Interactive Medical Image Labeling Tool to Construct a Robust Convolutional Neural Network Training Data Set: Development and Validation Study
Bibliographic record
Abstract
BACKGROUND: Skin ulcers are an important cause of morbidity and mortality everywhere in the world and occur due to several causes, including diabetes mellitus, peripheral neuropathy, immobility, pressure, arteriosclerosis, infections, and venous insufficiency. Ulcers are lesions that fail to undergo an orderly healing process and produce functional and anatomical integrity in the expected time. In most cases, the methods of analysis used nowadays are rudimentary, which leads to errors and the use of invasive and uncomfortable techniques on patients. There are many studies that use a convolutional neural network to classify the different tissues in a wound. To obtain good results, the network must be trained with a correctly labeled data set by an expert in wound assessment. Typically, it is difficult to label pixel by pixel using a professional photo editor software, as this requires extensive time and effort from a health professional. OBJECTIVE: The aim of this paper is to implement a new, fast, and accurate method of labeling wound samples for training a neural network to classify different tissues. METHODS: We developed a support tool and evaluated its accuracy and reliability. We also compared the support tool classification with a digital gold standard (labeling the data with an image editing software). RESULTS: The obtained comparison between the gold standard and the proposed method was 0.9789 for background, 0.9842 for intact skin, 0.8426 for granulation tissue, 0.9309 for slough, and 0.9871 for necrotic. The obtained speed on average was 2.6, compared to that of an advanced image editing user. CONCLUSIONS: This method increases tagging speed on average compared to an advanced image editing user. This increase is greater with untrained users. The samples obtained with the new system are indistinguishable from the samples made with the gold standard.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.007 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".