Imaging-Based Subtypes of Pancreatic Ductal Adenocarcinoma Exhibit Differential Growth and Metabolic Patterns in the Pre-Diagnostic Period: Implications for Early Detection

A retrieval-augmented chatbot based on GPT-4 provides appropriate differential diagnosis in gastrointestinal radiology: a proof of concept study

BACKGROUND

We investigated the potential of an imaging-aware GPT-4-based chatbot in providing diagnoses based on imaging descriptions of abdominal pathologies.

METHODS

Utilizing zero-shot learning via the LlamaIndex framework, GPT-4 was enhanced using the 96 documents from the Radiographics Top 10 Reading List on gastrointestinal imaging, creating a gastrointestinal imaging-aware chatbot (GIA-CB). To assess its diagnostic capability, 50 cases on a variety of abdominal pathologies were created, comprising radiological findings in fluoroscopy, MRI, and CT. We compared the GIA-CB to the generic GPT-4 chatbot (g-CB) in providing the primary and 2 additional differential diagnoses, using interpretations from senior-level radiologists as ground truth. The trustworthiness of the GIA-CB was evaluated by investigating the source documents as provided by the knowledge-retrieval mechanism. Mann-Whitney U test was employed.

RESULTS

The GIA-CB demonstrated a high capability to identify the most appropriate differential diagnosis in 39/50 cases (78%), significantly surpassing the g-CB in 27/50 cases (54%) (p = 0.006). Notably, the GIA-CB offered the primary differential in the top 3 differential diagnoses in 45/50 cases (90%) versus g-CB with 37/50 cases (74%) (p = 0.022) and always with appropriate explanations. The median response time was 29.8 s for GIA-CB and 15.7 s for g-CB, and the mean cost per case was $0.15 and $0.02, respectively.

CONCLUSIONS

The GIA-CB not only provided an accurate diagnosis for gastrointestinal pathologies, but also direct access to source documents, providing insight into the decision-making process, a step towards trustworthy and explainable AI. Integrating context-specific data into AI models can support evidence-based clinical decision-making.

RELEVANCE STATEMENT

A context-aware GPT-4 chatbot demonstrates high accuracy in providing differential diagnoses based on imaging descriptions, surpassing the generic GPT-4. It provided formulated rationale and source excerpts supporting the diagnoses, thus enhancing trustworthy decision-support.

KEY POINTS

• Knowledge retrieval enhances differential diagnoses in a gastrointestinal imaging-aware chatbot (GIA-CB). • GIA-CB outperformed the generic counterpart, providing formulated rationale and source excerpts. • GIA-CB has the potential to pave the way for AI-assisted decision support systems.

Potential of artificial intelligence in the risk stratification for and early detection of pancreatic cancer

Pancreatic ductal adenocarcinoma (PDAC) is the third most lethal cancer in the United States, with a 5-year life expectancy of 11%. Most symptoms manifest at an advanced stage of the disease when surgery is no longer appropriate. The dire prognosis of PDAC warrants new strategies to improve the outcomes of patients, and early detection has garnered significant attention. However, early detection of PDAC is most often incidental, emphasizing the importance of developing new early detection screening strategies. Due to the low incidence of the disease in the general population, much of the focus for screening has turned to individuals at high risk of PDAC. This enriches the screening population and balances the risks associated with pancreas interventions. The cancers that are found in these high-risk individuals by MRI and/or EUS screening show favorable 73% 5-year overall survival. Even with the emphasis on screening in enriched high-risk populations, only a minority of incident cancers are detected this way. One strategy to improve early detection outcomes is to integrate artificial intelligence (AI) into biomarker discovery and risk models. This expert review summarizes recent publications that have developed AI algorithms for the applications of risk stratification of PDAC using radiomics and electronic health records. Furthermore, this review illustrates the current uses of radiomics and biomarkers in AI for early detection of PDAC. Finally, various challenges and potential solutions are highlighted regarding the use of AI in medicine for early detection purposes.

Neoplastic cell enrichment of tumor tissues using coring and laser microdissection for proteomic and genomic analyses of pancreatic ductal adenocarcinoma

BACKGROUND

The identification of differentially expressed tumor-associated proteins and genomic alterations driving neoplasia is critical in the development of clinical assays to detect cancers and forms the foundation for understanding cancer biology. One of the challenges in the analysis of pancreatic ductal adenocarcinoma (PDAC) is the low neoplastic cellularity and heterogeneous composition of bulk tumors. To enrich neoplastic cells from bulk tumor tissue, coring, and laser microdissection (LMD) sampling techniques have been employed. In this study, we assessed the protein and KRAS mutation changes associated with samples obtained by these enrichment techniques and evaluated the fraction of neoplastic cells in PDAC for proteomic and genomic analyses.

METHODS

Three fresh frozen PDAC tumors and their tumor-matched normal adjacent tissues (NATs) were obtained from three sampling techniques using bulk, coring, and LMD; and analyzed by TMT-based quantitative proteomics. The protein profiles and characterizations of differentially expressed proteins in three sampling groups were determined. These three PDACs and samples of five additional PDACs obtained by the same three sampling techniques were also subjected to genomic analysis to characterize KRAS mutations.

RESULTS

The neoplastic cellularity of eight PDACs ranged from less than 10% to over 80% based on morphological review. Distinctive proteomic patterns and abundances of certain tumor-associated proteins were revealed when comparing the tumors and NATs by different sampling techniques. Coring and bulk tissues had comparable proteome profiles, while LMD samples had the most distinct proteome composition compared to bulk tissues. Further genomic analysis of bulk, cored, or LMD samples demonstrated that KRAS mutations were significantly enriched in LMD samples while coring was less effective in enriching for KRAS mutations when bulk tissues contained a relatively low neoplastic cellularity.

CONCLUSIONS

In addition to bulk tissues, samples from LMD and coring techniques can be used for proteogenomic studies. The greatest enrichment of neoplastic cellularity is obtained with the LMD technique.

Mass Transport Model of Radiation Response: Calibration and Application to Chemoradiation for Pancreatic Cancer

PURPOSE

The benefit of radiation therapy for pancreatic ductal adenocarcinoma (PDAC) remains unclear. We hypothesized that a new mechanistic mathematical model of chemotherapy and radiation response could predict clinical outcomes a priori, using a previously described baseline measurement of perfusion from computed tomography scans, normalized area under the enhancement curve (nAUC).

METHODS AND MATERIALS

We simplified an existing mass transport model that predicted cancer cell death by replacing previously unknown variables with averaged direct measurements from randomly selected pathologic sections of untreated PDAC. This allowed using nAUC as the sole model input to approximate tumor perfusion. We then compared the predicted cancer cell death to the actual cell death measured from corresponding resected tumors treated with neoadjuvant chemoradiation in a calibration cohort (n = 80) and prospective cohort (n = 25). After calibration, we applied the model to 2 separate cohorts for pathologic and clinical associations: targeted therapy cohort (n = 101), cetuximab/bevacizumab + radiosensitizing chemotherapy, and standard chemoradiation cohort (n = 81), radiosensitizing chemotherapy to 50.4 Gy in 28 fractions.

RESULTS

We established the relationship between pretreatment computed v nAUC to pathologically verified blood volume fraction of the tumor (r = 0.65; P = .009) and fractional tumor cell death (r = 0.97-0.99; P < .0001) in the calibration and prospective cohorts. On multivariate analyses, accounting for traditional covariates, nAUC independently associated with overall survival in all cohorts (mean hazard ratios, 0.14-0.31). Receiver operator characteristic analyses revealed discrimination of good and bad prognostic groups in the cohorts with area under the curve values of 0.64 to 0.71.

CONCLUSIONS

This work presents a new mathematical modeling approach to predict clinical response from chemotherapy and radiation for PDAC. Our findings indicate that oxygen/drug diffusion strongly influences clinical responses and that nAUC is a potential tool to select patients with PDAC for radiation therapy.

Prediction of anti-CD25 and 5-FU treatments efficacy for pancreatic cancer using a mathematical model

BACKGROUND

Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal disease with rising incidence and with 5-years overall survival of less than 8%. PDAC creates an immune-suppressive tumor microenvironment to escape immune-mediated eradication. Regulatory T (Treg) cells and myeloid-derived suppressor cells (MDSC) are critical components of the immune-suppressive tumor microenvironment. Shifting from tumor escape or tolerance to elimination is the major challenge in the treatment of PDAC.

RESULTS

In a mathematical model, we combine distinct treatment modalities for PDAC, including 5-FU chemotherapy and anti- CD25 immunotherapy to improve clinical outcome and therapeutic efficacy. To address and optimize 5-FU and anti- CD25 treatment (to suppress MDSCs and Tregs, respectively) schedule in-silico and simultaneously unravel the processes driving therapeutic responses, we designed an in vivo calibrated mathematical model of tumor-immune system (TIS) interactions. We designed a user-friendly graphical user interface (GUI) unit which is configurable for treatment timings to implement an in-silico clinical trial to test different timings of both 5-FU and anti- CD25 therapies. By optimizing combination regimens, we improved treatment efficacy. In-silico assessment of 5-FU and anti- CD25 combination therapy for PDAC significantly showed better treatment outcomes when compared to 5-FU and anti- CD25 therapies separately. Due to imprecise, missing, or incomplete experimental data, the kinetic parameters of the TIS model are uncertain that this can be captured by the fuzzy theorem. We have predicted the uncertainty band of cell/cytokines dynamics based on the parametric uncertainty, and we have shown the effect of the treatments on the displacement of the uncertainty band of the cells/cytokines. We performed global sensitivity analysis methods to identify the most influential kinetic parameters and simulate the effect of the perturbation on kinetic parameters on the dynamics of cells/cytokines.

CONCLUSION

Our findings outline a rational approach to therapy optimization with meaningful consequences for how we effectively design treatment schedules (timing) to maximize their success, and how we treat PDAC with combined 5-FU and anti- CD25 therapies. Our data revealed that a synergistic combinatorial regimen targeting the Tregs and MDSCs in both crisp and fuzzy settings of model parameters can lead to tumor eradication.

A Mathematical Model to Estimate Chemotherapy Concentration at the Tumor-Site and Predict Therapy Response in Colorectal Cancer Patients with Liver Metastases

Simple Summary

It is known that drug transport barriers in the tumor determine drug concentration at the tumor site, causing disparity from the systemic (plasma) drug concentration. However, current clinical standard of care still bases dosage and treatment optimization on the systemic concentration of drugs. Here, we present a proof of concept observational cohort study to accurately estimate drug concentration at the tumor site from mathematical modeling using biologic, clinical, and imaging/perfusion data, and correlate it with outcome in colorectal cancer liver metastases. We demonstrate that drug concentration at the tumor site, not in systemic circulation, can be used as a credible biomarker for predicting chemotherapy outcome, and thus our mathematical modeling approach can be applied prospectively in the clinic to personalize treatment design to optimize outcome.

Abstract

Chemotherapy remains a primary treatment for metastatic cancer, with tumor response being the benchmark outcome marker. However, therapeutic response in cancer is unpredictable due to heterogeneity in drug delivery from systemic circulation to solid tumors. In this proof-of-concept study, we evaluated chemotherapy concentration at the tumor-site and its association with therapy response by applying a mathematical model. By using pre-treatment imaging, clinical and biologic variables, and chemotherapy regimen to inform the model, we estimated tumor-site chemotherapy concentration in patients with colorectal cancer liver metastases, who received treatment prior to surgical hepatic resection with curative-intent. The differential response to therapy in resected specimens, measured with the gold-standard Tumor Regression Grade (TRG; from 1, complete response to 5, no response) was examined, relative to the model predicted systemic and tumor-site chemotherapy concentrations. We found that the average calculated plasma concentration of the cytotoxic drug was essentially equivalent across patients exhibiting different TRGs, while the estimated tumor-site chemotherapeutic concentration (eTSCC) showed a quadratic decline from TRG = 1 to TRG = 5 (p < 0.001). The eTSCC was significantly lower than the observed plasma concentration and dropped by a factor of ~5 between patients with complete response (TRG = 1) and those with no response (TRG = 5), while the plasma concentration remained stable across TRG groups. TRG variations were driven and predicted by differences in tumor perfusion and eTSCC. If confirmed in carefully planned prospective studies, these findings will form the basis of a paradigm shift in the care of patients with potentially curable colorectal cancer and liver metastases.