Accepted for/Published in: JMIR Medical Informatics
Date Submitted: Feb 6, 2026
Date Accepted: Oct 20, 2026
Development and Application of Deep Learning Models for Casefinding from Pathology Reports in a Hospital-Based Cancer Registry: Algorithm Development and Validation
ABSTRACT
Background:
Few studies have investigated automated casefinding in hospital-based cancer registries, despite their crucial role in supporting population-based surveillance. In these settings, data completeness is often compromised by referral patterns, yet the influence of weakly supervised learning strategies on model performance remains underexplored. This study evaluated the effects of different supervision approaches in training deep learning models for cancer casefinding.
Objective:
To evaluate the effects of different supervision approaches (inaccurate, inexact, and report-level labeling) on deep learning model performance for cancer casefinding in a hospital-based registry setting, and to assess model generalizability across patient-level and report-level classification tasks.
Methods:
Hospital-based cancer registry and pathology report data from Songklanagarind Hospital (2005-2023) were used for model development and evaluation. Three labeling approaches were implemented: inaccurate (all reports from cancer patients labeled positive), inexact (concatenated reports from cancer patients labeled positive), and report-level (only registrar-annotated tumor-related reports labeled positive). Model performance was evaluated on max-pooled, concatenated, and individual report classification tasks using SWEM (Simple Word Embedding Model), MLP (Multilayer perceptron), RNN (Recurrent Neural Network), CNN (Convolutional Neural Network) and HiSAN (Hierarchical Self-Attention Network) architectures
Results:
CNN models trained with inexact labeling (without temporal window constraints) achieved the highest performance across all evaluation tasks. While report-level labeling (strong supervision) yielded superior accuracy for individual reports, it suffered substantial degradation in patient-level evaluations. This suggests that models trained on curated data overfitted to primary tumor patterns, failing to recognize secondary evidence (e.g., metastatic or surveillance findings). Conversely, weakly supervised approaches enabled the detection of cancer patients via diverse report types. Error analysis revealed that false negatives were primarily driven by insufficient institutional data—specifically, referred patients lacking primary diagnostic reports—rather than algorithmic failure.
Conclusions:
Weakly supervised labeling approaches, particularly inexact labeling, yield more generalizable casefinding models than resource-intensive human annotation. CNN architectures demonstrated robust performance and feasibility for hospital-based implementation, effectively handling the noisy data inherent to referral centers.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.