Evaluating Temperature Scaling Calibration Effectiveness for CNNs under Varying Noise Levels in Brain Tumour Detection

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chanda, Ankur, Choudhury, Kushan, Roy, Shubhrodeep, Biswas, Shubhajit, Kuiry, Somenath
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914064184115200
author Chanda, Ankur
Choudhury, Kushan
Roy, Shubhrodeep
Biswas, Shubhajit
Kuiry, Somenath
author_facet Chanda, Ankur
Choudhury, Kushan
Roy, Shubhrodeep
Biswas, Shubhajit
Kuiry, Somenath
contents Precise confidence estimation in deep learning is vital for high-stakes fields like medical imaging, where overconfident misclassifications can have serious consequences. This work evaluates the effectiveness of Temperature Scaling (TS), a post-hoc calibration technique, in improving the reliability of convolutional neural networks (CNNs) for brain tumor classification. We develop a custom CNN and train it on a merged brain MRI dataset. To simulate real-world uncertainty, five types of image noise are introduced: Gaussian, Poisson, Salt & Pepper, Speckle, and Uniform. Model performance is evaluated using precision, recall, F1-score, accuracy, negative log-likelihood (NLL), and expected calibration error (ECE), both before and after calibration. Results demonstrate that TS significantly reduces ECE and NLL under all noise conditions without degrading classification accuracy. This underscores TS as an effective and computationally efficient approach to enhance decision confidence of medical AI systems, hence making model outputs more reliable in noisy or uncertain settings.
format Preprint
id arxiv_https___arxiv_org_abs_2509_24951
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Evaluating Temperature Scaling Calibration Effectiveness for CNNs under Varying Noise Levels in Brain Tumour Detection
Chanda, Ankur
Choudhury, Kushan
Roy, Shubhrodeep
Biswas, Shubhajit
Kuiry, Somenath
Computer Vision and Pattern Recognition
Precise confidence estimation in deep learning is vital for high-stakes fields like medical imaging, where overconfident misclassifications can have serious consequences. This work evaluates the effectiveness of Temperature Scaling (TS), a post-hoc calibration technique, in improving the reliability of convolutional neural networks (CNNs) for brain tumor classification. We develop a custom CNN and train it on a merged brain MRI dataset. To simulate real-world uncertainty, five types of image noise are introduced: Gaussian, Poisson, Salt & Pepper, Speckle, and Uniform. Model performance is evaluated using precision, recall, F1-score, accuracy, negative log-likelihood (NLL), and expected calibration error (ECE), both before and after calibration. Results demonstrate that TS significantly reduces ECE and NLL under all noise conditions without degrading classification accuracy. This underscores TS as an effective and computationally efficient approach to enhance decision confidence of medical AI systems, hence making model outputs more reliable in noisy or uncertain settings.
title Evaluating Temperature Scaling Calibration Effectiveness for CNNs under Varying Noise Levels in Brain Tumour Detection
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.24951