(308 days)
The Edison System is indicated for the non-invasive destruction of liver tumors, including unresectable liver tumors, using a non-thermal, mechanical process of focused ultrasound.
The HistoSonics Edison™ System (the "System") provides users with a means to identify, target and destroy tissue non-invasively via non-ionizing, non-thermal, mechanical process of focused ultrasound.
By delivering high amplitude, very short (microsecond), focused ultrasound pulses the device can induce acoustic cavitation at a known focal location in the form of a "bubble cloud". The bubble cloud is a cluster of microbubbles that form, rapidly expand and rapidly collapse, imparting stress and strain on target soft tissue. After a number of pulses, soft tissue within the bubble cloud is mechanically destroyed, resulting in a homogenous acellular lysate with limited to no recognizable cellular structures. The bubble cloud appears hypoechoic (bright) when viewed on diagnostic (B-mode) ultrasound. Additionally, the bubble cloud is detectable audibly.
The System is comprised of reusable medical equipment and disposable, single-patient use components. Reusable portions of the System include the Treatment Cart and a Support Arm and Frame that is used to contain ultrasound medium (degassed water) that acoustically couples the System to the patient. Disposable aspects of the System include a clear, acoustically transparent and deformable membrane that holds the ultrasound medium as well as various non-active accessories used during patient setup.
The System is functionally integrated with a GE LOGIQ E10s ultrasound system (K211524), which is provided by HistoSonics with each Edison System. The LOGIO E10s is provided with a preset configuration designed to optimize viewing of the bubble cloud. This configuration falls within the available performance parameters of the LOGIQ E10s covered by K211524.
The Treatment Cart is mobile and contains all hardware and software components necessary to localize, plan and deliver treatments. The Treatment Cart includes a large touchscreen user interface and control panel, a high voltage power supply, integrated amplifier circuitry, waveform generator boards, Control PC, and Treatment Arm/Micropositioning System with connected Treatment Head.
The System user interface guides the user step by step through the required workflow including Patient Preparation, Localize, Plan and Treat.
The treating physician uses the user interface to assess ultrasound images to localize the targeted tissue and define a Planned Treatment Volume (PTV) comprised of a target contour and margin contour. During targeting, the diagnostic ultrasound probe can be extended up to 5 cm in the Z axis to reduce the offset of the ultrasound medium, thereby improving the targeting process versus a non-extended probe. An optional workflow enables the physician to view DICOM images (MRI, CT and PET) adjacent to the live ultrasound to aid the target identification process. Additionally, an image fusion function is an optional workflow that allows physician to fuse the live ultrasound image from the GE LOGIQ E10s onto the previously obtained DICOM images.
The Treatment Arm/Micropositioning System is comprised of a six degree of freedom (6 DOF) dual encoded robotic arm and is used to direct the movement of the Treatment Head. The Treatment Arm provides mechanical support to the Treatment Head (containing the Therapy Transducer and the coaxially aligned GE LOGIQ E10s Diagnostic Imaging Probe) and allows gross and fine positioning prior to initiating therapy. The Treatment Head is available in two configurations with different maximum treatment depths to provide physicians with options based on target anatomy. Both Treatment Heads are supplied with the System.
Electronic signals from the Treatment Cart are applied to the Therapy Transducer to create a bubble cloud at a known focal point. Note that the bubble cloud location is fixed relative to the position of the Treatment Head. The System uses the software-controlled Micropositioning System to move the Treatment Head, and the resulting bubble cloud, through a programmed treatment pathway to enable treatment delivery at all bubble cloud locations included as part of the PTV.
Here's a breakdown of the acceptance criteria and the study proving the Edison System meets them, based on the provided text:
Acceptance Criteria and Device Performance
1. Table of Acceptance Criteria and Reported Device Performance
| Acceptance Criteria (Performance Goal) | Reported Device Performance | Comments |
|---|---|---|
| Primary Effectiveness Endpoint: Complete tumor ablation rate of 70% or higher (Technical Success), as determined by CT/MRI imaging obtained ≤36 hours after initial treatment. | 95.5% [95% CI 83.72 - 100%] of the lesions achieved technical success within 36 hours of the procedure. | This criterion was met. The observed rate significantly exceeded the performance goal. |
| Primary Safety Endpoint: Rate of 25% or less of index-procedure device-related major complications (CTCAE Grade 3 or higher) at 30-days. | 6.8% [95% CI 2.35 - 18.23%] of subjects had a reported procedure-related major complication (CTCAE ≥3) within 30 days post-procedure. | This criterion was met. The observed rate was well below the performance goal. |
| Secondary Effectiveness Endpoint: Technique efficacy defined as the lack of a nodular or mass-like area of enhancement within or along the edge of the treatment volume at 30-days post-procedure. | 83.3% [95% CI 67.65-92.11%] | No specific performance goal was designated for this secondary endpoint, but the result is provided. |
| Secondary Safety Endpoint: All adverse events reported within 30 days post-index procedure. | 101 AEs reported within 30 days, with 43.6% device-related. Most common non-serious ADEs: abdominal pain (22.7%), procedural pain (22.7%), pyrexia (15.9%). Serious AEs: pleuritic pain (2%), procedural pain (2%), sepsis (2%), liver failure (2%). | This was a descriptive endpoint, and no specific performance goal was set, but the rates were deemed consistent with established risk analyses and literature. |
Study Details
2. Sample Size and Data Provenance
- Test Set Sample Size:
- Effectiveness Analysis: 40 evaluable patients, corresponding to 44 total tumors/lesions.
- Safety Analysis: 44 subjects (all enrolled subjects).
- Data Provenance: Multicenter, non-randomized, prospective single-arm study. Data was pooled from 8 US sites (21 subjects) and 6 OUS (Outside US) sites in Europe (23 subjects).
3. Number of Experts and Qualifications for Ground Truth Establishment
- The text explicitly mentions "a third-party laboratory" and "Core Laboratory Adjudicated" for both primary and secondary effectiveness endpoints (technical success and technique efficacy), and "independent Clinical Events Committee (CEC) adjudicated" for safety endpoints.
- The exact number and specific qualifications of the experts within these "third-party laboratories" or "Clinical Events Committee" are not specified in the provided text.
4. Adjudication Method for the Test Set
- For effectiveness endpoints (Technical Success and Technique Efficacy): "Core Laboratory Adjudicated." This implies a centralized review by imaging experts. The specific method (e.g., 2+1, 3+1, etc., for consensus in case of disagreement) is not detailed.
- For safety endpoints (Major Complications and All Adverse Events): "Clinical Events Committee (CEC) Adjudicated." This indicates independent review of reported adverse events. The specific adjudication mechanism by the CEC is not detailed.
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
- No, an MRMC comparative effectiveness study was not done.
- The study was a single-arm study evaluating the device's performance against pre-specified performance goals, not a comparative study against other human-in-the-loop or unassisted human reader performance. Therefore, an effect size of human readers improving with AI vs. without AI assistance is not applicable to this study design.
6. Standalone Performance (Algorithm Only)
- Not applicable/Not evaluated as a standalone algorithm. The Edison System is a medical device that includes software for treatment planning and delivery, but it is not an AI algorithm intended for standalone diagnostic or clinical decision-making. Its performance is intrinsically linked to the human operator using the system's integrated features for localization, planning, and treatment. The "software" section refers to software validation and verification for a "Major" level of concern, implying robust engineering practices, but not a standalone AI performance evaluation.
7. Type of Ground Truth Used
- Effectiveness: Imaging (CT/MRI) interpretation by a third-party/Core Laboratory. The clinical study leveraged corroboration with histopathology results from preclinical "acute animal studies" to support effectiveness, in lieu of collecting post-treatment liver biopsy samples from human subjects. This suggests that the primary clinical effectiveness ground truth was based on imaging, supported by preclinical pathology.
- Safety: Clinical event adjudication by an independent Clinical Events Committee (CEC) based on reported adverse events, coupled with objective measures like lab tests and follow-up imaging.
8. Sample Size for the Training Set
- The provided document does not detail any specific training set or its sample size for the clinical study. This is expected, as the device is not presented as a machine learning model requiring a distinct clinical training dataset. The "software" section describes standard software development practices (hazard analysis, V&V testing), which would involve internal testing and validation, but not a "training set" in the machine learning sense for a clinical trial.
9. How the Ground Truth for the Training Set Was Established
- As there's no mention of a clinical "training set" or a machine learning component learning from clinical data for its primary function, the concept of ground truth establishment for a training set in this context is not applicable to the clinical study description.
- For the device's inherent functional software (e.g., image overlay, treatment planning tools), ground truth would be established through engineering validation, comparison to known anatomical models, and physical measurements (bench testing), which are described in the non-clinical performance sections (e.g., "Ultrasound Imaging Accuracy," "Therapy output effect... consistent across the planned treatment volume boundaries").
N/A