JKJoel Kullberg

Joel Kullberg

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Joel Kullberg has a total of 179 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Time-Driven Survival Analysis from FDG-PET/CT in Non-Small Cell Lung Cancer

    by Sambit Tarai, Ashish Chauhan, Elin Lundström, Johan Öfverstedt, Therese Sjöholm, Veronica Sanchez Rodriguez, Håkan Ahlström, Joel Kullberg

    Purpose: Automated medical image-based prediction of clinical outcomes, such as overall survival (OS), has great potential in improving patient prognostics and personalized treatment planning. We developed a deep regression framework using tissue-wise FDG-PET/CT projections as input, along with a temporal input representing a scalar time horizon (in days) to predict OS in patients with Non-Small Cell Lung Cancer (NSCLC). Methods: The proposed framework employed a ResNet-50 backbone to process input images and generate corresponding image embeddings. The embeddings were then combined with temporal data to produce OS probabilities as a function of time, effectively parameterizing the predictions based on time. The overall framework was developed using the U-CAN cohort (n = 556) and evaluated by comparing with a baseline method on the test set (n = 292). The baseline utilized the ResNet-50 architecture, processing only the images as input and providing OS predictions at pre-specified intervals, such as 2 or 5 year. Results: The incorporation of temporal data with image embeddings demonstrated an advantage in predicting OS, outperforming the baseline method with an improvement in AUC of 4.3%. The proposed model using clinical + IDP features achieved strong performance, and an ensemble of imaging and clinical + IDP models achieved the best overall performance (0.788), highlighting the complementary value of multimodal inputs. The proposed method also enabled risk stratification of patients into distinct categories (high vs. low risk). Heat maps from the saliency analysis highlighted tumor regions as key structures for the prediction. Conclusion: Our method provided an automated framework for predicting OS as a function of time, and demonstrates the potential of combining imaging and tabular data for improved survival prediction.

  • Open Access

    Tissue-specific and whole-body insulin sensitivity by integrated imaging and hyperinsulinemic euglycemic clamp: A repeatability study in people with T2DM and overweight/obesity

    by Iina Laitinen, Helena Litorp, Folke Sjöberg, Simon Ekström, Henrik Haraldsson, Stefan Pierrou, Alexander Korenyushkin, Joel Kullberg, Sudeepti Southekal, Ming Lu, et al.

    Assessment of glucose uptake by PET imaging under hyperinsulinemic euglycemic clamp (HEC) is an insightful tool for quantification of insulin resistance, a hallmark of diabetes and an area of interest in drug development. To enable the use of the method in metabolic trials, the repeatability of dynamic whole-body PET/MRI assessments of the tissue-specific glucose uptake and the total body glucose utilisation were investigated. The study included participants with type 2 diabetes mellitus (T2DM) and overweight/obesity, for two repeated examinations in standardised conditions. All participants signed informed consent, and the study plan was approved by the Swedish Ethical Review Authority (#2020-04140). After an overnight fast, HEC was established and a series of [18F]FDG-PET/MRI scans were performed. Total body glucose utilisation (M-value) was calculated for the duration of the scan and the tissue-specific metabolic rates of glucose uptake (MRGlu) were calculated using Patlak model. The repeatability was assessed by calculating the intraclass correlation coefficient (ICC). Repeatability was assessed in per protocol set of 12 participants (PPS, defined by a consistent HEC) and in full analysis set (FAS n = 16). The measured M-values and tissue MRGlu demonstrated varying levels of insulin resistance. M-value ICC was 0.95 (95% CI 0.86-0.99) for PPS, indicating excellent repeatability. Tissue-specific MRGlu repeatability was excellent for skeletal muscle (ICC 0.94), and good to at least fair for SAT, VAT, myocardium, and brain. The FAS had lower, but at least fair repeatability, emphasising the importance of standardisation in metabolic assessments. Dynamic [18F]FDG-PET/MRI provides quantitative information on tissue-specific insulin sensitivity during hyperinsulinemic euglycemic clamp with a reliability comparable to total body glucose utilisation assessment. The method has potential to add value in monitoring and evaluating T2DM treatment effects on glucose uptake and insulin resistance in interventional trials.


  • Automated Whole-Body PET/CT Lesion Tracking for Lesion-Wise Response Evaluation in Metastatic Breast Cancer

    by Hanna Jönsson, Alexander Korenyushkin, Jasper J. L. van Geel, Bertha Eisses, Elisabeth G. E. de Vries, Carolina P. Schröder, Håkan Ahlström, Joel Kullberg


  • Open Access

    Inclusion of tumor periphery in radiomics analysis of magnetic resonance images does not improve predictions of preoperative therapy response in patients with rectal cancer

    by Nafsika Korsavidou Hult, Sambit Tarai, Klara Hammarström, Joel Kullberg, Elin Lundström, Tomas Bjerner, Bengt Glimelius, Håkan Ahlström

    Background/Purpose: To evaluate the advantages of including versus excluding the tumor periphery and combining diffusion-weighted imaging (DWI) with T2-weighted imaging (T2w) for outcome predictions of preoperative radio(chemo)therapy in rectal cancer. Methods: Four analysis strategies, based on two segmentation methods and two magnetic resonance imaging (MRI) sequences, were evaluated in 106 patients examined with pretreatment MRI. One segmentation method included the tumor periphery in the region of interest (ROI) encompassing the whole tumor (wROI), considered as the reference segmentation approach, and one included only the central part (cROI). Relevant radiomics imaging features were extracted from either T2w alone or from both T2w and DWI and used by a machine learning algorithm for the prediction of pathologic complete response (pCR), neoadjuvant rectal (NAR) score, and disease recurrence. The area under the curve (AUC) was the performance measure. AUCs were compared with a bootstrapping method based on 10 4 bootstraps. Results: cROI applied to both T2w and DWI provided the highest numerical prediction of pCR (AUC 0.76), however, not significantly superior to the other strategies ( p ≥ 0.138). cROI applied to both T2w and DWI also yielded the highest numerical prediction of NAR score (AUC 0.84), showing advantages over wROI-based analysis strategies (AUC 0.66 and 0.69; p ≤ 0.008). When compared to cROI applied to T2w alone (AUC 0.73), the benefit was borderline statistically significant (p = 0.053 ). For prediction of disease recurrence, no differences were found between the analysis strategies. Conclusions: Inclusion of the tumor periphery in radiomic analysis of magnetic resonance images does not improve predictions of the preoperative therapy response in patients with rectal cancer. Excluding tumor periphery while adding DWI to T2w improves prediction of the NAR score, although it does not affect pCR or recurrence prediction. Graphical


  • Open Access

    Whole-body tumor segmentation from FDG-PET/CT: Leveraging a segmentation prior from tissue-wise projections

    by Sambit Tarai, Elin Lundström, Nouman Ahmad, Robin Strand, Håkan Ahlström, Joel Kullberg


  • Leveraging Point Annotations in Segmentation Learning with Boundary Loss

    by Eva Breznik, Hoel Kervadec, Filip Malmberg, Joel Kullberg, Håkan Ahlström, Marleen de Bruijne, Robin Strand


  • Prediction of Total Metabolic Tumor Volume from Tissue-Wise FDG-PET/CT Projections, Interpreted Using Cohort Saliency Analysis

    by Sambit Tarai, Elin Lundström, Johan Öfverstedt, Hanna Jönsson, Nouman Ahmad, Håkan Ahlström, Joel Kullberg


  • Open Access

    Short-term effects of obesity surgery versus low-energy diet on body composition and tissue-specific glucose uptake: a randomised clinical study using whole-body integrated 18F-FDG-PET/MRI

    by Jan W. Eriksson, Maria J. Pereira, Christakis Kagios, Sofia Kvernby, Elin Lundström, Giovanni Fanni, Martin H. Lundqvist, Björn C. L. Carlsson, Magnus Sundbom, Sambit Tarai, et al.

    Aims/hypothesis: Obesity surgery (OS) and diet-induced weight loss rapidly improve insulin resistance. We aim to investigate the impact of either Roux-en-Y gastric bypass (RYGB) or sleeve gastrectomy (SG) surgery compared with a diet low in energy (low-calorie diet; LCD) on body composition, glucose control and insulin sensitivity, assessed both at the global and tissue-specific level in individuals with obesity but not diabetes. Methods: In this parallel group randomised controlled trial, patients on a waiting list for OS were randomised (no blinding, sealed envelopes) to either undergo surgery directly or undergo an LCD before surgery. At baseline and 4 weeks after surgery (n=15, 11 RYGB and 4 SG) or 4 weeks after the start of LCD (n=9), investigations were carried out, including an OGTT and hyperinsulinaemic–euglycaemic clamps during which concomitant simultaneous whole-body [18F]fluorodeoxyglucose-positron emission tomography (PET)/MRI was performed. The primary outcome was HOMA-IR change. Results: One month after bariatric surgery and initiation of LCD, both treatments induced similar reductions in body weight (mean ± SD: −7.7±1.4 kg and −7.4±2.2 kg, respectively), adipose tissue volume (7%) and liver fat content (2% units). HOMA-IR, a main endpoint, was significantly reduced following OS (−26.3% [95% CI −49.5, −3.0], p=0.009) and non-significantly following LCD (−20.9% [95% CI −58.2, 16.5). For both groups, there were similar reductions in triglycerides and LDL-cholesterol. Fasting plasma glucose and insulin were also significantly reduced only following OS. There was an increase in glucose AUC in response to an OGTT in the OS group (by 20%) but not in the LCD group. During hyperinsulinaemia, only the OS group showed a significantly increased PET-derived glucose uptake rate in skeletal muscle but a reduced uptake in the heart and abdominal adipose tissue. Both liver and brain glucose uptake rates were unchanged after surgery or LCD. Whole-body glucose disposal and endogenous glucose production were not significantly affected. Conclusions/interpretation: The short-term metabolic effects seen 4 weeks after OS are not explained by loss of body fat alone. Thus OS, but not LCD, led to reductions in fasting plasma glucose and insulin resistance as well as to distinct changes in insulin-stimulated glucose fluxes to different tissues. Such effects may contribute to the prevention or reversal of type 2 diabetes following OS. Moreover, the full effects on whole-body insulin resistance and plasma glucose require a longer time than 4 weeks. Trial registration: ClinicalTrials.gov NCT02988011 Funding: This work was supported by AstraZeneca R&D, the Swedish Diabetes Foundation, the European Union’s Horizon Europe Research project PAS GRAS, the European Commission via the Marie Sklodowska Curie Innovative Training Network TREATMENT, EXODIAB, the Family Ernfors Foundation, the P.O. Zetterling Foundation, Novo Nordisk Foundation, the Agnes and Mac Rudberg Foundation and the Uppsala University Hospital ALF grants Graphical


  • Open Access

    A whole-body diffusion MRI normal atlas: development, evaluation and initial use

    by Therese Sjöholm, Sambit Tarai, Filip Malmberg, Robin Strand, Alexander Korenyushkin, Gunilla Enblad, Håkan Ahlström, Joel Kullberg

    Background: Statistical atlases can provide population-based descriptions of healthy volunteers and/or patients and can be used for region- and voxel-based analysis. This work aims to develop whole-body diffusion atlases of healthy volunteers scanned at 1.5T and 3T. Further aims include evaluating the atlases by establishing whole-body Apparent Diffusion Coefficient (ADC) values of healthy tissues and including healthy tissue deviations in an automated tumour segmentation task. Methods: Multi-station whole-body Diffusion Weighted Imaging (DWI) and water-fat Magnetic Resonance Imaging (MRI) of healthy volunteers (n = 45) were acquired at 1.5T (n = 38) and/or 3T (n = 29), with test-retest imaging for five subjects per scanner. Using deformable image registration, whole-body MRI data was registered and composed into normal atlases. Healthy tissue ADCmean was manually measured for ten tissues, with test-retest percentage Repeatability Coefficient (%RC), and effect of age, sex and scanner assessed. Voxel-wise whole-body analyses using the normal atlases were studied with ADC correlation analyses and an automated tumour segmentation task. For the latter, lymphoma patient MRI scans (n = 40) with and without information about healthy tissue deviations were entered into a 3D U-Net architecture. Results: Sex- and Body Mass Index (BMI)-stratified whole-body high b-value DWI and ADC normal atlases were created at 1.5T and 3T. %RC of healthy tissue ADCmean varied depending on tissue assessed (4–48% at 1.5T, 6–70% at 3T). Scanner differences in ADCmean were visualised in Bland-Altman analyses of dually scanned subjects. Sex differences were measurable for liver, muscle and bone at 1.5T, and muscle at 3T. Volume of Interest (VOI)-based multiple linear regression, and voxel-based correlations in normal atlas space, showed that age and ADC were negatively associated for liver and bone at 1.5T, and positively associated with brain tissue at 1.5T and 3T. Adding voxel-wise information about healthy tissue deviations in an automated tumour segmentation task gave numerical improvements in the segmentation metrics Dice score, sensitivity and precision. Conclusions: Whole-body DWI and ADC normal atlases were created at 1.5T and 3T, and applied in whole-body voxel-wise analyses.


  • Open Access

    Diet composition, nutrient substitutions and circulating fatty acids in relation to ectopic and visceral fat depots

    by Michael Fridén, Andrés Martínez Mora, Lars Lind, Ulf Risérus, Joel Kullberg, Fredrik Rosqvist