Prediction of Effectiveness and Toxicities of Immune Checkpoint Inhibitors Using Real-World Patient Data.

PURPOSE: Although immune checkpoint inhibitors (ICIs) have improved outcomes in certain patients with cancer, they can also cause life-threatening immunotoxicities. Predicting immunotoxicity risks alongside response could provide a personalized risk-benefit profile, inform therapeutic decision making, and improve clinical trial cohort selection. We aimed to build a machine learning (ML) framework using routine electronic health record (EHR) data to predict hepatitis, colitis, pneumonitis, and 1-year overall survival. METHODS: Real-world EHR data of more than 2,200 patients treated with ICI through December 31, 2018, were used to develop predictive models. Using a prediction time point of ICI initiation, a 1-year prediction time window was applied to create binary labels for the four outcomes for each patient. Feature engineering involved aggregating laboratory measurements over appropriate time windows (60-365 days). Patients were randomly partitioned into training (80%) and test (20%) sets. Random forest classifiers were developed using a rigorous model development framework. RESULTS: The patient cohort had a median age of 63 years and was 61.8% male. Patients predominantly had melanoma (37.8%), lung cancer (27.3%), or genitourinary cancer (16.4%). They were treated with PD-1 (60.4%), PD-L1 (9.0%), and CTLA-4 (19.7%) ICIs. Our models demonstrate reasonably strong performance, with AUCs of 0.739, 0.729, 0.755, and 0.752 for the pneumonitis, hepatitis, colitis, and 1-year overall survival models, respectively. Each model relies on an outcome-specific feature set, though some features are shared among models. CONCLUSION: To our knowledge, this is the first ML solution that assesses individual ICI risk-benefit profiles based predominantly on routine structured EHR data. As such, use of our ML solution will not require additional data collection or documentation in the clinic.

Accelerated curation of checkpoint inhibitor-induced colitis cases from electronic health records.

Objective: Automatically identifying patients at risk of immune checkpoint inhibitor (ICI)-induced colitis allows physicians to improve patientcare. However, predictive models require training data curated from electronic health records (EHR). Our objective is to automatically identify notes documenting ICI-colitis cases to accelerate data curation. Materials and Methods: We present a data pipeline to automatically identify ICI-colitis from EHR notes, accelerating chart review. The pipeline relies on BERT, a state-of-the-art natural language processing (NLP) model. The first stage of the pipeline segments long notes using keywords identified through a logistic classifier and applies BERT to identify ICI-colitis notes. The next stage uses a second BERT model tuned to identify false positive notes and remove notes that were likely positive for mentioning colitis as a side-effect. The final stage further accelerates curation by highlighting the colitis-relevant portions of notes. Specifically, we use BERT's attention scores to find high-density regions describing colitis. Results: The overall pipeline identified colitis notes with 84% precision and reduced the curator note review load by 75%. The segment BERT classifier had a high recall of 0.98, which is crucial to identify the low incidence (<10%) of colitis. Discussion: Curation from EHR notes is a burdensome task, especially when the curation topic is complicated. Methods described in this work are not only useful for ICI colitis but can also be adapted for other domains. Conclusion: Our extraction pipeline reduces manual note review load and makes EHR data more accessible for research.

Validation of prognostic biomarker scores for predicting progression of dementia in patients with amnestic mild cognitive impairment.

OBJECTIVE: The objective of this study was to develop and validate a practical computerized prognostic model that uses baseline psychometric and imaging data, including results of PET imaging of amyloid deposition, to predict the progression to dementia in patients at risk for Alzheimer's disease (AD). PATIENTS AND METHODS: Data from patients in a phase II trial of [F]flutemetamol for PET imaging of brain amyloid and from the Alzheimer's Disease Neuroimaging Initiative were used to train the prognostic model to yield a disease state index (DSI), a measure of the similarity of an individual patient's data to data from patients in specific diagnostic groups. Inputs to the model included amyloid PET results, MRI measurements of hippocampal volume, and the results of psychometric tests. The model was subsequently validated by using data from a prospective study of an independent cohort of patients with mild cognitive impairment. RESULTS: In total, data from 223 patients of the 233 enroled were suitable for analysis. The DSI predicted by the model and the risk of progression to AD dementia within 3 years were higher for patients with amyloid deposition and neurodegeneration than for patients with amyloid deposition without neurodegeneration. Rates of non-AD dementia among patients with neurodegeneration at baseline were consistent with the results of other studies. The results were consistent with the Jack model of AD progression. CONCLUSION: The DSI from the model that included psychometric, MRI, and PET amyloid data provides useful prognostic information in cases of mild cognitive impairment.

Validation of an electronic image reader training programme for interpretation of [18F]flutemetamol ß-amyloid PET brain images.

OBJECTIVES: An electronic training programme (ETP) was developed for interpretation of images during routine clinical use of the PET amyloid imaging agent [F]flutemetamol injection (VIZAMYL). This study was carried out to validate the ETP. MATERIALS AND METHODS: Five nuclear medicine technologists (NMTs) and five readers previously inexperienced in amyloid image interpretation were required to self-train using the ETP and pass a test to participate. A total of 305 [F]flutemetamol PET images were then tested as the validation set, following preassessment and reorientation (where required) by one of five NMTs. Next, a new set of readers blinded to clinical information independently assessed all 305 images. Images had been acquired in previous studies from patients representing the full spectrum of cognitive capacity. When available, a standard of truth determined by histopathology or clinical history was used to derive sensitivity and specificity for image interpretation from this validation set. Randomly selected images (n=29) were read in duplicate to measure intrareader reproducibility. Images were read first without, and subsequently with anatomic images, if available. RESULTS: All NMTs and all readers scored 100% on the qualifying test. The interpretation of 135 cases without anatomic image support resulted in sensitivity ranging from 84% to 94% (majority 94%, median 92%) and specificity ranging from 77% to 96% (majority 92%, median 81%). Inter-reader agreement was very high, with most κ scores more than 0.8. Intrareader reproducibility ranged from 93 to 100%. CONCLUSION: The self-guided ETP effectively trained new amyloid PET image readers to accurately and reproducibly interpret [F]flutemetamol PET images.

129 Xe chemical shift in human blood and pulmonary blood oxygenation measurement in humans using hyperpolarized 129 Xe NMR.

PURPOSE: To evaluate the dependency of the 129 Xe-red blood cell (RBC) chemical shift on blood oxygenation, and to use this relation for noninvasive measurement of pulmonary blood oxygenation in vivo with hyperpolarized 129 Xe NMR. METHODS: Hyperpolarized 129 Xe was equilibrated with blood samples of varying oxygenation in vitro, and NMR was performed at 1.5 T and 3 T. Dynamic in vivo NMR during breath hold apnea was performed at 3 T on two healthy volunteers following inhalation of hyperpolarized 129 Xe. RESULTS: The 129 Xe chemical shift in RBCs was found to increase nonlinearly with blood oxygenation at 1.5 T and 3 T. During breath hold apnea, the 129 Xe chemical shift in RBCs exhibited a periodic time modulation and showed a net decrease in chemical shift of ∼1 ppm over a 35 s breath hold, corresponding to a decrease of 7-10 % in RBC oxygenation. The 129 Xe-RBC signal amplitude showed a modulation with the same frequency as the 129 Xe-RBC chemical shift. CONCLUSION: The feasibility of using the 129 Xe-RBC chemical shift to measure pulmonary blood oxygenation in vivo has been demonstrated. Correlation between 129 Xe-RBC signal and 129 Xe-RBC chemical shift modulations in the lung warrants further investigation, with the aim to better quantify temporal blood oxygenation changes in the cardiopulmonary vascular circuit. Magn Reson Med 77:1399-1408, 2017. © 2016 The Authors Magnetic Resonance in Medicine published by Wiley Periodicals, Inc. on behalf of International Society for Magnetic Resonance in Medicine. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Relaxation and exchange dynamics of hyperpolarized 129Xe in human blood.

PURPOSE: (129) Xe-blood NMR was performed over the full blood oxygenation range to evaluate (129) Xe relaxation and exchange dynamics in human blood. METHODS: Hyperpolarized (129) Xe was equilibrated with blood and isolated plasma, and NMR was performed at 1.5 T. RESULTS: The (129) Xe relaxation rate was found to increase nonlinearly with decreasing blood oxygenation. Three constants were extrapolated: rsO2 = 11.1, a "relaxivity index" characterizing the rate of change of (129) Xe relaxation as a function of blood oxygenation, and 1/T1oHb = 0.13 s(-1) and 1/T1dHb = 0.42 s(-1) , the (129) Xe relaxation rates in oxygenated blood and deoxygenated blood, respectively. In addition, rate constants, ka = 0.022 ms(-1) and kb = 0.062 ms(-1) , were determined for xenon diffusing between red blood cells (RBCs) and plasma (hematocrit = 48%). The (129) Xe-O2 relaxivity in plasma, rO2 = 0.075 s(-1) mM(-1) , and the (129) Xe relaxation rate in isolated plasma (without dissolved O2 ), 1/T1,b0 = 0.046 s(-1) , were also calculated. Finally, intrinsic (129) Xe-RBC relaxation rates, 1/T1,aoHb = 0.19 s(-1) and 1/T1,adHb = 0.84 s(-1) , in oxygenated blood and deoxygenated blood, respectively, were calculated. CONCLUSION: The relaxation and exchange analysis performed in this study should provide a sound experimental basis upon which to design future MR experiments for dissolved xenon transport from the lungs to distal tissues.

Experimental validation of the hyperpolarized 129 Xe chemical shift saturation recovery technique in healthy volunteers and subjects with interstitial lung disease.

PURPOSE: To assess the sensitivity of the hyperpolarized 129 Xe chemical shift saturation recovery (CSSR) technique for noninvasive quantification of changes to lung microstructure and function in idiopathic pulmonary fibrosis (IPF) and systemic sclerosis (SSc). METHODS: Ten healthy volunteers, four subjects with SSc and four with IPF were scanned at 1.5 T. A CSSR pulse sequence was implemented using binomial-composite radiofrequency pulses to monitor 129 Xe magnetization in tissues and blood plasma (T/P) and red blood cells (RBCs). The dynamics of 129 Xe uptake into these compartments were fitted with three existing analytical models of gas diffusion to extract parameters of lung physiology. These parameters were quantitatively compared between models. RESULTS: Uptake of xenon into the pulmonary capillaries was impaired in subjects with IPF and SSc. Statistically significant septal thickening was measured by 129 Xe CSSR in IPF patients. Preliminary data suggests age-dependent alterations to septal thickness in healthy volunteers. These findings were reproduced using each of the literature models. CSSR-derived parameters were compared with gold-standard indicators of pulmonary function; diffusing capacity of carbon monoxide and pulmonary transit-time. CONCLUSIONS: CSSR with hyperpolarized 129 Xe is sensitive to pathology-induced degradation of lung structure/function and shows promise for quantification of disease severity and monitoring treatment response. Magn Reson Med 74:196-207, 2015. © 2014 Wiley Periodicals, Inc.

Quantitative analysis of hyperpolarized 129Xe ventilation imaging in healthy volunteers and subjects with chronic obstructive pulmonary disease.

In this study, hyperpolarized (129) Xe MR ventilation and (1) H anatomical images were obtained from three subject groups: young healthy volunteers (HVs), subjects with chronic obstructive pulmonary disease (COPD) and age-matched controls (AMCs). Ventilation images were quantified by two methods: an expert reader-based ventilation defect score percentage (VDS%) and a semi-automated segmentation-based ventilation defect percentage (VDP). Reader-based values were assigned by two experienced radiologists and resolved by consensus. In the semi-automated analysis, (1) H anatomical images and (129) Xe ventilation images were both segmented following registration to obtain the thoracic cavity volume and ventilated volume, respectively, which were then expressed as a ratio to obtain the VDP. Ventilation images were also characterized by generating signal intensity histograms from voxels within the thoracic cavity volume, and heterogeneity was analyzed using the coefficient of variation (CV). The reader-based VDS% correlated strongly with the semi-automatically generated VDP (r = 0.97, p < 0.0001) and with CV (r = 0.82, p < 0.0001). Both (129) Xe ventilation defect scoring metrics readily separated the three groups from one another and correlated significantly with the forced expiratory volume in 1 s (FEV1 ) (VDS%: r = -0.78, p = 0.0002; VDP: r = -0.79, p = 0.0003; CV: r = -0.66, p = 0.0059) and other pulmonary function tests. In the healthy subject groups (HVs and AMCs), the prevalence of ventilation defects also increased with age (VDS%: r = 0.61, p = 0.0002; VDP: r = 0.63, p = 0.0002). Moreover, ventilation histograms and their associated CVs distinguished between subjects with COPD with similar ventilation defect scores, but visibly different ventilation patterns.

Direct visualisation of collateral ventilation in COPD with hyperpolarised gas MRI.

BACKGROUND: Collateral ventilation has been proposed as a mechanism of compensation of respiratory function in obstructive lung diseases but observations of it in vivo are limited. The assessment of collateral ventilation with an imaging technique might help to gain insight into lung physiology and assist the planning of new bronchoscopic techniques for treating emphysema. OBJECTIVE: To obtain images of delayed ventilation that might be related to collateral ventilation over the period of a single breath-hold in patients with chronic obstructive pulmonary disease (COPD). METHODS: Time-resolved breath-hold hyperpolarised (3)He MRI was used to obtain images of the progressive influx of polarised gas into initially non-ventilated defects. RESULTS: A time-series of images showed that (3)He moves into lung regions which were initially non-ventilated. Ventilation defects with delayed filling were observed in 8 of the 10 patients scanned. CONCLUSIONS: A method for direct imaging of delayed ventilation within a single breath-hold has been demonstrated in patients with COPD. Images of what is believed to be collateral ventilation and slow filling of peripheral airspaces due to increased flow resistance are presented. The technique provides 3D whole-lung coverage with sensitivity to regional information, and is non-invasive and non-ionising.

Chronic obstructive pulmonary disease: safety and tolerability of hyperpolarized 129Xe MR imaging in healthy volunteers and patients.

PURPOSE: To evaluate the safety and tolerability of inhaling multiple 1-L volumes of undiluted hyperpolarized xenon 129 ((129)Xe) followed by up to a 16-second breath hold and magnetic resonance (MR) imaging. MATERIALS AND METHODS: This study was approved by the institutional review board and was HIPAA compliant. Written informed consent was obtained. Forty-four subjects (19 men, 25 women; mean age, 46.1 years ± 18.8 [standard deviation]) were enrolled, consisting of 24 healthy volunteers, 10 patients with chronic obstructive pulmonary disease (COPD), and 10 age-matched control subjects. All subjects received three or four 1-L volumes of undiluted hyperpolarized (129)Xe, followed by breath-hold MR imaging. Oxygen saturation, heart rate and rhythm, and blood pressure were continuously monitored. These parameters, along with respiratory rate and subjective symptoms, were assessed after each dose. Subjects' serum biochemistry and hematology were recorded at screening and at 24-hour follow-up. A 12-lead electrocardiogram (ECG) was obtained at these times and also within 2 hours prior to and 1 hour after (129)Xe MR imaging. Xenon-related symptoms were evaluated for relationship to subject group by using a χ(2) test and to subject age by using logistic regression. Changes in vital signs were tested for significance across subject group and time by using a repeated-measures multivariate analysis of variance test. RESULTS: The 44 subjects tolerated all xenon inhalations, no subjects withdrew, and no serious adverse events occurred. No significant changes in vital signs (P > .27) were observed, and no subjects exhibited changes in laboratory test or ECG results at follow-up that were deemed clinically important or required intervention. Most subjects (91%) did experience transient xenon-related symptoms, most commonly dizziness (59%), paresthesia (34%), euphoria (30%), and hypoesthesia (30%). All symptoms resolved without clinical intervention in 1.6 minutes ± 0.9. CONCLUSION: Inhalation of hyperpolarized (129)Xe is well tolerated in healthy subjects and in those with mild or moderate COPD. Subjects do experience mild, transient, xenon-related symptoms, consistent with its known anesthetic properties.

Hyperpolarized molecules in solution.

Hyperpolarization is a technique to enhance the nuclear polarization and thereby increase the available signal in magnetic resonance (MR). This chapter provides an introduction to the concept of hyperpolarization as well as an overview of dynamic nuclear polarization (DNP) and para-hydrogen induced polarization (PHIP), two methods used to generate hyperpolarized molecules in aqueous solution.

Applications of hyperpolarized agents in solutions.

This chapter provides an overview of pulse sequences adapted to hyperpolarized MR imaging. Applications of hyperpolarized agents in aqueous solution are reviewed. Vascular (e.g., angiography, perfusion, and catheter tracking) as well as metabolic (e.g., oncology, cardiology, neurology, and pH mapping) applications are covered. Due to the rapid development of new applications for hyperpolarized agents, a review format has been used for this chapter instead of a strict protocol/procedure structure.

Diffusion-weighted hyperpolarized 129Xe MRI in healthy volunteers and subjects with chronic obstructive pulmonary disease.

Given its greater availability and lower cost, (129) Xe apparent diffusion coefficient (ADC) MRI offers an alternative to (3) He ADC MRI. To demonstrate the feasibility of hyperpolarized (129) Xe ADC MRI, we present results from healthy volunteers (HV), chronic obstructive pulmonary disease (COPD) subjects, and age-matched healthy controls (AMC). The mean parenchymal ADC was 0.036 ± 0.003 cm(2) sec(-1) for HV, 0.043 ± 0.006 cm(2) sec(-1) for AMC, and 0.056 ± 0.008 cm(2) sec(-1) for COPD subjects with emphysema. In healthy individuals, but not the COPD group, ADC decreased significantly in the anterior-posterior direction by ∼ 22% (P = 0.006, AMC; 0.0059, HV), likely because of gravity-induced tissue compression. The COPD group exhibited a significantly larger superior-inferior ADC reduction (∼ 28%) than the healthy groups (∼ 24%) (P = 0.00018, HV; P = 3.45 × 10(-5) , AMC), consistent with smoking-related tissue destruction in the superior lung. Superior-inferior gradients in healthy subjects may result from regional differences in xenon concentration. ADC was significantly correlated with pulmonary function tests (forced expiratory volume in 1 sec, r = -0.77, P = 0.0002; forced expiratory volume in 1 sec/forced vital capacity, r = -0.77, P = 0.0002; diffusing capacity of carbon monoxide in the lung/alveolar volume (V(A) ), r = -0.77, P = 0.0002). In healthy groups, ADC increased with age by 0.0002 cm(2) sec(-1) year(-1) (r = 0.56, P = 0.02). This study shows that (129) Xe ADC MRI is clinically feasible, sufficiently sensitive to distinguish HV from subjects with emphysema, and detects age- and posture-dependent changes.

Hyperpolarized Xe MR imaging of alveolar gas uptake in humans.

BACKGROUND: One of the central physiological functions of the lungs is to transfer inhaled gases from the alveoli to pulmonary capillary blood. However, current measures of alveolar gas uptake provide only global information and thus lack the sensitivity and specificity needed to account for regional variations in gas exchange. METHODS AND PRINCIPAL FINDINGS: Here we exploit the solubility, high magnetic resonance (MR) signal intensity, and large chemical shift of hyperpolarized (HP) (129)Xe to probe the regional uptake of alveolar gases by directly imaging HP (129)Xe dissolved in the gas exchange tissues and pulmonary capillary blood of human subjects. The resulting single breath-hold, three-dimensional MR images are optimized using millisecond repetition times and high flip angle radio-frequency pulses, because the dissolved HP (129)Xe magnetization is rapidly replenished by diffusive exchange with alveolar (129)Xe. The dissolved HP (129)Xe MR images display significant, directional heterogeneity, with increased signal intensity observed from the gravity-dependent portions of the lungs. CONCLUSIONS: The features observed in dissolved-phase (129)Xe MR images are consistent with gravity-dependent lung deformation, which produces increased ventilation, reduced alveolar size (i.e., higher surface-to-volume ratios), higher tissue densities, and increased perfusion in the dependent portions of the lungs. Thus, these results suggest that dissolved HP (129)Xe imaging reports on pulmonary function at a fundamental level.

Relationship between structural changes and hyperpolarized gas magnetic resonance imaging in chronic obstructive pulmonary disease using computational simulations with realistic alveolar geometry.

Both the development of accurate models of lung function and their quantitative validation can be significantly enhanced by the use of functional imaging techniques. The advent of hyperpolarized noble gas magnetic resonance imaging (MRI) technology has increased the amount of local, functional information we can obtain from the lung. In particular, application of (3)He to measure apparent diffusion coefficients has enabled some measure of lung microstructure and airspace size within the lung. Models mimicking image acquisition in hyperpolarized gas MRI can improve understanding of the relationship between image findings and lung structure, and can be used to improve the definition of imaging protocols. In this paper, we review the state of the art in hyperpolarized gas MRI modelling. We also present our own results, obtained using a Monte Carlo approach and a realistic alveolar sac geometry, which has previously been applied in functional lung studies. In this way, we demonstrate the potential for models combining lung function and image acquisition, which could provide valuable tools in both basic studies and clinical practice.

Detecting tumor response to treatment using hyperpolarized 13C magnetic resonance imaging and spectroscopy.

Measurements of early tumor responses to therapy have been shown, in some cases, to predict treatment outcome. We show in lymphoma-bearing mice injected intravenously with hyperpolarized [1-(13)C]pyruvate that the lactate dehydrogenase-catalyzed flux of (13)C label between the carboxyl groups of pyruvate and lactate in the tumor can be measured using (13)C magnetic resonance spectroscopy and spectroscopic imaging, and that this flux is inhibited within 24 h of chemotherapy. The reduction in the measured flux after drug treatment and the induction of tumor cell death can be explained by loss of the coenzyme NAD(H) and decreases in concentrations of lactate and enzyme in the tumors. The technique could provide a new way to assess tumor responses to treatment in the clinic.

Characterization of diffusing capacity and perfusion of the rat lung in a lipopolysaccaride disease model using hyperpolarized 129Xe.

The ability to quantify pulmonary diffusing capacity and perfusion using dynamic hyperpolarized (129)Xe NMR spectroscopy is demonstrated. A model of alveolar gas exchange was developed, which, in conjunction with (129)Xe NMR, enables quantification of average alveolar wall thickness, pulmonary perfusion, capillary diffusion length, and mean transit time. The technique was employed to compare a group of naïve rats (n = 10) with a group of rats with acute inflammatory lung injury (n = 10), caused by instillation of lipopolysaccaride (LPS). The measured structural and perfusion-related parameters were in agreement with reported values from studies using non-NMR methods. Significant differences between the groups were found in total diffusion length (control 8.5 +/- 0.5 microm, LPS 9.9 +/- 0.6 microm, P < 0.001), in capillary diffusion length (control 2.9 +/- 0.4 microm, LPS 3.9 +/- 1.0 microm, P < 0.05), and in pulmonary hematocrit (control 0.55 +/- 0.06, LPS 0.43 +/- 0.08, P < 0.01), whereas no differences were observed in alveolar wall thickness, pulmonary perfusion, and mean transit time. These results demonstrate the ability of the method to distinguish two main aspects of lung function, namely, diffusing capacity and pulmonary perfusion.

Increase of signal-to-noise of more than 10,000 times in liquid state NMR.

Extract: Two major applications exist for Nuclear Magnetic Resonance (NMR): spectroscopy and imaging. NMR spectroscopy has gained acceptance as one of the major analytical techniques, due to the detailed information that can be obtained about molecular structure, dynamics and intra- and inter-molecular interactions. Magnetic resonance imaging (MRI) is a non-invasive technique with superior soft tissue contrast and broad diagnostic value. The technique has gained wide clinical acceptance and is of great importance in diagnostic medicine. However, despite significant technological advancements (increasing field strength and cooling of electronics), the application of NMR is limited by an intrinsically low sensitivity, as compared to other analytical methods. Fundamentally, the low sensitivity originates from the low magnetic energy of nuclear spins, compared to the thermal energy at room temperature. At a magnetic field strength of 1.5 Tesla and room temperature, the proton spins are polarized to only 5 parts per million, and an improvement of 200,000 is thus theoretically possible. For other nuclei bearing lower magnetic moments (1/4 for 13C and 1/10 for 15N, respectively, compared to 1H), the theoretical enhancement factor is proportionally greater. The weak nuclear polarization is generally compensated by a high concentration (i.e., a large number of nuclear spins). However, the sensitivity of several other nuclei is further reduced by the low natural abundance of the NMR-active isotope (1.1 % for 13C and 0.36 % for 15N, respectively).

Diffusion MRI for prediction of response of rectal cancer to chemoradiation.

Prediction of tumour response before onset of treatment could have considerable clinical benefit. Since the apparent diffusion coefficient (ADC) of a tumour's water content can show the extent of necrosis, we looked for a possible correlation of ADC with response to treatment. We measured mean tumour water ADC before and after chemotherapy and chemoradiation in 14 patients with locally advanced rectal cancer, with a quantitative magnetic resonance diffusion imaging sequence. We found a strong negative correlation between mean pretreatment tumour water ADC and percentage size change of tumours after chemotherapy (r=-0.67, p=0.01) and chemoradiation (r=-0.83, p=0.001). Persistence of low ADC in responders after chemotherapy could represent loss of a non-viable fraction of the treated tumour.