Concepedia

Code-free deep learning for multi-modality medical image classification

Edward Korot, Zeyu Guan, Daniel Ferraz, Siegfried K. Wagner, Gongyu Zhang, Xiaoxuan Liu, Livia Faes, Nikolas Pontikos, Samuel G. Finlayson, Hagar Khalid,

Nature Machine Intelligence · 2021 · 169 citations · 36 references

DOI Full text

Open access

Concepts

ComputingComputer ScienceHealth InformaticsBiomedical ImagingDiagnostic ImagingMachine LearningComputer-aided DiagnosisDeep LearningImage ClassificationMedical Image AnalysisMedical ImagingClassification ModelsDataset CurationDeep Feature LearningDeep Learning-based ClassificationMachine Learning For HealthcareMedical Image ClassificationResearch Data Curation

TL;DR

Large technology companies have created code‑free cloud platforms that enable researchers and clinicians without coding experience to develop deep‑learning models. The study evaluates the performance and feature sets of six such platforms by training image‑classification models on four representative medical imaging datasets. The authors compared the platforms by building classification models on four cross‑sectional and en‑face imaging datasets. The platforms achieved F1 scores between 72.0 and 93.9, with optical coherence tomography consistently yielding higher performance, indicating potential applications in dataset curation, mobile edge models for low‑connectivity regions, and as baselines for iterative deep‑learning development.

Abstract

Abstract A number of large technology companies have created code-free cloud-based platforms that allow researchers and clinicians without coding experience to create deep learning algorithms. In this study, we comprehensively analyse the performance and featureset of six platforms, using four representative cross-sectional and en-face medical imaging datasets to create image classification models. The mean (s.d.) F1 scores across platforms for all model–dataset pairs were as follows: Amazon, 93.9 (5.4); Apple, 72.0 (13.6); Clarifai, 74.2 (7.1); Google, 92.0 (5.4); MedicMind, 90.7 (9.6); Microsoft, 88.6 (5.3). The platforms demonstrated uniformly higher classification performance with the optical coherence tomography modality. Potential use cases given proper validation include research dataset curation, mobile ‘edge models’ for regions without internet access, and baseline models against which to compare and iterate bespoke deep learning approaches.