Abstract:Universal lesion detection has great value for clinical practice as it aims to detect various types of lesions in multiple organs on medical images. Deep learning methods have shown promising results, but demanding large volumes of annotated data for training. However, annotating medical images is costly and requires specialized knowledge. The diverse forms and contrasts of objects in medical images make fully annotation even more challenging, resulting in incomplete annotations. Directly training ULD detectors on such datasets can yield suboptimal results. Pseudo-label-based methods examine the training data and mine unlabelled objects for retraining, which have shown to be effective to tackle this issue. Presently, top-performing methods rely on a dynamic label-mining mechanism, operating at the mini-batch level. However, the model's performance varies at different iterations, leading to inconsistencies in the quality of the mined labels and limits their performance enhancement. Inspired by the observation that deep models learn concepts with increasing complexity, we introduce an innovative exploratory training to assess the reliability of mined lesions over time. Specifically, we introduce a teacher-student detection model as basis, where the teacher's predictions are combined with incomplete annotations to train the student. Additionally, we design a prediction bank to record high-confidence predictions. Each sample is trained several times, allowing us to get a sequence of records for each sample. If a prediction consistently appears in the record sequence, it is likely to be a true object, otherwise it may just a noise. This serves as a crucial criterion for selecting reliable mined lesions for retraining. Our experimental results substantiate that the proposed framework surpasses state-of-the-art methods on two medical image datasets, demonstrating its superior performance.

Embracing Annotation Efficient Learning (AEL) for Digital Pathology and Natural Images

The HABP2 G534E variant is an unlikely cause of familial non-medullary thyroid cancer.

Active Image Synthesis for Efficient Labeling

AL-Annotator: an Active Learning-based Cervical Cell Annotation System

Towards Good Practices for Efficiently Annotating Large-Scale Image Classification Datasets

The ALCHEmist: Automated Labeling 500x CHEaper Than LLM Data Annotators

Label Smarter, Not Harder: CleverLabel for Faster Annotation of Ambiguous Image Classification with Higher Quality

Democratizing Pathological Image Segmentation with Lay Annotators via Molecular-empowered Learning

Annotation-efficient deep learning for automatic medical image segmentation

Tackling the Incomplete Annotation Issue in Universal Lesion Detection Task By Exploratory Training

Label-Efficient Deep Learning in Medical Image Analysis: Challenges and Future Directions

Modular Deep Active Learning Framework for Image Annotation: A Technical Report for the Ophthalmo-AI Project

No Need to Sacrifice Data Quality for Quantity: Crowd-Informed Machine Annotation for Cost-Effective Understanding of Visual Data

Iterative annotation to ease neural network training: Specialized machine learning in medical image analysis

Annotation Cost Efficient Active Learning for Content Based Image Retrieval

An Adaptive Query Approach for Extracting Medical Images for Disease Detection Applications

Annotating Ambiguous Images: General Annotation Strategy for High-Quality Data with Real-World Biomedical Validation

Label Assistant: A Workflow for Assisted Data Annotation in Image Segmentation Tasks

An End-to-End Framework For Universal Lesion Detection With Missing Annotations

Improvement of AMPLICOR Human Immunodeficiency Virus Type 1 Viral Load Test (Version 1.5) by Addition of a Coprecipitant during the RNA Isolation Step

Focused Active Learning for Histopathological Image Classification