Medical News Observer

Key Points:

• DeepCatch auto-segments anatomical structures from CT images and calculates their volumes and proportions.
• The software is intended to be used in conjunction with professional clinical judgment.
• It has demonstrated substantial equivalence to predicate devices, ensuring its safety and effectiveness.

DeepCatch is medical image processing software designed to analyze CT images and auto-segment anatomical structures, including skin, bone, muscle, visceral fat, subcutaneous fat, internal organs, and the central nervous system. It calculates and provides the volume and proportions of these structures along with relevant 3D models. The device utilizes whole-body CT images and is designed to complement professional clinical judgment.

Manufacturer Information

Medical IP Co., Ltd develops this medical image processing software.

Regulatory Approval Information

DeepCatch received its regulatory approval under the 510(k) premarket notification process, confirming that it is substantially equivalent to legally marketed predicate devices. The device is subject to general control provisions of the Federal Food, Drug, and Cosmetic Act, which includes requirements for annual registration, device listing, good manufacturing practice, labeling, and prohibitions against misbranding and adulteration. DeepCatch received FDA clearance based on its substantial equivalence to predicate devices MEDIP PRO (K191026) and SYNAPSE 3D LUNG AND ABDOMEN ANALYSIS (K130542).

Indications

DeepCatch analyzes CT images to auto-segment anatomical structures, calculate their volume and proportions, and provide relevant 3D models. This helps obtain accurate values for the volume and proportion of anatomical structures through the secondary utilization of CT images obtained for various medical purposes. The device is designed for professional use in conjunction with clinical judgment.

Contraindications

The device must be part of a comprehensive diagnostic and treatment planning process under professional supervision. It is not intended for standalone diagnostic purposes or as a replacement for professional medical evaluation and judgment.

Study Results

Non-Clinical Test Summary

DeepCatch was validated for its intended use and demonstrated compliance with voluntary electrical safety and electromagnetic compatibility standards. The software, containing a moderate level of concern, was verified and validated according to FDA guidance.

Performance Test Results:

• Internal datasets (n=100) and external datasets (n=580) showed a Dice Similarity Coefficient (DSC) mean of greater than or equal to 90%.
• Volume, area, and ratio differences between ground truth (GT) and measurement results were within acceptable limits, demonstrating less than 10% for volume and area, less than 1% for ratios, and less than 5% for body circumference.

Comparative Performance Tests:

• Comparison with MEDIP PRO (n=100) and Synapse 3D (n=100) showed no significant difference in DSC values, indicating comparable performance. DeepCatch showed better performance for muscle segmentation and specific area measurements compared to Synapse 3D.

U.S.-Based Performance Test:

• Testing on U.S.-based datasets (n=167) confirmed no performance bias towards a particular population, with DSC means of over 90% for all anatomical structures and measurement errors within acceptable limits.

Limitations

DeepCatch is not intended for primary diagnostic use and must be used with professional clinical judgment. The software’s performance is contingent on the quality of input CT images, and the segmentation accuracy may vary based on individual patient anatomy and image quality.

Conclusion

DeepCatch is a reliable and effective medical image processing software that is substantially equivalent to its predicate devices in terms of safety and effectiveness. Its robust performance in segmenting anatomical structures and calculating their volumes and proportions makes it a valuable tool for clinical applications in radiology.

Future Implications

Future enhancements of DeepCatch may include expanding its capabilities to integrate with other diagnostic tools and improving its algorithms for greater accuracy and efficiency. Additionally, exploring its application in different imaging modalities and incorporating advanced AI and machine learning techniques could enhance its utility in clinical practice. The continued development and refinement of DeepCatch will likely contribute to improved patient outcomes and more precise diagnostic and treatment planning in the medical field.

Device Details

• Trade/Device Name: DeepCatch
• Regulation Number: 21 CFR 892.2050
• Device Classification Name: Medical Image Management and Processing System
• Regulatory Class: Class II
• Product Code: QIH
• Device Class: II
• Review Panel: Radiology

Manufacturer Details

• Name of Manufacturer: Medical IP Co., Ltd.
• Address: SNUH Cancer Research Center 806, 101 Daehak-ro, Jongno-gu, Seoul, KR 0308
• Contact Name: Jun-Sik Yoon
• Telephone No.: +82 10-8277-2909
• Email Address: jsyoon@medicalip.com
• Registration No.: 3016579137

Reference

“DeepCatch K223556 510(k) Premarket Notification.” Accessed Online. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm?ID=K223556.