elimination through chemotherapy alone. Sensitivity analysis confirms the robustness of the model, particularly under metronomic treatment protocols, aligning with experimental observations regarding metronomic-to-MTD dosage ratios. Furthermore, the results emphasize the importance of synergistic effects from combination therapies. This biologically consistent framework provides valuable insights into tumor-immune interactions and offers a foundation for optimizing therapeutic strategies in cancer treatment.

T. Schaffer, A. Brawanski, S. Wein et al.

A U-Net based deep learning architecture is designed to segment brain tumors as they appear on various MRI modalities. Special emphasis is lent to the non-enhancing tumor compartment. The latter has not been considered anymore in recent brain tumor segmentation challenges like the MICCAI challenges. However, it is considered to be indicative of the survival time of the patient as well as of areas of further tumor growth. Hence it deems essential to have means to automatically delineate its extension within the tumor.

Yu Chen^*+, Haoyi Liu^*+, Shiyu Liu^*+ et al.

Introduction Adolescent tobacco use has become a serious global public health problem, and effective tobacco control public service advertisements (PSAs) are crucial for reducing adolescent smoking rates. The study aims to employ a mixedmethods approach combining quantitative surveys and qualitative focus groups to evaluate the effectiveness of different types of tobacco control PSAs among Chinese adolescents, identify effective advertising characteristics and content elements, and provide empirical evidence for optimizing youth tobacco control communication strategies. Methods A total of 125 students aged 10–18 years were recruited from six primary and secondary schools in Beijing and Kunming from November 2020 to April 2021. Participants completed Likert-scale ratings measuring advertisement effectiveness after viewing eight tobacco control PSAs and participated in focus group interviews. Quantitative data were analyzed using independent samples t-tests, Spearman correlation analysis, and multivariable logistic regression analysis, while qualitative data were analyzed using thematic analysis. All statistical tests were two-tailed with significance set at p<0.05. Results Quantitative analysis revealed that PSAs employing ‘testimonials’ and ‘disease’ frameworks were most strongly associated with prevention intentions, while those using ‘celebrity endorsement’, ‘humor’ and ‘appearance damage’ frameworks showed the weakest associations. Kunming adolescents showed significantly higher advertisement acceptance scores than Beijing adolescents (mean difference=0.21; 95% CI: 0.04–0.38, p<0.05). The 10-item effectiveness scale demonstrated good internal consistency (Cronbach’s α=0.82). Qualitative analysis identified effective characteristics including presentation of specific health hazards, use of testimonials, and fear appeals; ineffective characteristics included non-specific harm presentation, use of humorous elements, and appearance damage content. Conclusions Tobacco control PSA design should consider strategies combining disease warnings with real-life testimonials, avoid humorous advertisements and industry-sponsored messaging, and consider regional cultural differences. Distribution through online and social media platforms frequently used by adolescents may enhance reach. Future longitudinal research with broader geographical sampling is needed to confirm these findings.

Frank Bastian, Hassan Alkhayuon, Kieran Mulchrone et al.

We propose a simple dynamic model of cancer development that captures carcinogenesis and subsequent cancer progression. A central idea of the model is to include the immune system as an extinction threshold, similar to the strong Allee effect in population biology. We first identify the limitations of commonly used Allee effect models in reproducing typical cancer progression. We then address these limitations by deriving a new model that incorporates: (i) random mutations of stem cells at a rate that increases with age and (ii) immune response whose strength may also vary over time. Our model accurately reproduces a wide range of real-world cancer data: the typical age-specific cumulative risk of most human cancers, the progression of breast cancer in mice, and the unusual age-specific cumulative risk of breast cancer in women. In the last case, we use a moving extinction threshold to reflect the different immune response at different phases of the menstrual cycle and menopausal treatment. This provides new insights into the effects of hormone replacement therapy and menstrual cycle length. This moving threshold approach can be applied to a variety of other cancer scenarios where the immune response or other important factors may vary over time.

Pedro R. A. S. Bassi, Xinze Zhou, Wenxuan Li et al.

Early tumor detection save lives. Each year, more than 300 million computed tomography (CT) scans are performed worldwide, offering a vast opportunity for effective cancer screening. However, detecting small or early-stage tumors on these CT scans remains challenging, even for experts. Artificial intelligence (AI) models can assist by highlighting suspicious regions, but training such models typically requires extensive tumor masks--detailed, voxel-wise outlines of tumors manually drawn by radiologists. Drawing these masks is costly, requiring years of effort and millions of dollars. In contrast, nearly every CT scan in clinical practice is already accompanied by medical reports describing the tumor's size, number, appearance, and sometimes, pathology results--information that is rich, abundant, and often underutilized for AI training. We introduce R-Super, which trains AI to segment tumors that match their descriptions in medical reports. This approach scales AI training with large collections of readily available medical reports, substantially reducing the need for manually drawn tumor masks. When trained on 101,654 reports, AI models achieved performance comparable to those trained on 723 masks. Combining reports and masks further improved sensitivity by +13% and specificity by +8%, surpassing radiologists in detecting five of the seven tumor types. Notably, R-Super enabled segmentation of tumors in the spleen, gallbladder, prostate, bladder, uterus, and esophagus, for which no public masks or AI models previously existed. This study challenges the long-held belief that large-scale, labor-intensive tumor mask creation is indispensable, establishing a scalable and accessible path toward early detection across diverse tumor types. We plan to release our trained models, code, and dataset at https://github.com/MrGiovanni/R-Super

Ugur Dinc, Jibak Sarkar, Philipp Schubert et al.

Introduction: Large language models (LLM) have shown great potential in clinical decision support. GPT-5 is a novel LLM system that has been specifically marketed towards oncology use. Methods: Performance was assessed using two complementary benchmarks: (i) the ACR Radiation Oncology In-Training Examination (TXIT, 2021), comprising 300 multiple-choice items, and (ii) a curated set of 60 authentic radiation oncologic vignettes representing diverse disease sites and treatment indications. For the vignette evaluation, GPT-5 was instructed to generate concise therapeutic plans. Four board-certified radiation oncologists rated correctness, comprehensiveness, and hallucinations. Inter-rater reliability was quantified using Fleiss' \k{appa}. Results: On the TXIT benchmark, GPT-5 achieved a mean accuracy of 92.8%, outperforming GPT-4 (78.8%) and GPT-3.5 (62.1%). Domain-specific gains were most pronounced in Dose and Diagnosis. In the vignette evaluation, GPT-5's treatment recommendations were rated highly for correctness (mean 3.24/4, 95% CI: 3.11-3.38) and comprehensiveness (3.59/4, 95% CI: 3.49-3.69). Hallucinations were rare with no case reaching majority consensus for their presence. Inter-rater agreement was low (Fleiss' \k{appa} 0.083 for correctness), reflecting inherent variability in clinical judgment. Errors clustered in complex scenarios requiring precise trial knowledge or detailed clinical adaptation. Discussion: GPT-5 clearly outperformed prior model variants on the radiation oncology multiple-choice benchmark. Although GPT-5 exhibited favorable performance in generating real-world radiation oncology treatment recommendations, correctness ratings indicate room for further improvement. While hallucinations were infrequent, the presence of substantive errors underscores that GPT-5-generated recommendations require rigorous expert oversight before clinical implementation.

Roham Maiti, Debasmita Bhoumik

Brain plays a crucial role in regulating body functions and cognitive processes, with brain tumors posing significant risks to human health. Precise and prompt detection is a key factor in proper treatment and better patient outcomes. Traditional methods for detecting brain tumors, that include biopsies, MRI, and CT scans often face challenges due to their high costs and the need for specialized medical expertise. Recent developments in machine learning (ML) and deep learning (DL) has exhibited strong capabilities in automating the identification and categorization of brain tumors from medical images, especially MRI scans. However, these classical ML models have limitations, such as high computational demands, the need for large datasets, and long training times, which hinder their accessibility and efficiency. Our research uses MobileNET model for efficient detection of these tumors. The novelty of this project lies in building an accurate tumor detection model which use less computing re-sources and runs in less time followed by efficient decision making through the use of image processing technique for accurate results. The suggested method attained an average accuracy of 98.5%.

Jong Hyun Kim, Jongseong Jang

The application of machine learning to transcriptomics data has led to significant advances in cancer research. However, the high dimensionality and complexity of RNA sequencing (RNA-seq) data pose significant challenges in pan-cancer studies. This study hypothesizes that gene sets derived from single-cell RNA sequencing (scRNA-seq) data will outperform those selected using bulk RNA-seq in pan-cancer downstream tasks. We analyzed scRNA-seq data from 181 tumor biopsies across 13 cancer types. High-dimensional weighted gene co-expression network analysis (hdWGCNA) was performed to identify relevant gene sets, which were further refined using XGBoost for feature selection. These gene sets were applied to downstream tasks using TCGA pan-cancer RNA-seq data and compared to six reference gene sets and oncogenes from OncoKB evaluated with deep learning models, including multilayer perceptrons (MLPs) and graph neural networks (GNNs). The XGBoost-refined hdWGCNA gene set demonstrated higher performance in most tasks, including tumor mutation burden assessment, microsatellite instability classification, mutation prediction, cancer subtyping, and grading. In particular, genes such as DPM1, BAD, and FKBP4 emerged as important pan-cancer biomarkers, with DPM1 consistently significant across tasks. This study presents a robust approach for feature selection in cancer genomics by integrating scRNA-seq data and advanced analysis techniques, offering a promising avenue for improving predictive accuracy in cancer research.

Anahita Fathi Kazerooni, Nastaran Khalili, Xinyang Liu et al.

Pediatric tumors of the central nervous system are the most common cause of cancer-related death in children. The five-year survival rate for high-grade gliomas in children is less than 20%. Due to their rarity, the diagnosis of these entities is often delayed, their treatment is mainly based on historic treatment concepts, and clinical trials require multi-institutional collaborations. Here we present the CBTN-CONNECT-DIPGR-ASNR-MICCAI BraTS-PEDs challenge, focused on pediatric brain tumors with data acquired across multiple international consortia dedicated to pediatric neuro-oncology and clinical trials. The CBTN-CONNECT-DIPGR-ASNR-MICCAI BraTS-PEDs challenge brings together clinicians and AI/imaging scientists to lead to faster development of automated segmentation techniques that could benefit clinical trials, and ultimately the care of children with brain tumors.

Rethinam Senthil

Daryl Lim Joon, Colleen Berry, Benjamin Harris et al.

IntroductionImage guidance with gold fiducials improves outcomes of prostate radiotherapy. However, gold produces artefact on CT imaging, interfering with contouring and verification. The purpose of this study was to compare polymer to standard gold fiducials using radiotherapy imaging modalities to assess the visibility and artefact.MethodsTwenty eight patients with locally advanced prostate cancer were enrolled, half had three polymer fiducials implanted into the prostate and half underwent insertion of gold fiducials. Patients were imaged with CT, T2 weighted MRI, cone-beam CT (CBCT) and planar KV images. Fiducials were scored for visibility and assessed for CT artefact in surrounding prostate tissue. The artefact was quantified from Hounsfield number histograms and separated into percentile ranges and proportion of voxels in HU normal tissue range of a 2cm sphere surrounding the fiducial.ResultsGold and polymer fiducials were sufficiently visible for CT and CBCT verification. The gold fiducials could be visualized well on KV planar imaging; however, the polymer markers were obscured by pelvic bones. Neither polymer nor gold fiducials could be visualized on MRI. The polymer fiducial produced less artefact than gold on CT, having less voxel spread for the HU percentile ranges and a greater proportion of voxels in the normal tissue range.ConclusionsPolymer fiducials are a more suitable fiducial than gold for CT/CBCT in prostate cancer radiotherapy, demonstrating minimal artefact and good visibility on CT. However, they were not well seen on MRI or KV imaging and thus not suitable for co-registration or planar KV verification.

Vítor Teixeira, Kate Davies, Peter Campbell et al.

Jacqueline P. Whitehouse, Jacqueline P. Whitehouse, Hilary Hii et al.

IntroductionEpendymomas (EPN) are the third most common malignant brain cancer in children. Treatment strategies for pediatric EPN have remained unchanged over recent decades, with 10-year survival rates stagnating at just 67% for children aged 0-14 years. Moreover, a proportion of patients who survive treatment often suffer long-term neurological side effects as a result of therapy. It is evident that there is a need for safer, more effective treatments for pediatric EPN patients. There are ten distinct subgroups of EPN, each with their own molecular and prognostic features. To identify and facilitate the testing of new treatments for EPN, in vivo laboratory models representative of the diverse molecular subtypes are required. Here, we describe the establishment of a patient-derived orthotopic xenograft (PDOX) model of posterior fossa A (PFA) EPN, derived from a metastatic cranial lesion.MethodsPatient and PDOX tumors were analyzed using immunohistochemistry, DNA methylation profiling, whole genome sequencing (WGS) and RNA sequencing.ResultsBoth patient and PDOX tumors classified as PFA EPN by methylation profiling, and shared similar histological features consistent with this molecular subgroup. RNA sequencing revealed that gene expression patterns were maintained across the primary and metastatic tumors, as well as the PDOX. Copy number profiling revealed gains of chromosomes 7, 8 and 19, and loss of chromosomes 2q and 6q in the PDOX and matched patient tumor. No clinically significant single nucleotide variants were identified, consistent with the low mutation rates observed in PFA EPN. Overexpression of EZHIP RNA and protein, a common feature of PFA EPN, was also observed. Despite the aggressive nature of the tumor in the patient, this PDOX was unable to be maintained past two passages in vivo.DiscussionOthers who have successfully developed PDOX models report some of the lowest success rates for EPN compared to other pediatric brain cancer types attempted, with loss of tumorigenicity not uncommon, highlighting the challenges of propagating these tumors in the laboratory. Here, we discuss our collective experiences with PFA EPN PDOX model generation and propose potential approaches to improve future success in establishing preclinical EPN models.

Chengyang Song, Fangjian Shang, Wei Tu et al.

Abstract Background Disco-interaction protein 2 homologue B (DIP2B) plays an important role in DNA methylation. There have been many reports on DIP2B in various diseases, but neither the diagnostic value nor the prognostic value of DIP2B across cancer types has been deeply explored. Methods The expression levels of DIP2B in 33 cancer types were analysed based on data sets from The Cancer Genome Atlas (TCGA) and the Genotype-Tissue Expression (GTEx) database. The relationships of DIP2B expression with immune cell infiltration and immune-related gene expression were studied via the CIBERSORT, ESTIMATE and TISIDB tools. Gene set variation analysis (GSVA) was performed to identify pathways related to DIP2B. DIP2B knockdown by siRNA was performed in breast cancer cell lines to investigate the effect on proliferation, apoptosis and migration. The relationships of DIP2B expression with clinicopathological features and prognosis were analysed based on immunohistochemistry. Results DIP2B was highly expressed in 26 of 33 cancer types and was significantly associated with poor overall survival (OS) in breast invasive carcinoma (BRCA), mesothelioma and chromophobe renal cell carcinoma (each P < 0.05). DIP2B showed a negative correlation with the immune score, the infiltration levels of key immune killer cells (CD8 + T cells, activated NK cells and plasma cells), and the expression of major histocompatibility complex–related genes and chemokine-related genes in BRCA. Subtype analysis showed that DIP2B expression was associated with poor OS in Her-2 + BRCA patients (P < 0.05). DIP2B showed a negative correlation with immune killer cell infiltration and immune regulatory genes in BRCA subtypes. In BRCA, the GSVA results revealed that genes correlating positively with DIP2B were enriched in cancer-related pathways (PI3K-AKT) and cell-cycle-related pathways (MITOTIC_SPINDLE, G2M_CHECKPOINT and E2F_TARGETS), while genes correlating negatively with DIP2B were enriched in DNA_REPAIR. Knockdown of the DIP2B gene induced a reduction in proliferation and migration and an increase in apoptosis in breast cancer cell lines. DIP2B expression was associated with lymph node metastasis and poor histological grade in BRCA according to immunohistochemistry (each P < 0.05). DIP2B expression predicted reduced disease-free survival and OS in BRCA patients (each P < 0.05), especially those with the Her-2 + subtype (P = 0.023 and P = 0.069). Conclusions DIP2B may be a prognostic biomarker for BRCA, especially for the Her-2 + subtype. DIP2B is associated with a “cold” tumour immune microenvironment in BRCA and might serve as a future target for immunotherapy.

Radwa Adel, Ercan Engin Kuruoglu

Complex gene interactions play a significant role in cancer progression, driving cellular behaviors that contribute to tumor growth, invasion, and metastasis. Gene co-expression networks model the functional connectivity between genes under various biological conditions. Understanding the system-level evolution of these networks in cancer is critical for elucidating disease mechanisms and informing the development of targeted therapies. While previous studies have primarily focused on structural differences between cancer and normal cell co-expression networks, this study applies graph frequency analysis to cancer transcriptomic signals defined on gene co-expression networks, highlighting the graph spectral characteristics of cancer systems. Using a range of graph frequency filters, we showed that cancer cells display distinctive patterns in the graph frequency content of their gene transcriptomic signals, effectively distinguishing between cancer types and stages. The transformation of the original gene feature space into the graph spectral space captured more intricate cancer properties, as validated by significantly higher F-statistic scores for graph frequency-filtered gene features compared to those in the original space.

Mariateresa Giglio MD, Angela Preziosa MD, Roberta Mele MD et al.

Background In cancer patients with limited life expectancy, an implant of an intrathecal (IT) drug delivery system connected to a subcutaneous port (IDDS-SP) has been proposed as a successful strategy, but conflicting results are reported on quality of life (QoL). The aim of this prospective observational study is to report the effects on pain, mood and QoL of an IT combination therapy delivered by an IDDS-SP in malignant refractory pain. Methods Adult patients in which IT therapy was recommended were recruited. An IT therapy with morphine and levobupivacaine was started: VASPI score, depression and anxiety (evaluated by the Edmonton Symptom Assessment System -ESAS-), the Pittsburgh Sleep Quality Index (PSQI), the 5-level EuroQol 5D version (EQ-5D-5L) and the requirements of breakthrough cancer pain (BTcP) medications were registered, with adverse events rate and the satisfaction of patients scored as Patient Global Impression of Change (PGIC). Results Fifty patients, (16 F/34 M) were enrolled (age 69 ± 12). All had advanced cancer with metastasis. The median daily VASPI score was 75, the median depression score was 6, and the median anxiety score was 4, median PSQI was 16. At 28 days, a significant reduction in VASPI score was registered as well as in depression and anxiety item. Also, PSQI decreased significantly. The EQ-5D-5 L showed a significant improvement in all components at 14 and 28 days. Patient Global Impression of Change scores showed high level of satisfaction. A low incidence of adverse events and a reduction in BTCP episodes were also registered. Conclusion Intrathecal combination therapy delivered by an IDDS-SP could ensure adequate control of cancer related symptoms, such as pain, depression, anxiety and sleep disturbances. These effects, with low rate of AEs and reduced BTcP episodes, could explain the improvement in QoL and the overall high levels of patients’ satisfaction.

YANG Shaofei, YANG Guanghua, PU Jiaxi, LU Minhao, ZHAO Zhiqing, ZHAO Bin

Objective To evaluate arterial covered stent used in the resection of complex carotid body tumor (CBT) and postoperative adverse events. Methods Twelve patients with CBT Shamblin type &#x02161; 3 cases and type &#x02162; 9 cases from 2018 to 2020 who underwent intraoperative combined arterial covered stent reconstruction (4 cases) or prereconstruction of internal carotid artery (8 cases) were included in this study. Statistics and analysis were done for general clinical data, intraoperative evaluation and postoperative adverse events. Results The operation was successful for all 12 cases. Average operative time was (160&#x000b1;55) min and blocking time was (25&#x000b1;6) min during carotid artery reconstruction. There was without need to block during internal carotid artery prereconstruction. Intraoperative bleeding was (342&#x000b1;101) mL. It was shown by follow-up one year later that there were 2 cases of temporary nerve injury and 1 case of postoperative covered stent obstruction. Both permanent nerve injury and transient ischemic attack were not found. Conclusions Stent graft used to reconstruct during operation or prereconstruction of internal carotid artery in CBT Shamblin type &#x02161; and type &#x02162; could reduce vascular injury, shorten blocking time of carotid artery, and lower adverse events. However, the treatment remained to be studied with large-sample controlled trial.

Buhai Wang, Yaqin Fan, Liying Zhang et al.

Introduction: Pembrolizumab is widely used in advanced non-small-cell lung cancer (NSCLC) patients with positive programmed death-ligand 1 (PD-L1). However, efficacy evaluation along treatment by serial monitoring of circulating tumor DNA (ctDNA) using next-generation sequencing remained to be well studied. Methods: Nine PD-L1 positive advanced NSCLC patients were prospectively enrolled and received pembrolizumab monotherapy. Pretreatment tissue and/or plasma samples were collected as baseline reference. Serial plasma samples were collected after 3 and 6 weeks of treatment as well as at disease progression. All samples underwent targeted next-generation sequencing. Results: The median progression-free survival (mPFS) and median overall survival (mOS) were 4.43 and 25.53 months, respectively. In total, 3 patients achieved partial response (PR) or stable disease (SD) for more than 6 months and were thus classified into the durable clinical benefit (DCB) group, whereas the rest 6 were grouped as nondurable benefit (NDB) patients. Molecular profiling of baseline samples revealed that TP53 and APC were the 2 most frequently mutated genes in all patients, whereas POT1 and SETD2 mutations were enriched in DCB and NDB groups, respectively. Higher tumor mutational burden (TMB) was observed in DCB patients than NDB group. During serial ctDNA monitoring, 2 DCB patients showed a dramatic ctDNA reduction while 75% of NDB patients’ ctDNA concentration increased at week 6. Several acquired mutations might contribute to the pembrolizumab resistance, including CDKN2A frameshift and MITF nonsense mutations. Conclusions: Genomic profiling of peripheral blood samples can be applied to dynamically monitor disease progression. The reduction in ctDNA concentration during treatment implied DCBs.

Jinfeng Ning, Fengjiao Wang, Jianlong Bu et al.

Abstract Aberrant epigenetic drivers or suppressors contribute to LUAD progression and drug resistance, including KRAS, PTEN, Keap1. Human Plant Homeodomain (PHD) finger protein 1 (PHF1) coordinates with H3K36me3 to increase nucleosomal DNA accessibility. Previous studies revealed that PHF1 is markedly upregulated in various tumors and enhances cell proliferation, migration and tumorigenesis. However, its roles in LUAD are still unknown. We aimed to depict the biological roles of PHF1 and identify useful targets for clinical treatment of LUAD. Based on the bioinformatic analysis, we found that PHF1 was down-regulated in LUAD samples and low PHF1 expressions correlated with unfavorable clinical characteristics. Patients with low PHF1 had poorer survival outcomes relative to those with high PHF1. Targeting PHF1 potentiated cell growth, migration and in vivo proliferation. Mechanistically, FTO mediated the stabilization of PHF1 mRNA by demethylating m6A, which particularly prevented YTHDF2 from degrading PHF1 transcripts. Of note, FTO also expressed lowly in LUAD that predicts poor prognosis of patients. FTO inhibition promoted LUAD progression, and PHF1 overexpression could reverse the effect. Lastly, down-regulated FTO/PHF1 axis could mainly elevate FOXM1 expression to potentiate the self-renewal capacity. Targeting FOXM1 was effective to suppress PHF1low/− LUAD growth. Collectively, our findings revealed that FTO positively regulates PHF1 expression and determined the tumor-suppressive role of FTO/PHF1 axis, thereby highlighting insights into its epigenetic remodeling mechanisms in LUAD progression and treatment.

Jianan Chen, Anne L. Martel

Cancer stage is a large determinant of patient prognosis and management in many cancer types, and is often assessed using medical imaging modalities, such as CT and MRI. These medical images contain rich information that can be explored to stratify patients within each stage group to further improve prognostic algorithms. Although the majority of cancer deaths result from metastatic and multifocal disease, building imaging biomarkers for patients with multiple tumors has been a challenging task due to the lack of annotated datasets and standard study framework. In this paper, we process two public datasets to set up a benchmark cohort of 341 patient in total for studying outcome prediction of multifocal metastatic cancer. We identify the lack of expressiveness in common multiple instance classification networks and propose two injective multiple instance pooling functions that are better suited to outcome prediction. Our results show that multiple instance learning with injective pooling functions can achieve state-of-the-art performance in the non-small-cell lung cancer CT and head and neck CT outcome prediction benchmarking tasks. We will release the processed multifocal datasets, our code and the intermediate files i.e. extracted radiomic features to support further transparent and reproducible research.