Development of a Whole Slide Imaging System on Smartphones and Evaluation With Frozen Section Samples
Bibliographic record
Abstract
BACKGROUND: The aim was to develop scalable Whole Slide Imaging (sWSI), a WSI system based on mainstream smartphones coupled with regular optical microscopes. This ultra-low-cost solution should offer diagnostic-ready imaging quality on par with standalone scanners, supporting both oil and dry objective lenses of different magnifications, and reasonably high throughput. These performance metrics should be evaluated by expert pathologists and match those of high-end scanners. OBJECTIVE: The aim was to develop scalable Whole Slide Imaging (sWSI), a whole slide imaging system based on smartphones coupled with optical microscopes. This ultra-low-cost solution should offer diagnostic-ready imaging quality on par with standalone scanners, supporting both oil and dry object lens of different magnification. All performance metrics should be evaluated by expert pathologists and match those of high-end scanners. METHODS: In the sWSI design, the digitization process is split asynchronously between light-weight clients on smartphones and powerful cloud servers. The client apps automatically capture FoVs at up to 12-megapixel resolution and process them in real-time to track the operation of users, then give instant feedback of guidance. The servers first restitch each pair of FoVs, then automatically correct the unknown nonlinear distortion introduced by the lens of the smartphone on the fly, based on pair-wise stitching, before finally combining all FoVs into one gigapixel VS for each scan. These VSs can be viewed using Internet browsers anywhere. In the evaluation experiment, 100 frozen section slides from patients randomly selected among in-patients of the participating hospital were scanned by both a high-end Leica scanner and sWSI. All VSs were examined by senior pathologists whose diagnoses were compared against those made using optical microscopy as ground truth to evaluate the image quality. RESULTS: The sWSI system is developed for both Android and iPhone smartphones and is currently being offered to the public. The image quality is reliable and throughput is approximately 1 FoV per second, yielding a 15-by-15 mm slide under 20X object lens in approximately 30-35 minutes, with little training required for the operator. The expected cost for setup is approximately US $100 and scanning each slide costs between US $1 and $10, making sWSI highly cost-effective for infrequent or low-throughput usage. In the clinical evaluation of sample-wise diagnostic reliability, average accuracy scores achieved by sWSI-scan-based diagnoses were as follows: 0.78 for breast, 0.88 for uterine corpus, 0.68 for thyroid, and 0.50 for lung samples. The respective low-sensitivity rates were 0.05, 0.05, 0.13, and 0.25 while the respective low-specificity rates were 0.18, 0.08, 0.20, and 0.25. The participating pathologists agreed that the overall quality of sWSI was generally on par with that produced by high-end scanners, and did not affect diagnosis in most cases. Pathologists confirmed that sWSI is reliable enough for standard diagnoses of most tissue categories, while it can be used for quick screening of difficult cases. CONCLUSIONS: As an ultra-low-cost alternative to whole slide scanners, diagnosis-ready VS quality and robustness for commercial usage is achieved in the sWSI solution. Operated on main-stream smartphones installed on normal optical microscopes, sWSI readily offers affordable and reliable WSI to resource-limited or infrequent clinical users.
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 | 0.002 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.006 | 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".