Table 1 Summary of existing methods and research gaps.

Sharma et al. (2025)⁷⁴

ML Intelligent

System Review

Multiple HD

Datasets

–

No spectrogram-based,

no percussion/palpation classification

Dhanka & Maini (2024)⁷⁵

HyOPTXGBoost

+ HyOPTRF with Optuna

Statlog HD

Dataset

96.30%

Single-organ focus; lacks

generalization to other signals

Dhanka et al. (2023)⁷⁶

XGBoost + Logistic Regression

Cleveland

Dataset

91.85%

Binary classification

only; no multi-region signal processing

Viji et al. (2024)⁷⁷

STO-IWGAN Hybrid

Energy System

Dataset

95%

Non-medical domain;

optimization only

Dhanka & Maini (2025)⁷⁸

Hybrid XGBoost

with Optuna

Cardiovascular

Dataset

95.45%

Focus on outlier handling; not

spectrogram or ensemble-based

This study

CNN + SVM + RF

Ensemble

Spectrograms from iApp

(Percussion & Palpation)

95.4%

First to classify 8 anatomical

regions from spectrograms