Kaustubh Chakradeo

Kaustubh Chakradeo

Auto-generated profile, based on publicly available information

Kaustubh Chakradeo

University of Copenhagen

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
25 January 2024 in IEEE Access
Verified
Journal Article
Open access
DJDiana Susan Joseph
Diana Susan Joseph
PPPranav M. Pawar
Pranav M. Pawar
KCKaustubh Chakradeo
Kaustubh Chakradeo
Agriculture plays a significant role in meeting food needs and providing food security for the increasingly growing global population, which has increased by 0.88% since 2022. Plant diseases can reduce food production and affect food security. Worldwide crop loss due to plant disease is estimated to be around 14.1%. The lack of proper identification of plant disease at the early stages of infection can result in inappropriate disease control measures. Therefore, the automatic identification and diagnosis of plant diseases are highly recommended. Lack of availability of large amounts of data that are not processed to a large extent is one of the main challenges in plant disease diagnosis. In the current manuscript, we developed datasets for food grains specifically for rice, wheat, and maize to address the identified challenges. The developed datasets consider the common diseases (two bacterial diseases and two fungal diseases of rice, four fungal diseases of maize, and four fungal diseases of wheat) that affect crop yields and cause damage to the whole plant. The datasets developed were applied to eight fine-tuned deep learning models with the same training hyperparameters. The experimental results based on eight fine-tuned deep learning models show that, while recognizing maize leaf diseases, the models Xception and MobileNet performed best with a testing accuracy of 0.9580 and 0.9464 respectively. Similarly, while recognizing the wheat leaf diseases, the models MobileNetV2 and MobileNet performed best with a testing accuracy of 0.9632 and 0.9628 respectively. The Xception and Inception V3 models performed best, with a testing accuracy of 0.9728 and 0.9620, respectively, for recognizing rice leaf diseases. The research also proposes a new convolutional neural network (CNN) model trained from scratch on all three food grain datasets developed. The proposed model performs well and shows a testing accuracy of 0.9704, 0.9706, and 0.9808 respectively on the maize, rice, and wheat datasets.

26 April 2025 in Philosophy & Technology
Verified
Journal Article
Open access
PDF file
IWIsaac Wagner
Isaac Wagner
KCKaustubh Chakradeo
Kaustubh Chakradeo
We apply the concept of robustness from the philosophy of science to human–AI collaboration in diagnostic radiology, introducing diagnostic complementarity as a way to understand how radiologists and AI systems can productively work together in the context of double-reading. Diagnostic complementarity refers to the idea that two readers (e.g. radiologist and AI) have different diagnostic strengths and limitations, such that their combined performance exceeds that of either working alone. We argue that state-of-the-art AI diagnostic systems—convolutional neural networks (CNNs)—realize diagnostic complementarity with radiologists, and thus can serve as effective second readers for radiologist first readers in double-reading. In the course of making our argument, we clarify diagnostic complementarity and its epistemic benefits, explain how it can be discovered through theoretical and empirical means, and show how our points can be used to evaluate not only the utility of CNNs as second readers, but also of future, more advanced AI systems.

18 September 2024 in medRxiv
Verified
Preprint
Open access
PDF file
DLDaniel J Laydon
Daniel J Laydon
DSDavid L Smith
David L Smith
KCKaustubh Chakradeo
Kaustubh Chakradeo
MKMark P Khurana
Mark P Khurana
JOJaffer Okiring
Jaffer Okiring
SBSamir Bhatt
Samir Bhatt
Mosquitoes display marked sensitivity to weather, particularly temperature and precipitation. Therefore, climate change is expected to profoundly affect malaria epidemiology in its transmission, spatiotemporal distribution and consequent disease burden. However, malaria transmission is also complicated by other factors (e.g. urbanization, socioeconomic development, genetics, drug resistance) which together constitute a highly complex, dynamical system, where the influence of any single factor can be masked by others. In this study, we aim to re-evaluate the evidence underlying the widespread belief that climate change will increase worldwide malaria transmission. We review two broad types of study that have contributed to this evidence-base: i) studies that project changes in transmission due to inferred relationships between environment and mosquito entomology, and ii) regression-based studies that look for associations between environmental variables and malaria prevalence. We then employ a simple statistical model to show that environmental variables alone do not account for the observed spatiotemporal variation in malaria prevalence. Our review raises concerns regarding the relationship between climate change and malaria. We find that, while climate change’s effect on malaria is highly plausible, empirical evidence is much less certain. Future research on climate change and malaria must become integrated into malaria control programs, and understood in context as one factor among many. Our work outlines gaps in modelling that we believe are priorities for future research.

25 November 2024 in arXiv
Verified
Preprint
Open access
PDF file
KCKaustubh Chakradeo
Kaustubh Chakradeo
PNPernille Nielsen
Pernille Nielsen
LGLise Mette Rahbek Gjerdrum
Lise Mette Rahbek Gjerdrum
GHGry Sahl Hansen
Gry Sahl Hansen
DDDavid A Duchêne
David A Duchêne
SBSamir Bhatt
Samir Bhatt
As global life expectancy increases, so does the burden of chronic diseases, yet individuals exhibit considerable variability in the rate at which they age. Identifying biomarkers that distinguish fast from slow ageing is crucial for understanding the biology of ageing, enabling early disease detection, and improving prevention strategies. Using contrastive deep learning, we show that skin biopsy images alone are sufficient to determine an individual's age. We then use visual features in histopathology slides of the skin biopsies to construct a novel biomarker of ageing. By linking with comprehensive health registers in Denmark, we demonstrate that visual features in histopathology slides of skin biopsies predict mortality and the prevalence of chronic age-related diseases. Our work highlights how routinely collected health data can provide additional value when used together with deep learning, by creating a new biomarker for ageing which can be actively used to determine mortality over time.

16 May 2025 in medRxiv
Verified
Preprint
Open access
PDF file
KCKaustubh Chakradeo
Kaustubh Chakradeo
AKAlexandros Katsiferis
Alexandros Katsiferis
NSNeil Scheidwasser
Neil Scheidwasser
IHIwona Hawryluk
Iwona Hawryluk
KBKatherine E Battle
Katherine E Battle
SBSamir Bhatt
Samir Bhatt
Malaria remains a leading cause of morbidity and mortality worldwide, with sub-Saharan Africa bearing the highest burden. Stalled progress under an inadequate budget and the expectation that funding would be further constrained [1, 2] call for an evaluation of a worst-case scenario to appreciate malaria transmission potential in the absence of interventions. We examine a scenario in which all funding for interventions ceases for several years, allowing population immunity to adjust to a new equilibrium amid a surge in transmission, while economic development, climate, and demographics remain static, which we define as the transmission niche or “baseline” prevalence. The baseline proposed here is a set of stratified PfPr 2-10 observations to select only samples with low intervention histories, preserving diverse epidemiological context. We further developed a bespoke contrastive deep learning architecture applied to 30-meter resolution images from Landsat 8 satellites, generating a detailed vector covariate set based on observable land status, which helps produce state-of-the-art malaria estimates, an improvement over all previous covariate-based models. These features were integrated into a Bayesian MCMC model with regularisation to estimate baseline PfPr 2-10 and R 0 to generate a map of malaria incidence over all of Sub-Saharan Africa. Using population data from the World Malaria Report 2024, our model estimates 422 (275–582) million cases across sub-Saharan Africa for this worst-case baseline scenario, reflecting a 131% increase compared to previous business-as-usual baseline scenario estimates [3]. Our analyses highlight the effect of long-term benefits of two decades of investments for malaria control and the critical need for sustained intervention efforts and informed policy-making to mitigate potential resurgences in malaria transmission.

18 February 2026 in medRxiv
Verified
Preprint
Open access
PDF file
AFAlice Fracchia
Alice Fracchia
JRJonas Julius Rudbaek
Jonas Julius Rudbaek
KCKaustubh Chakradeo
Kaustubh Chakradeo
TJTine Jess
Tine Jess
FOFilip Ottosson
Filip Ottosson
ASAleksejs Sazonovs
Aleksejs Sazonovs
Background: Gestational exposures may contribute to the newborn’s lifetime risk of inflammatory bowel disease (IBD). While gestational influences are associated with IBD onset, the causality and confounding of such exposures are difficult to ascertain. The neonatal metabolome provides a metabolic snapshot of gestational influences. Objective: We tested the neonatal metabolome’s ability to predict future IBD, to assess whether gestational exposures are reflected in early molecular precursors of the disease. Methods: We profiled dried blood spots from 520 newborns who later developed IBD and matched controls using high-resolution untargeted mass spectrometry metabolomics (1,350 QC-passing metabolites). Genotyping was available for 1,009 of these individuals. PERMANOVA confirmed assay sensitivity to gestational exposures, gradient boosting was used for prediction. Results: The neonatal metabolome significantly captured maternal smoking, birth weight, and gestational age (p < 0.001), but explained minimal variance in IBD status (R 2 = 0.09%, p = 0.390) and showed no predictive power for IBD (AUC = 0.51, 95% CI 0.50–0.52, p = 0.585). Stratifying by disease subtype and age of onset did not improve performance. In contrast, genetic risk scores were modestly predictive (CD: AUC = 0.64, p < 5.11×10⁻ 14 ; UC: AUC = 0.63, p < 7.65×10⁻¹²), but uncorrelated with neonatal metabolomic profiles (CD: p = 0.650; UC: p = 0.970), suggesting a later-age effect. Conclusions: Using a large, comprehensively profiled cohort, we demonstrate that neonatal metabolomic profiles sensitively capture gestational signatures, but not the overall future IBD risk. Our findings suggest that most IBD risk accumulates later in life, beyond gestational molecular imprints.